diff --git a/clang-tools-extra/clang-tidy/google/AvoidCStyleCastsCheck.h b/clang-tools-extra/clang-tidy/google/AvoidCStyleCastsCheck.h
index 72f96d0945625..075b5f0ce450a 100644
--- a/clang-tools-extra/clang-tidy/google/AvoidCStyleCastsCheck.h
+++ b/clang-tools-extra/clang-tidy/google/AvoidCStyleCastsCheck.h
@@ -25,6 +25,9 @@ namespace readability {
 /// This check is similar to `-Wold-style-cast`, but it suggests automated fixes
 /// in some cases. The reported locations should not be different from the
 /// ones generated by `-Wold-style-cast`.
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-readability-casting.html
 class AvoidCStyleCastsCheck : public ClangTidyCheck {
 public:
   AvoidCStyleCastsCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/clang-tidy/google/AvoidUnderscoreInGoogletestNameCheck.h b/clang-tools-extra/clang-tidy/google/AvoidUnderscoreInGoogletestNameCheck.h
index 6a690f60da1c2..a9989ee51bcf6 100644
--- a/clang-tools-extra/clang-tidy/google/AvoidUnderscoreInGoogletestNameCheck.h
+++ b/clang-tools-extra/clang-tidy/google/AvoidUnderscoreInGoogletestNameCheck.h
@@ -18,6 +18,9 @@ namespace readability {
 
 // Check for underscores in the names of googletest tests, per
 // https://github.com/google/googletest/blob/master/googletest/docs/faq.md#why-should-test-suite-names-and-test-names-not-contain-underscore
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-readability-avoid-underscore-in-googletest-name.html
 class AvoidUnderscoreInGoogletestNameCheck : public ClangTidyCheck {
 public:
   using ClangTidyCheck::ClangTidyCheck;
diff --git a/clang-tools-extra/clang-tidy/google/DefaultArgumentsCheck.h b/clang-tools-extra/clang-tidy/google/DefaultArgumentsCheck.h
index 574965dc556a6..089d463ff9852 100644
--- a/clang-tools-extra/clang-tidy/google/DefaultArgumentsCheck.h
+++ b/clang-tools-extra/clang-tidy/google/DefaultArgumentsCheck.h
@@ -18,6 +18,9 @@ namespace google {
 /// Checks that default parameters are not given for virtual methods.
 ///
 /// See https://google.github.io/styleguide/cppguide.html#Default_Arguments
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-default-arguments.html
 class DefaultArgumentsCheck : public ClangTidyCheck {
 public:
   DefaultArgumentsCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/clang-tidy/google/ExplicitConstructorCheck.h b/clang-tools-extra/clang-tidy/google/ExplicitConstructorCheck.h
index 721ad4efc9797..573b0e18f90db 100644
--- a/clang-tools-extra/clang-tidy/google/ExplicitConstructorCheck.h
+++ b/clang-tools-extra/clang-tidy/google/ExplicitConstructorCheck.h
@@ -18,6 +18,9 @@ namespace google {
 /// Checks that all single-argument constructors are explicit.
 ///
 /// See https://google.github.io/styleguide/cppguide.html#Explicit_Constructors
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-explicit-constructor.html
 class ExplicitConstructorCheck : public ClangTidyCheck {
 public:
   ExplicitConstructorCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/clang-tidy/google/ExplicitMakePairCheck.h b/clang-tools-extra/clang-tidy/google/ExplicitMakePairCheck.h
index dae66c8afa8b9..56ca5b20966e1 100644
--- a/clang-tools-extra/clang-tidy/google/ExplicitMakePairCheck.h
+++ b/clang-tools-extra/clang-tidy/google/ExplicitMakePairCheck.h
@@ -22,6 +22,9 @@ namespace build {
 /// specified explicitly, and such use isn't intended in any case.
 ///
 /// Corresponding cpplint.py check name: 'build/explicit_make_pair'.
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-build-explicit-make-pair.html
 class ExplicitMakePairCheck : public ClangTidyCheck {
 public:
   ExplicitMakePairCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/clang-tidy/google/GlobalNamesInHeadersCheck.h b/clang-tools-extra/clang-tidy/google/GlobalNamesInHeadersCheck.h
index aa569dedcdb4b..801d362aa5094 100644
--- a/clang-tools-extra/clang-tidy/google/GlobalNamesInHeadersCheck.h
+++ b/clang-tools-extra/clang-tidy/google/GlobalNamesInHeadersCheck.h
@@ -27,6 +27,9 @@ namespace readability {
 ///
 ///     For extension-less header files, using an empty string or leaving an
 ///     empty string between ";" if there are other filename extensions.
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-global-names-in-headers.html
 class GlobalNamesInHeadersCheck : public ClangTidyCheck {
 public:
   GlobalNamesInHeadersCheck(StringRef Name, ClangTidyContext *Context);
diff --git a/clang-tools-extra/clang-tidy/google/IntegerTypesCheck.h b/clang-tools-extra/clang-tidy/google/IntegerTypesCheck.h
index 0b528dab2476d..5b2d2f501c91a 100644
--- a/clang-tools-extra/clang-tidy/google/IntegerTypesCheck.h
+++ b/clang-tools-extra/clang-tidy/google/IntegerTypesCheck.h
@@ -25,6 +25,9 @@ namespace runtime {
 /// with `u?intXX(_t)?`.
 ///
 /// Corresponding cpplint.py check: 'runtime/int'.
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-runtime-int.html
 class IntegerTypesCheck : public ClangTidyCheck {
 public:
   IntegerTypesCheck(StringRef Name, ClangTidyContext *Context);
diff --git a/clang-tools-extra/clang-tidy/google/OverloadedUnaryAndCheck.h b/clang-tools-extra/clang-tidy/google/OverloadedUnaryAndCheck.h
index 33cf4d62a45a6..7f513387ed616 100644
--- a/clang-tools-extra/clang-tidy/google/OverloadedUnaryAndCheck.h
+++ b/clang-tools-extra/clang-tidy/google/OverloadedUnaryAndCheck.h
@@ -21,6 +21,9 @@ namespace runtime {
 /// https://google.github.io/styleguide/cppguide.html#Operator_Overloading
 ///
 /// Corresponding cpplint.py check name: 'runtime/operator'.
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-runtime-operator.html
 class OverloadedUnaryAndCheck : public ClangTidyCheck {
 public:
   OverloadedUnaryAndCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/clang-tidy/google/TodoCommentCheck.h b/clang-tools-extra/clang-tidy/google/TodoCommentCheck.h
index 8c32dddc28373..ee14997f1506d 100644
--- a/clang-tools-extra/clang-tidy/google/TodoCommentCheck.h
+++ b/clang-tools-extra/clang-tidy/google/TodoCommentCheck.h
@@ -19,6 +19,9 @@ namespace readability {
 /// Finds TODO comments without a username or bug number.
 ///
 /// Corresponding cpplint.py check: 'readability/todo'
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-readability-todo.html
 class TodoCommentCheck : public ClangTidyCheck {
 public:
   TodoCommentCheck(StringRef Name, ClangTidyContext *Context);
diff --git a/clang-tools-extra/clang-tidy/google/UnnamedNamespaceInHeaderCheck.h b/clang-tools-extra/clang-tidy/google/UnnamedNamespaceInHeaderCheck.h
index 52af823e6a0e1..22110215a5efe 100644
--- a/clang-tools-extra/clang-tidy/google/UnnamedNamespaceInHeaderCheck.h
+++ b/clang-tools-extra/clang-tidy/google/UnnamedNamespaceInHeaderCheck.h
@@ -30,6 +30,9 @@ namespace build {
 /// https://google.github.io/styleguide/cppguide.html#Namespaces
 ///
 /// Corresponding cpplint.py check name: 'build/namespaces'.
+///
+/// For the user-facing documentation see:
+/// http://clang.llvm.org/extra/clang-tidy/checks/google-build-namespaces.html
 class UnnamedNamespaceInHeaderCheck : public ClangTidyCheck {
 public:
   UnnamedNamespaceInHeaderCheck(StringRef Name, ClangTidyContext *Context);
diff --git a/clang-tools-extra/clang-tidy/google/UsingNamespaceDirectiveCheck.h b/clang-tools-extra/clang-tidy/google/UsingNamespaceDirectiveCheck.h
index 97a1afbf160fe..c1b86c953fc59 100644
--- a/clang-tools-extra/clang-tidy/google/UsingNamespaceDirectiveCheck.h
+++ b/clang-tools-extra/clang-tidy/google/UsingNamespaceDirectiveCheck.h
@@ -31,6 +31,9 @@ namespace build {
 ///   \endcode
 ///
 /// Corresponding cpplint.py check name: `build/namespaces`.
+///
+/// For the user-facing documentation see:
+/// https://clang.llvm.org/extra/clang-tidy/checks/google-build-using-namespace.html
 class UsingNamespaceDirectiveCheck : public ClangTidyCheck {
 public:
   UsingNamespaceDirectiveCheck(StringRef Name, ClangTidyContext *Context)
diff --git a/clang-tools-extra/clang-tidy/performance/UnnecessaryCopyInitialization.cpp b/clang-tools-extra/clang-tidy/performance/UnnecessaryCopyInitialization.cpp
index 03b4450d8ca8c..24847d80657c1 100644
--- a/clang-tools-extra/clang-tidy/performance/UnnecessaryCopyInitialization.cpp
+++ b/clang-tools-extra/clang-tidy/performance/UnnecessaryCopyInitialization.cpp
@@ -63,8 +63,10 @@ void UnnecessaryCopyInitialization::registerMatchers(MatchFinder *Finder) {
                    declStmt(
                        has(varDecl(hasLocalStorage(),
                                    hasType(qualType(
-                                       hasCanonicalType(
-                                           matchers::isExpensiveToCopy()),
+                                       hasCanonicalType(allOf(
+                                           matchers::isExpensiveToCopy(),
+                                           unless(hasDeclaration(namedDecl(
+                                               hasName("::std::function")))))),
                                        unless(hasDeclaration(namedDecl(
                                            matchers::matchesAnyListedName(
                                                AllowedTypes)))))),
diff --git a/clang-tools-extra/test/clang-tidy/checkers/performance-unnecessary-copy-initialization.cpp b/clang-tools-extra/test/clang-tidy/checkers/performance-unnecessary-copy-initialization.cpp
index 7a70bc18a28c8..9055529c7a3fd 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/performance-unnecessary-copy-initialization.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/performance-unnecessary-copy-initialization.cpp
@@ -405,3 +405,58 @@ void negativeInitialzedFromFreeFunctionWithNonDefaultArg() {
   ExpensiveToCopyType Orig;
   const ExpensiveToCopyType Copy = freeFunctionWithDefaultArg(&Orig);
 }
+
+namespace std {
+inline namespace __1 {
+
+template <class>
+class function;
+template <class R, class... Args>
+class function<R(Args...)> {
+public:
+  function();
+  function(const function &other);
+  R operator()(Args &&...args) const;
+};
+
+} // namespace __1
+} // namespace std
+
+void negativeStdFunction() {
+  std::function<int()> Orig;
+  std::function<int()> Copy = Orig;
+  int i = Orig();
+}
+
+using Functor = std::function<int()>;
+
+void negativeAliasedStdFunction() {
+  Functor Orig;
+  Functor Copy = Orig;
+  int i = Orig();
+}
+
+typedef std::function<int()> TypedefFunc;
+
+void negativeTypedefedStdFunction() {
+  TypedefFunc Orig;
+  TypedefFunc Copy = Orig;
+  int i = Orig();
+}
+
+namespace fake {
+namespace std {
+template <class R, class... Args>
+struct function {
+  // Custom copy constructor makes it expensive to copy;
+  function(const function &);
+};
+} // namespace std
+
+void positiveFakeStdFunction(std::function<void(int)> F) {
+  auto Copy = F;
+  // CHECK-MESSAGES: [[@LINE-1]]:8: warning: local copy 'Copy' of the variable 'F' is never modified;
+  // CHECK-FIXES: const auto& Copy = F;
+}
+
+} // namespace fake
diff --git a/clang/CMakeLists.txt b/clang/CMakeLists.txt
index 7592d84ea9ed7..9e40cf1b2a976 100644
--- a/clang/CMakeLists.txt
+++ b/clang/CMakeLists.txt
@@ -449,7 +449,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     PATTERN "*.def"
     PATTERN "*.h"
     PATTERN "config.h" EXCLUDE
-    PATTERN ".svn" EXCLUDE
     )
 
   install(DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}/include/clang
diff --git a/clang/docs/ClangCommandLineReference.rst b/clang/docs/ClangCommandLineReference.rst
index 86bc9f637ab7e..6864f6058a3ff 100644
--- a/clang/docs/ClangCommandLineReference.rst
+++ b/clang/docs/ClangCommandLineReference.rst
@@ -3335,6 +3335,8 @@ X86
 
 .. option:: -mtsxldtrk, -mno-tsxldtrk
 
+.. option:: -muintr, -mno-uintr
+
 .. option:: -mvaes, -mno-vaes
 
 .. option:: -mvpclmulqdq, -mno-vpclmulqdq
diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index 540cf91326056..56337177e060e 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -200,6 +200,10 @@ X86 Support in Clang
   implies -mtune=<cpu>. -mtune=generic is the default with no -march or -mtune
   specified.
 
+- Support for ``HRESET`` instructions has been added.
+
+- Support for ``UINTR`` instructions has been added.
+
 Internal API Changes
 --------------------
 
diff --git a/clang/include/clang/AST/APValue.h b/clang/include/clang/AST/APValue.h
index 0ee48f35a20a1..6cda588ffe748 100644
--- a/clang/include/clang/AST/APValue.h
+++ b/clang/include/clang/AST/APValue.h
@@ -235,8 +235,10 @@ class APValue {
   struct UninitArray {};
   struct UninitStruct {};
 
-  friend class ASTReader;
+  friend class ASTRecordReader;
   friend class ASTWriter;
+  friend class ASTImporter;
+  friend class ASTNodeImporter;
 
 private:
   ValueKind Kind;
@@ -569,11 +571,9 @@ class APValue {
     *(APFixedPoint *)(char *)Data.buffer = std::move(FX);
   }
   void setVector(const APValue *E, unsigned N) {
-    assert(isVector() && "Invalid accessor");
-    ((Vec*)(char*)Data.buffer)->Elts = new APValue[N];
-    ((Vec*)(char*)Data.buffer)->NumElts = N;
+    MutableArrayRef<APValue> InternalElts = setVectorUninit(N);
     for (unsigned i = 0; i != N; ++i)
-      ((Vec*)(char*)Data.buffer)->Elts[i] = E[i];
+      InternalElts[i] = E[i];
   }
   void setComplexInt(APSInt R, APSInt I) {
     assert(R.getBitWidth() == I.getBitWidth() &&
@@ -594,11 +594,7 @@ class APValue {
   void setLValue(LValueBase B, const CharUnits &O,
                  ArrayRef<LValuePathEntry> Path, bool OnePastTheEnd,
                  bool IsNullPtr);
-  void setUnion(const FieldDecl *Field, const APValue &Value) {
-    assert(isUnion() && "Invalid accessor");
-    ((UnionData*)(char*)Data.buffer)->Field = Field;
-    *((UnionData*)(char*)Data.buffer)->Value = Value;
-  }
+  void setUnion(const FieldDecl *Field, const APValue &Value);
   void setAddrLabelDiff(const AddrLabelExpr* LHSExpr,
                         const AddrLabelExpr* RHSExpr) {
     ((AddrLabelDiffData*)(char*)Data.buffer)->LHSExpr = LHSExpr;
@@ -656,6 +652,24 @@ class APValue {
     new ((void*)(char*)Data.buffer) AddrLabelDiffData();
     Kind = AddrLabelDiff;
   }
+
+private:
+  /// The following functions are used as part of initialization, during
+  /// deserialization and importing. Reserve the space so that it can be
+  /// filled in by those steps.
+  MutableArrayRef<APValue> setVectorUninit(unsigned N) {
+    assert(isVector() && "Invalid accessor");
+    Vec *V = ((Vec *)(char *)Data.buffer);
+    V->Elts = new APValue[N];
+    V->NumElts = N;
+    return {V->Elts, V->NumElts};
+  }
+  MutableArrayRef<LValuePathEntry>
+  setLValueUninit(LValueBase B, const CharUnits &O, unsigned Size,
+                  bool OnePastTheEnd, bool IsNullPtr);
+  MutableArrayRef<const CXXRecordDecl *>
+  setMemberPointerUninit(const ValueDecl *Member, bool IsDerivedMember,
+                         unsigned Size);
 };
 
 } // end namespace clang.
diff --git a/clang/include/clang/AST/ASTContext.h b/clang/include/clang/AST/ASTContext.h
index 97facd5d3b68d..75daa122deebc 100644
--- a/clang/include/clang/AST/ASTContext.h
+++ b/clang/include/clang/AST/ASTContext.h
@@ -289,8 +289,8 @@ class ASTContext : public RefCountedBase<ASTContext> {
   /// Mapping from GUIDs to the corresponding MSGuidDecl.
   mutable llvm::FoldingSet<MSGuidDecl> MSGuidDecls;
 
-  /// Used to cleanups APValues stored in the AST.
-  mutable llvm::SmallVector<APValue *, 0> APValueCleanups;
+  /// Mapping from APValues to the corresponding TemplateParamObjects.
+  mutable llvm::FoldingSet<TemplateParamObjectDecl> TemplateParamObjectDecls;
 
   /// A cache mapping a string value to a StringLiteral object with the same
   /// value.
@@ -2876,6 +2876,11 @@ class ASTContext : public RefCountedBase<ASTContext> {
   /// GUID value.
   MSGuidDecl *getMSGuidDecl(MSGuidDeclParts Parts) const;
 
+  /// Return the template parameter object of the given type with the given
+  /// value.
+  TemplateParamObjectDecl *getTemplateParamObjectDecl(QualType T,
+                                                      const APValue &V) const;
+
   /// Parses the target attributes passed in, and returns only the ones that are
   /// valid feature names.
   ParsedTargetAttr filterFunctionTargetAttrs(const TargetAttr *TD) const;
diff --git a/clang/include/clang/AST/ASTImporter.h b/clang/include/clang/AST/ASTImporter.h
index 205d7ec67754f..a6d822ba2ea6d 100644
--- a/clang/include/clang/AST/ASTImporter.h
+++ b/clang/include/clang/AST/ASTImporter.h
@@ -14,6 +14,7 @@
 #ifndef LLVM_CLANG_AST_ASTIMPORTER_H
 #define LLVM_CLANG_AST_ASTIMPORTER_H
 
+#include "clang/AST/APValue.h"
 #include "clang/AST/DeclBase.h"
 #include "clang/AST/DeclarationName.h"
 #include "clang/AST/ExprCXX.h"
@@ -503,6 +504,13 @@ class TypeSourceInfo;
     /// "to" context, or the import error.
     llvm::Expected<CXXBaseSpecifier *> Import(const CXXBaseSpecifier *FromSpec);
 
+    /// Import the given APValue from the "from" context into
+    /// the "to" context.
+    ///
+    /// \return the equivalent APValue in the "to" context or the import
+    /// error.
+    llvm::Expected<APValue> Import(const APValue &FromValue);
+
     /// Import the definition of the given declaration, including all of
     /// the declarations it contains.
     LLVM_NODISCARD llvm::Error ImportDefinition(Decl *From);
diff --git a/clang/include/clang/AST/DeclTemplate.h b/clang/include/clang/AST/DeclTemplate.h
index 5f35257390912..641647659c17e 100644
--- a/clang/include/clang/AST/DeclTemplate.h
+++ b/clang/include/clang/AST/DeclTemplate.h
@@ -3226,7 +3226,7 @@ class VarTemplateDecl : public RedeclarableTemplateDecl {
   static bool classofKind(Kind K) { return K == VarTemplate; }
 };
 
-// \brief Declaration of a C++2a concept.
+/// Declaration of a C++2a concept.
 class ConceptDecl : public TemplateDecl, public Mergeable<ConceptDecl> {
 protected:
   Expr *ConstraintExpr;
@@ -3255,6 +3255,9 @@ class ConceptDecl : public TemplateDecl, public Mergeable<ConceptDecl> {
     return isa<TemplateTypeParmDecl>(getTemplateParameters()->getParam(0));
   }
 
+  ConceptDecl *getCanonicalDecl() override { return getFirstDecl(); }
+  const ConceptDecl *getCanonicalDecl() const { return getFirstDecl(); }
+
   // Implement isa/cast/dyncast/etc.
   static bool classof(const Decl *D) { return classofKind(D->getKind()); }
   static bool classofKind(Kind K) { return K == Concept; }
@@ -3264,6 +3267,74 @@ class ConceptDecl : public TemplateDecl, public Mergeable<ConceptDecl> {
   friend class ASTDeclWriter;
 };
 
+/// A template parameter object.
+///
+/// Template parameter objects represent values of class type used as template
+/// arguments. There is one template parameter object for each such distinct
+/// value used as a template argument across the program.
+///
+/// \code
+/// struct A { int x, y; };
+/// template<A> struct S;
+/// S<A{1, 2}> s1;
+/// S<A{1, 2}> s2; // same type, argument is same TemplateParamObjectDecl.
+/// \endcode
+class TemplateParamObjectDecl : public ValueDecl,
+                                public Mergeable<TemplateParamObjectDecl>,
+                                public llvm::FoldingSetNode {
+private:
+  /// The value of this template parameter object.
+  APValue Value;
+
+  TemplateParamObjectDecl(DeclContext *DC, QualType T, const APValue &V)
+      : ValueDecl(TemplateParamObject, DC, SourceLocation(), DeclarationName(),
+                  T),
+        Value(V) {}
+
+  static TemplateParamObjectDecl *Create(const ASTContext &C, QualType T,
+                                         const APValue &V);
+  static TemplateParamObjectDecl *CreateDeserialized(ASTContext &C,
+                                                     unsigned ID);
+
+  /// Only ASTContext::getTemplateParamObjectDecl and deserialization
+  /// create these.
+  friend class ASTContext;
+  friend class ASTReader;
+  friend class ASTDeclReader;
+
+public:
+  /// Print this template parameter object in a human-readable format.
+  void printName(llvm::raw_ostream &OS) const override;
+
+  /// Print this object as an equivalent expression.
+  void printAsExpr(llvm::raw_ostream &OS) const;
+
+  /// Print this object as an initializer suitable for a variable of the
+  /// object's type.
+  void printAsInit(llvm::raw_ostream &OS) const;
+
+  const APValue &getValue() const { return Value; }
+
+  static void Profile(llvm::FoldingSetNodeID &ID, QualType T,
+                      const APValue &V) {
+    ID.AddPointer(T.getCanonicalType().getAsOpaquePtr());
+    V.Profile(ID);
+  }
+  void Profile(llvm::FoldingSetNodeID &ID) {
+    Profile(ID, getType(), getValue());
+  }
+
+  TemplateParamObjectDecl *getCanonicalDecl() override {
+    return getFirstDecl();
+  }
+  const TemplateParamObjectDecl *getCanonicalDecl() const {
+    return getFirstDecl();
+  }
+
+  static bool classof(const Decl *D) { return classofKind(D->getKind()); }
+  static bool classofKind(Kind K) { return K == TemplateParamObject; }
+};
+
 inline NamedDecl *getAsNamedDecl(TemplateParameter P) {
   if (auto *PD = P.dyn_cast<TemplateTypeParmDecl *>())
     return PD;
diff --git a/clang/include/clang/AST/ExprCXX.h b/clang/include/clang/AST/ExprCXX.h
index 926021e9c6ed6..61a23ddaa368b 100644
--- a/clang/include/clang/AST/ExprCXX.h
+++ b/clang/include/clang/AST/ExprCXX.h
@@ -4233,8 +4233,10 @@ class SubstNonTypeTemplateParmExpr : public Expr {
   friend class ASTReader;
   friend class ASTStmtReader;
 
-  /// The replaced parameter.
-  NonTypeTemplateParmDecl *Param;
+  /// The replaced parameter and a flag indicating if it was a reference
+  /// parameter. For class NTTPs, we can't determine that based on the value
+  /// category alone.
+  llvm::PointerIntPair<NonTypeTemplateParmDecl*, 1, bool> ParamAndRef;
 
   /// The replacement expression.
   Stmt *Replacement;
@@ -4245,10 +4247,10 @@ class SubstNonTypeTemplateParmExpr : public Expr {
 public:
   SubstNonTypeTemplateParmExpr(QualType Ty, ExprValueKind ValueKind,
                                SourceLocation Loc,
-                               NonTypeTemplateParmDecl *Param,
+                               NonTypeTemplateParmDecl *Param, bool RefParam,
                                Expr *Replacement)
       : Expr(SubstNonTypeTemplateParmExprClass, Ty, ValueKind, OK_Ordinary),
-        Param(Param), Replacement(Replacement) {
+        ParamAndRef(Param, RefParam), Replacement(Replacement) {
     SubstNonTypeTemplateParmExprBits.NameLoc = Loc;
     setDependence(computeDependence(this));
   }
@@ -4261,7 +4263,14 @@ class SubstNonTypeTemplateParmExpr : public Expr {
 
   Expr *getReplacement() const { return cast<Expr>(Replacement); }
 
-  NonTypeTemplateParmDecl *getParameter() const { return Param; }
+  NonTypeTemplateParmDecl *getParameter() const {
+    return ParamAndRef.getPointer();
+  }
+
+  bool isReferenceParameter() const { return ParamAndRef.getInt(); }
+
+  /// Determine the substituted type of the template parameter.
+  QualType getParameterType(const ASTContext &Ctx) const;
 
   static bool classof(const Stmt *s) {
     return s->getStmtClass() == SubstNonTypeTemplateParmExprClass;
diff --git a/clang/include/clang/AST/RecursiveASTVisitor.h b/clang/include/clang/AST/RecursiveASTVisitor.h
index 6f07b92f25323..5e83cded06520 100644
--- a/clang/include/clang/AST/RecursiveASTVisitor.h
+++ b/clang/include/clang/AST/RecursiveASTVisitor.h
@@ -1970,6 +1970,8 @@ DEF_TRAVERSE_DECL(MSPropertyDecl, { TRY_TO(TraverseDeclaratorHelper(D)); })
 
 DEF_TRAVERSE_DECL(MSGuidDecl, {})
 
+DEF_TRAVERSE_DECL(TemplateParamObjectDecl, {})
+
 DEF_TRAVERSE_DECL(FieldDecl, {
   TRY_TO(TraverseDeclaratorHelper(D));
   if (D->isBitField())
diff --git a/clang/include/clang/Basic/BuiltinsX86_64.def b/clang/include/clang/Basic/BuiltinsX86_64.def
index f66ae78f7e81f..3e186af82ff71 100644
--- a/clang/include/clang/Basic/BuiltinsX86_64.def
+++ b/clang/include/clang/Basic/BuiltinsX86_64.def
@@ -94,6 +94,12 @@ TARGET_BUILTIN(__builtin_ia32_cvtusi2sd64, "V2dV2dUOiIi", "ncV:128:", "avx512f")
 TARGET_BUILTIN(__builtin_ia32_cvtusi2ss64, "V4fV4fUOiIi", "ncV:128:", "avx512f")
 TARGET_BUILTIN(__builtin_ia32_directstore_u64, "vULi*ULi", "n", "movdiri")
 
+// UINTR
+TARGET_BUILTIN(__builtin_ia32_clui, "v", "n", "uintr")
+TARGET_BUILTIN(__builtin_ia32_stui, "v", "n", "uintr")
+TARGET_BUILTIN(__builtin_ia32_testui, "Uc", "n", "uintr")
+TARGET_BUILTIN(__builtin_ia32_senduipi, "vUWi", "n", "uintr")
+
 // AMX
 TARGET_BUILTIN(__builtin_ia32_tile_loadconfig, "vvC*", "n", "amx-tile")
 TARGET_BUILTIN(__builtin_ia32_tile_storeconfig, "vvC*", "n", "amx-tile")
diff --git a/clang/include/clang/Basic/CodeGenOptions.def b/clang/include/clang/Basic/CodeGenOptions.def
index 641b99edaed11..0ab9054f0bb5d 100644
--- a/clang/include/clang/Basic/CodeGenOptions.def
+++ b/clang/include/clang/Basic/CodeGenOptions.def
@@ -356,6 +356,9 @@ ENUM_CODEGENOPT(DefaultTLSModel, TLSModel, 2, GeneralDynamicTLSModel)
 /// Bit size of immediate TLS offsets (0 == use the default).
 VALUE_CODEGENOPT(TLSSize, 8, 0)
 
+/// The default stack protector guard offset to use.
+VALUE_CODEGENOPT(StackProtectorGuardOffset, 32, (unsigned)-1)
+
 /// Number of path components to strip when emitting checks. (0 == full
 /// filename)
 VALUE_CODEGENOPT(EmitCheckPathComponentsToStrip, 32, 0)
diff --git a/clang/include/clang/Basic/CodeGenOptions.h b/clang/include/clang/Basic/CodeGenOptions.h
index ca391bf8f1861..f658c9b8a7819 100644
--- a/clang/include/clang/Basic/CodeGenOptions.h
+++ b/clang/include/clang/Basic/CodeGenOptions.h
@@ -327,6 +327,15 @@ class CodeGenOptions : public CodeGenOptionsBase {
   /// by sanitizer coverage pass.
   std::vector<std::string> SanitizeCoverageAllowlistFiles;
 
+  /// The guard style used for stack protector to get a initial value, this
+  /// value usually be gotten from TLS or get from __stack_chk_guard, or some
+  /// other styles we may implement in the future.
+  std::string StackProtectorGuard;
+
+  /// The TLS base register when StackProtectorGuard is "tls".
+  /// On x86 this can be "fs" or "gs".
+  std::string StackProtectorGuardReg;
+
   /// Path to blocklist file specifying which objects
   /// (files, functions) listed for instrumentation by sanitizer
   /// coverage pass should actually not be instrumented.
diff --git a/clang/include/clang/Basic/DeclNodes.td b/clang/include/clang/Basic/DeclNodes.td
index 866988ee3f015..4771a3549426b 100644
--- a/clang/include/clang/Basic/DeclNodes.td
+++ b/clang/include/clang/Basic/DeclNodes.td
@@ -41,6 +41,7 @@ def Named : DeclNode<Decl, "named declarations", 1>;
     def OMPDeclareReduction : DeclNode<Value>, DeclContext;
     def OMPDeclareMapper : DeclNode<Value>, DeclContext;
     def MSGuid : DeclNode<Value>;
+    def TemplateParamObject : DeclNode<Value>;
     def Declarator : DeclNode<Value, "declarators", 1>;
       def Field : DeclNode<Declarator, "non-static data members">;
         def ObjCIvar : DeclNode<Field>;
diff --git a/clang/include/clang/Basic/DiagnosticDriverKinds.td b/clang/include/clang/Basic/DiagnosticDriverKinds.td
index 5d49f41dd15ff..fd80504ca8794 100644
--- a/clang/include/clang/Basic/DiagnosticDriverKinds.td
+++ b/clang/include/clang/Basic/DiagnosticDriverKinds.td
@@ -223,6 +223,7 @@ def err_drv_dllexport_inlines_and_fallback : Error<
 
 def err_drv_invalid_value : Error<"invalid value '%1' in '%0'">;
 def err_drv_invalid_int_value : Error<"invalid integral value '%1' in '%0'">;
+def err_drv_invalid_value_with_suggestion : Error<"invalid value '%1' in '%0','%2'">;
 def err_drv_invalid_remap_file : Error<
     "invalid option '%0' not of the form <from-file>;<to-file>">;
 def err_drv_invalid_gcc_output_type : Error<
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 2de36ff88607d..190bb3cba4ef7 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -2025,8 +2025,8 @@ def err_destructor_template : Error<
 
 // C++ initialization
 def err_init_conversion_failed : Error<
-  "cannot initialize %select{a variable|a parameter|return object|"
-  "statement expression result|an "
+  "cannot initialize %select{a variable|a parameter|template parameter|"
+  "return object|statement expression result|an "
   "exception object|a member subobject|an array element|a new value|a value|a "
   "base class|a constructor delegation|a vector element|a block element|a "
   "block element|a complex element|a lambda capture|a compound literal "
@@ -2168,7 +2168,7 @@ def warn_unsequenced_mod_use : Warning<
   "unsequenced modification and access to %0">, InGroup<Unsequenced>;
 
 def select_initialized_entity_kind : TextSubstitution<
-  "%select{copying variable|copying parameter|"
+  "%select{copying variable|copying parameter|initializing template parameter|"
   "returning object|initializing statement expression result|"
   "throwing object|copying member subobject|copying array element|"
   "allocating object|copying temporary|initializing base subobject|"
@@ -4533,6 +4533,10 @@ def note_not_structural_rvalue_ref_field : Note<
 def note_not_structural_subobject : Note<
   "%0 is not a structural type because it has a "
   "%select{non-static data member|base class}1 of non-structural type %2">;
+def warn_cxx17_compat_template_nontype_parm_type : Warning<
+  "non-type template parameter of type %0 is incompatible with "
+  "C++ standards before C++20">,
+  DefaultIgnore, InGroup<CXXPre20Compat>;
 def warn_cxx14_compat_template_nontype_parm_auto_type : Warning<
   "non-type template parameters declared with %0 are incompatible with C++ "
   "standards before C++17">,
diff --git a/clang/include/clang/Basic/OpenCLExtensions.def b/clang/include/clang/Basic/OpenCLExtensions.def
index 1ae36b32fb0a9..d67cb3ff019b4 100644
--- a/clang/include/clang/Basic/OpenCLExtensions.def
+++ b/clang/include/clang/Basic/OpenCLExtensions.def
@@ -23,6 +23,16 @@
 //   core - minimum OpenCL version when the extension becomes optional core
 //          feature or core feature. ~0U indicates not a core feature or an
 //          optional core feature.
+//
+// As per The OpenCL Extension Specification, Section 1.2, in this file, an
+// extension is defined if and only it either:
+//  * affects the OpenCL language semantics or its syntax,
+//  * adds built-in functions to the language.
+//
+// For such an extension, a preprocessor #define that matches the extension
+// name must be created and a #pragma is required if and only if the
+// compilation flow is impacted, e.g. due to a difference of syntax or
+// semantics in the language compared to the core standard.
 
 #ifndef OPENCLEXT_INTERNAL
 #ifndef OPENCLEXT
@@ -34,8 +44,6 @@
 
 // OpenCL 1.0.
 OPENCLEXT_INTERNAL(cl_khr_3d_image_writes, 100, 200)
-// fprounding mode is special since it is not mentioned beyond 1.0
-OPENCLEXT_INTERNAL(cl_khr_select_fprounding_mode, 100, 110)
 OPENCLEXT_INTERNAL(cl_khr_byte_addressable_store, 100, 110)
 OPENCLEXT_INTERNAL(cl_khr_fp16, 100, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_fp64, 100, 120)
@@ -45,35 +53,19 @@ OPENCLEXT_INTERNAL(cl_khr_local_int32_base_atomics, 100, 110)
 OPENCLEXT_INTERNAL(cl_khr_local_int32_extended_atomics, 100, 110)
 OPENCLEXT_INTERNAL(cl_khr_int64_base_atomics, 100, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_int64_extended_atomics, 100, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_gl_sharing, 100, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_icd, 100, ~0U)
-
-// OpenCL 1.1.
-OPENCLEXT_INTERNAL(cl_khr_gl_event, 110, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_d3d10_sharing, 110, ~0U)
 
 // EMBEDDED_PROFILE
 OPENCLEXT_INTERNAL(cles_khr_int64, 110, ~0U)
 
 // OpenCL 1.2.
-OPENCLEXT_INTERNAL(cl_khr_context_abort, 120, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_d3d11_sharing, 120, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_depth_images, 120, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_dx9_media_sharing, 120, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_image2d_from_buffer, 120, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_initialize_memory, 120, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_gl_depth_images, 120, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_gl_msaa_sharing, 120, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_spir, 120, ~0U)
 
 // OpenCL 2.0.
-OPENCLEXT_INTERNAL(cl_khr_egl_event, 200, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_egl_image, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_mipmap_image, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_mipmap_image_writes, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_srgb_image_writes, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_subgroups, 200, ~0U)
-OPENCLEXT_INTERNAL(cl_khr_terminate_context, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_subgroup_extended_types, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_subgroup_non_uniform_vote, 200, ~0U)
 OPENCLEXT_INTERNAL(cl_khr_subgroup_ballot, 200, ~0U)
diff --git a/clang/include/clang/Basic/TargetInfo.h b/clang/include/clang/Basic/TargetInfo.h
index 7253b5ea9abed..26dc6eacb2041 100644
--- a/clang/include/clang/Basic/TargetInfo.h
+++ b/clang/include/clang/Basic/TargetInfo.h
@@ -581,8 +581,9 @@ class TargetInfo : public virtual TransferrableTargetInfo,
   /// Determine whether constrained floating point is supported on this target.
   virtual bool hasStrictFP() const { return HasStrictFP; }
 
-  /// Return the alignment that is suitable for storing any
-  /// object with a fundamental alignment requirement.
+  /// Return the alignment that is the largest alignment ever used for any
+  /// scalar/SIMD data type on the target machine you are compiling for
+  /// (including types with an extended alignment requirement).
   unsigned getSuitableAlign() const { return SuitableAlign; }
 
   /// Return the default alignment for __attribute__((aligned)) on
diff --git a/clang/include/clang/Basic/arm_mve.td b/clang/include/clang/Basic/arm_mve.td
index 25daae2a0a25d..94f76dd729b0b 100644
--- a/clang/include/clang/Basic/arm_mve.td
+++ b/clang/include/clang/Basic/arm_mve.td
@@ -406,7 +406,7 @@ def vabdq: Intrinsic<Vector, (args Vector:$a, Vector:$b),
 }
 
 multiclass VectorVectorArithmetic<string operation, dag extraArgs = (?),
-                                  int wantXVariant = 1> {
+                                  bit wantXVariant = 1> {
   defm "" : IntrinsicMX<
       Vector, (args Vector:$a, Vector:$b, Predicate:$pred),
       !con((IRInt<operation, [Vector, Predicate]> $a, $b),
@@ -415,7 +415,7 @@ multiclass VectorVectorArithmetic<string operation, dag extraArgs = (?),
 
 multiclass VectorScalarArithmetic<string operation, string basename,
                                   dag extraArgs = (?),
-                                  int wantXVariant = 1> {
+                                  bit wantXVariant = 1> {
   defm "" : IntrinsicMXNameOverride<
       Vector, (args Vector:$a, unpromoted<Scalar>:$b, Predicate:$pred),
       !con((IRInt<operation, [Vector, Predicate]> $a, (splat $b)),
@@ -451,7 +451,7 @@ let params = T.Usual in {
 }
 
 multiclass DblVectorVectorArithmetic<string operation, dag extraArgs = (?),
-                                     int wantXVariant = 1> {
+                                     bit wantXVariant = 1> {
   defm "" : IntrinsicMX<
       DblVector, (args Vector:$a, Vector:$b, DblPredicate:$pred),
       !con((IRInt<operation, [DblVector, Vector, DblPredicate]> $a, $b),
@@ -460,7 +460,7 @@ multiclass DblVectorVectorArithmetic<string operation, dag extraArgs = (?),
 
 multiclass DblVectorScalarArithmetic<string operation, string basename,
                                      dag extraArgs = (?),
-                                     int wantXVariant = 1> {
+                                     bit wantXVariant = 1> {
   defm "" : IntrinsicMXNameOverride<
       DblVector, (args Vector:$a, unpromoted<Scalar>:$b, DblPredicate:$pred),
       !con((IRInt<operation, [DblVector, Vector, DblPredicate]> $a, (splat $b)),
diff --git a/clang/include/clang/Basic/arm_mve_defs.td b/clang/include/clang/Basic/arm_mve_defs.td
index 4038a18027f8c..1a090c08cc853 100644
--- a/clang/include/clang/Basic/arm_mve_defs.td
+++ b/clang/include/clang/Basic/arm_mve_defs.td
@@ -519,7 +519,7 @@ class NameOverride<string basename_> {
 // polymorph differently (typically because the type of the inactive
 // parameter can be used as a disambiguator if it's present).
 multiclass IntrinsicMX<Type rettype, dag arguments, dag cg,
-                       int wantXVariant = 1,
+                       bit wantXVariant = 1,
                        string nameSuffix = "",
                        PolymorphicNameType pnt_m = PNT_Type,
                        PolymorphicNameType pnt_x = PNT_Type> {
@@ -532,7 +532,7 @@ multiclass IntrinsicMX<Type rettype, dag arguments, dag cg,
     let pnt = pnt_m;
   }
 
-  foreach unusedVar = !if(!eq(wantXVariant, 1), [1], []<int>) in {
+  if wantXVariant then {
     // The _x variant leaves off that parameter, and simply uses an
     // undef value of the same type.
 
@@ -546,7 +546,7 @@ multiclass IntrinsicMX<Type rettype, dag arguments, dag cg,
 // Same as above, but with an additional parameter 'basename' which overrides
 // the C intrinsic base name
 multiclass IntrinsicMXNameOverride<Type rettype, dag arguments, dag cg,
-                                   string basename, int wantXVariant = 1,
+                                   string basename, bit wantXVariant = 1,
                                    string nameSuffix = "",
                                    PolymorphicNameType pnt_m = PNT_Type,
                                    PolymorphicNameType pnt_x = PNT_Type> {
@@ -556,7 +556,7 @@ multiclass IntrinsicMXNameOverride<Type rettype, dag arguments, dag cg,
     let pnt = pnt_m;
   }
 
-  foreach unusedVar = !if(!eq(wantXVariant, 1), [1], []<int>) in {
+  if wantXVariant then {
     def "_x" # nameSuffix:
       Intrinsic<rettype, arguments, (seq (undef rettype):$inactive, cg)>,
       NameOverride<basename # "_x" # nameSuffix> {
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index ac00407352efe..fc7055ef08a0a 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -2670,6 +2670,12 @@ def mrecip : Flag<["-"], "mrecip">, Group<m_Group>;
 def mrecip_EQ : CommaJoined<["-"], "mrecip=">, Group<m_Group>, Flags<[CC1Option]>;
 def mprefer_vector_width_EQ : Joined<["-"], "mprefer-vector-width=">, Group<m_Group>, Flags<[CC1Option]>,
   HelpText<"Specifies preferred vector width for auto-vectorization. Defaults to 'none' which allows target specific decisions.">;
+def mstack_protector_guard_EQ : Joined<["-"], "mstack-protector-guard=">, Group<m_Group>, Flags<[CC1Option]>,
+  HelpText<"Use the given guard (global, tls) for addressing the stack-protector guard">;
+def mstack_protector_guard_offset_EQ : Joined<["-"], "mstack-protector-guard-offset=">, Group<m_Group>, Flags<[CC1Option]>,
+  HelpText<"Use the given offset for addressing the stack-protector guard">;
+def mstack_protector_guard_reg_EQ : Joined<["-"], "mstack-protector-guard-reg=">, Group<m_Group>, Flags<[CC1Option]>,
+  HelpText<"Use the given reg for addressing the stack-protector guard">;
 def mpie_copy_relocations : Flag<["-"], "mpie-copy-relocations">, Group<m_Group>,
   Flags<[CC1Option]>,
   HelpText<"Use copy relocations support for PIE builds">;
@@ -3381,6 +3387,8 @@ def mtbm : Flag<["-"], "mtbm">, Group<m_x86_Features_Group>;
 def mno_tbm : Flag<["-"], "mno-tbm">, Group<m_x86_Features_Group>;
 def mtsxldtrk : Flag<["-"], "mtsxldtrk">, Group<m_x86_Features_Group>;
 def mno_tsxldtrk : Flag<["-"], "mno-tsxldtrk">, Group<m_x86_Features_Group>;
+def muintr : Flag<["-"], "muintr">, Group<m_x86_Features_Group>;
+def mno_uintr : Flag<["-"], "mno-uintr">, Group<m_x86_Features_Group>;
 def mvaes : Flag<["-"], "mvaes">, Group<m_x86_Features_Group>;
 def mno_vaes : Flag<["-"], "mno-vaes">, Group<m_x86_Features_Group>;
 def mvpclmulqdq : Flag<["-"], "mvpclmulqdq">, Group<m_x86_Features_Group>;
diff --git a/clang/include/clang/Sema/Initialization.h b/clang/include/clang/Sema/Initialization.h
index ca9e0a198cb91..6976e7c95c8b2 100644
--- a/clang/include/clang/Sema/Initialization.h
+++ b/clang/include/clang/Sema/Initialization.h
@@ -55,6 +55,9 @@ class alignas(8) InitializedEntity {
     /// The entity being initialized is a function parameter.
     EK_Parameter,
 
+    /// The entity being initialized is a non-type template parameter.
+    EK_TemplateParameter,
+
     /// The entity being initialized is the result of a function call.
     EK_Result,
 
@@ -175,7 +178,8 @@ class alignas(8) InitializedEntity {
   };
 
   union {
-    /// When Kind == EK_Variable, EK_Member or EK_Binding, the variable.
+    /// When Kind == EK_Variable, EK_Member, EK_Binding, or
+    /// EK_TemplateParameter, the variable, binding, or template parameter.
     VD Variable;
 
     /// When Kind == EK_RelatedResult, the ObjectiveC method where
@@ -281,6 +285,17 @@ class alignas(8) InitializedEntity {
     return Entity;
   }
 
+  /// Create the initialization entity for a template parameter.
+  static InitializedEntity
+  InitializeTemplateParameter(QualType T, NonTypeTemplateParmDecl *Param) {
+    InitializedEntity Entity;
+    Entity.Kind = EK_TemplateParameter;
+    Entity.Type = T;
+    Entity.Parent = nullptr;
+    Entity.Variable = {Param, false, false};
+    return Entity;
+  }
+
   /// Create the initialization entity for the result of a function.
   static InitializedEntity InitializeResult(SourceLocation ReturnLoc,
                                             QualType Type, bool NRVO) {
@@ -441,6 +456,10 @@ class alignas(8) InitializedEntity {
             getKind() == EK_Parameter_CF_Audited);
   }
 
+  bool isParamOrTemplateParamKind() const {
+    return isParameterKind() || getKind() == EK_TemplateParameter;
+  }
+
   /// Determine whether this initialization consumes the
   /// parameter.
   bool isParameterConsumed() const {
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 4c01e94241554..c50749c0877ed 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -3523,7 +3523,8 @@ class Sema final {
   ExprResult CheckConvertedConstantExpression(Expr *From, QualType T,
                                               llvm::APSInt &Value, CCEKind CCE);
   ExprResult CheckConvertedConstantExpression(Expr *From, QualType T,
-                                              APValue &Value, CCEKind CCE);
+                                              APValue &Value, CCEKind CCE,
+                                              NamedDecl *Dest = nullptr);
 
   /// Abstract base class used to perform a contextual implicit
   /// conversion from an expression to any type passing a filter.
diff --git a/clang/include/clang/Serialization/ASTBitCodes.h b/clang/include/clang/Serialization/ASTBitCodes.h
index 726c383ade807..52ddd57492025 100644
--- a/clang/include/clang/Serialization/ASTBitCodes.h
+++ b/clang/include/clang/Serialization/ASTBitCodes.h
@@ -1286,6 +1286,9 @@ class TypeIdx {
       /// A MSGuidDecl record.
       DECL_MS_GUID,
 
+      /// A TemplateParamObjectDecl record.
+      DECL_TEMPLATE_PARAM_OBJECT,
+
       /// A VarDecl record.
       DECL_VAR,
 
diff --git a/clang/lib/AST/APValue.cpp b/clang/lib/AST/APValue.cpp
index 8d402ee8e3dca..53aa09fa9b364 100644
--- a/clang/lib/AST/APValue.cpp
+++ b/clang/lib/AST/APValue.cpp
@@ -414,7 +414,18 @@ void APValue::swap(APValue &RHS) {
   std::swap(Data, RHS.Data);
 }
 
+/// Profile the value of an APInt, excluding its bit-width.
+static void profileIntValue(llvm::FoldingSetNodeID &ID, const llvm::APInt &V) {
+  for (unsigned I = 0, N = V.getBitWidth(); I < N; I += 32)
+    ID.AddInteger((uint32_t)V.extractBitsAsZExtValue(std::min(32u, N - I), I));
+}
+
 void APValue::Profile(llvm::FoldingSetNodeID &ID) const {
+  // Note that our profiling assumes that only APValues of the same type are
+  // ever compared. As a result, we don't consider collisions that could only
+  // happen if the types are different. (For example, structs with different
+  // numbers of members could profile the same.)
+
   ID.AddInteger(Kind);
 
   switch (Kind) {
@@ -428,25 +439,22 @@ void APValue::Profile(llvm::FoldingSetNodeID &ID) const {
     return;
 
   case Struct:
-    ID.AddInteger(getStructNumBases());
     for (unsigned I = 0, N = getStructNumBases(); I != N; ++I)
       getStructBase(I).Profile(ID);
-    ID.AddInteger(getStructNumFields());
     for (unsigned I = 0, N = getStructNumFields(); I != N; ++I)
       getStructField(I).Profile(ID);
     return;
 
   case Union:
     if (!getUnionField()) {
-      ID.AddPointer(nullptr);
+      ID.AddInteger(0);
       return;
     }
-    ID.AddPointer(getUnionField()->getCanonicalDecl());
+    ID.AddInteger(getUnionField()->getFieldIndex() + 1);
     getUnionValue().Profile(ID);
     return;
 
   case Array: {
-    ID.AddInteger(getArraySize());
     if (getArraySize() == 0)
       return;
 
@@ -502,46 +510,46 @@ void APValue::Profile(llvm::FoldingSetNodeID &ID) const {
   }
 
   case Vector:
-    ID.AddInteger(getVectorLength());
     for (unsigned I = 0, N = getVectorLength(); I != N; ++I)
       getVectorElt(I).Profile(ID);
     return;
 
   case Int:
-    // We don't need to include the sign bit; it's implied by the type.
-    getInt().APInt::Profile(ID);
+    profileIntValue(ID, getInt());
     return;
 
   case Float:
-    getFloat().Profile(ID);
+    profileIntValue(ID, getFloat().bitcastToAPInt());
     return;
 
   case FixedPoint:
-    // We don't need to include the fixed-point semantics; they're
-    // implied by the type.
-    getFixedPoint().getValue().APInt::Profile(ID);
+    profileIntValue(ID, getFixedPoint().getValue());
     return;
 
   case ComplexFloat:
-    getComplexFloatReal().Profile(ID);
-    getComplexFloatImag().Profile(ID);
+    profileIntValue(ID, getComplexFloatReal().bitcastToAPInt());
+    profileIntValue(ID, getComplexFloatImag().bitcastToAPInt());
     return;
 
   case ComplexInt:
-    getComplexIntReal().APInt::Profile(ID);
-    getComplexIntImag().APInt::Profile(ID);
+    profileIntValue(ID, getComplexIntReal());
+    profileIntValue(ID, getComplexIntImag());
     return;
 
   case LValue:
     getLValueBase().Profile(ID);
     ID.AddInteger(getLValueOffset().getQuantity());
-    ID.AddInteger(isNullPointer());
-    ID.AddInteger(isLValueOnePastTheEnd());
-    // For uniqueness, we only need to profile the entries corresponding
-    // to union members, but we don't have the type here so we don't know
-    // how to interpret the entries.
-    for (LValuePathEntry E : getLValuePath())
-      E.Profile(ID);
+    ID.AddInteger((isNullPointer() ? 1 : 0) |
+                  (isLValueOnePastTheEnd() ? 2 : 0) |
+                  (hasLValuePath() ? 4 : 0));
+    if (hasLValuePath()) {
+      ID.AddInteger(getLValuePath().size());
+      // For uniqueness, we only need to profile the entries corresponding
+      // to union members, but we don't have the type here so we don't know
+      // how to interpret the entries.
+      for (LValuePathEntry E : getLValuePath())
+        E.Profile(ID);
+    }
     return;
 
   case MemberPointer:
@@ -882,17 +890,33 @@ void APValue::setLValue(LValueBase B, const CharUnits &O, NoLValuePath,
   LVal.IsNullPtr = IsNullPtr;
 }
 
-void APValue::setLValue(LValueBase B, const CharUnits &O,
-                        ArrayRef<LValuePathEntry> Path, bool IsOnePastTheEnd,
-                        bool IsNullPtr) {
+MutableArrayRef<APValue::LValuePathEntry>
+APValue::setLValueUninit(LValueBase B, const CharUnits &O, unsigned Size,
+                         bool IsOnePastTheEnd, bool IsNullPtr) {
   assert(isLValue() && "Invalid accessor");
-  LV &LVal = *((LV*)(char*)Data.buffer);
+  LV &LVal = *((LV *)(char *)Data.buffer);
   LVal.Base = B;
   LVal.IsOnePastTheEnd = IsOnePastTheEnd;
   LVal.Offset = O;
-  LVal.resizePath(Path.size());
-  memcpy(LVal.getPath(), Path.data(), Path.size() * sizeof(LValuePathEntry));
   LVal.IsNullPtr = IsNullPtr;
+  LVal.resizePath(Size);
+  return {LVal.getPath(), Size};
+}
+
+void APValue::setLValue(LValueBase B, const CharUnits &O,
+                        ArrayRef<LValuePathEntry> Path, bool IsOnePastTheEnd,
+                        bool IsNullPtr) {
+  MutableArrayRef<APValue::LValuePathEntry> InternalPath =
+      setLValueUninit(B, O, Path.size(), IsOnePastTheEnd, IsNullPtr);
+  memcpy(InternalPath.data(), Path.data(),
+         Path.size() * sizeof(LValuePathEntry));
+}
+
+void APValue::setUnion(const FieldDecl *Field, const APValue &Value) {
+  assert(isUnion() && "Invalid accessor");
+  ((UnionData *)(char *)Data.buffer)->Field =
+      Field ? Field->getCanonicalDecl() : nullptr;
+  *((UnionData*)(char*)Data.buffer)->Value = Value;
 }
 
 const ValueDecl *APValue::getMemberPointerDecl() const {
@@ -929,15 +953,27 @@ void APValue::MakeArray(unsigned InitElts, unsigned Size) {
   Kind = Array;
 }
 
-void APValue::MakeMemberPointer(const ValueDecl *Member, bool IsDerivedMember,
-                                ArrayRef<const CXXRecordDecl*> Path) {
+MutableArrayRef<APValue::LValuePathEntry>
+setLValueUninit(APValue::LValueBase B, const CharUnits &O, unsigned Size,
+                bool OnePastTheEnd, bool IsNullPtr);
+
+MutableArrayRef<const CXXRecordDecl *>
+APValue::setMemberPointerUninit(const ValueDecl *Member, bool IsDerivedMember,
+                                unsigned Size) {
   assert(isAbsent() && "Bad state change");
-  MemberPointerData *MPD = new ((void*)(char*)Data.buffer) MemberPointerData;
+  MemberPointerData *MPD = new ((void *)(char *)Data.buffer) MemberPointerData;
   Kind = MemberPointer;
   MPD->MemberAndIsDerivedMember.setPointer(
       Member ? cast<ValueDecl>(Member->getCanonicalDecl()) : nullptr);
   MPD->MemberAndIsDerivedMember.setInt(IsDerivedMember);
-  MPD->resizePath(Path.size());
+  MPD->resizePath(Size);
+  return {MPD->getPath(), MPD->PathLength};
+}
+
+void APValue::MakeMemberPointer(const ValueDecl *Member, bool IsDerivedMember,
+                                ArrayRef<const CXXRecordDecl *> Path) {
+  MutableArrayRef<const CXXRecordDecl *> InternalPath =
+      setMemberPointerUninit(Member, IsDerivedMember, Path.size());
   for (unsigned I = 0; I != Path.size(); ++I)
-    MPD->getPath()[I] = Path[I]->getCanonicalDecl();
+    InternalPath[I] = Path[I]->getCanonicalDecl();
 }
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 2ebc14d6c62c2..1a6a8e77f722b 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -1005,9 +1005,6 @@ ASTContext::~ASTContext() {
 
   for (const auto &Value : ModuleInitializers)
     Value.second->~PerModuleInitializers();
-
-  for (APValue *Value : APValueCleanups)
-    Value->~APValue();
 }
 
 void ASTContext::setTraversalScope(const std::vector<Decl *> &TopLevelDecls) {
@@ -4885,9 +4882,16 @@ TemplateArgument ASTContext::getInjectedTemplateArg(NamedDecl *Param) {
 
     Arg = TemplateArgument(ArgType);
   } else if (auto *NTTP = dyn_cast<NonTypeTemplateParmDecl>(Param)) {
+    QualType T =
+        NTTP->getType().getNonPackExpansionType().getNonLValueExprType(*this);
+    // For class NTTPs, ensure we include the 'const' so the type matches that
+    // of a real template argument.
+    // FIXME: It would be more faithful to model this as something like an
+    // lvalue-to-rvalue conversion applied to a const-qualified lvalue.
+    if (T->isRecordType())
+      T.addConst();
     Expr *E = new (*this) DeclRefExpr(
-        *this, NTTP, /*enclosing*/ false,
-        NTTP->getType().getNonPackExpansionType().getNonLValueExprType(*this),
+        *this, NTTP, /*enclosing*/ false, T,
         Expr::getValueKindForType(NTTP->getType()), NTTP->getLocation());
 
     if (NTTP->isParameterPack())
@@ -11006,6 +11010,27 @@ ASTContext::getMSGuidDecl(MSGuidDecl::Parts Parts) const {
   return New;
 }
 
+TemplateParamObjectDecl *
+ASTContext::getTemplateParamObjectDecl(QualType T, const APValue &V) const {
+  assert(T->isRecordType() && "template param object of unexpected type");
+
+  // C++ [temp.param]p8:
+  //   [...] a static storage duration object of type 'const T' [...]
+  T.addConst();
+
+  llvm::FoldingSetNodeID ID;
+  TemplateParamObjectDecl::Profile(ID, T, V);
+
+  void *InsertPos;
+  if (TemplateParamObjectDecl *Existing =
+          TemplateParamObjectDecls.FindNodeOrInsertPos(ID, InsertPos))
+    return Existing;
+
+  TemplateParamObjectDecl *New = TemplateParamObjectDecl::Create(*this, T, V);
+  TemplateParamObjectDecls.InsertNode(New, InsertPos);
+  return New;
+}
+
 bool ASTContext::AtomicUsesUnsupportedLibcall(const AtomicExpr *E) const {
   const llvm::Triple &T = getTargetInfo().getTriple();
   if (!T.isOSDarwin())
diff --git a/clang/lib/AST/ASTDiagnostic.cpp b/clang/lib/AST/ASTDiagnostic.cpp
index 99ce46e83123e..2bc731717b982 100644
--- a/clang/lib/AST/ASTDiagnostic.cpp
+++ b/clang/lib/AST/ASTDiagnostic.cpp
@@ -1834,7 +1834,14 @@ class TemplateDiff {
     if (VD) {
       if (AddressOf)
         OS << "&";
-      OS << VD->getName();
+      else if (auto *TPO = dyn_cast<TemplateParamObjectDecl>(VD)) {
+        // FIXME: Diffing the APValue would be neat.
+        // FIXME: Suppress this and use the full name of the declaration if the
+        // parameter is a pointer or reference.
+        TPO->printAsInit(OS);
+        return;
+      }
+      VD->printName(OS);
       return;
     }
 
diff --git a/clang/lib/AST/ASTImporter.cpp b/clang/lib/AST/ASTImporter.cpp
index 100eefa5743c2..49f2f18bf33a1 100644
--- a/clang/lib/AST/ASTImporter.cpp
+++ b/clang/lib/AST/ASTImporter.cpp
@@ -397,6 +397,7 @@ namespace clang {
     Error ImportImplicitMethods(const CXXRecordDecl *From, CXXRecordDecl *To);
 
     Expected<CXXCastPath> ImportCastPath(CastExpr *E);
+    Expected<APValue> ImportAPValue(const APValue &FromValue);
 
     using Designator = DesignatedInitExpr::Designator;
 
@@ -6698,18 +6699,11 @@ ExpectedStmt ASTNodeImporter::VisitAddrLabelExpr(AddrLabelExpr *E) {
 ExpectedStmt ASTNodeImporter::VisitConstantExpr(ConstantExpr *E) {
   Error Err = Error::success();
   auto ToSubExpr = importChecked(Err, E->getSubExpr());
+  auto ToResult = importChecked(Err, E->getAPValueResult());
   if (Err)
     return std::move(Err);
 
-  // TODO : Handle APValue::ValueKind that require importing.
-
-  APValue::ValueKind Kind = E->getResultAPValueKind();
-  if (Kind == APValue::Int || Kind == APValue::Float ||
-      Kind == APValue::FixedPoint || Kind == APValue::ComplexFloat ||
-      Kind == APValue::ComplexInt)
-    return ConstantExpr::Create(Importer.getToContext(), ToSubExpr,
-                                E->getAPValueResult());
-  return ConstantExpr::Create(Importer.getToContext(), ToSubExpr);
+  return ConstantExpr::Create(Importer.getToContext(), ToSubExpr, ToResult);
 }
 ExpectedStmt ASTNodeImporter::VisitParenExpr(ParenExpr *E) {
   Error Err = Error::success();
@@ -7871,7 +7865,8 @@ ExpectedStmt ASTNodeImporter::VisitSubstNonTypeTemplateParmExpr(
     return std::move(Err);
 
   return new (Importer.getToContext()) SubstNonTypeTemplateParmExpr(
-      ToType, E->getValueKind(), ToExprLoc, ToParameter, ToReplacement);
+      ToType, E->getValueKind(), ToExprLoc, ToParameter,
+      E->isReferenceParameter(), ToReplacement);
 }
 
 ExpectedStmt ASTNodeImporter::VisitTypeTraitExpr(TypeTraitExpr *E) {
@@ -8810,6 +8805,11 @@ ASTImporter::Import(const CXXBaseSpecifier *BaseSpec) {
   return Imported;
 }
 
+llvm::Expected<APValue> ASTImporter::Import(const APValue &FromValue) {
+  ASTNodeImporter Importer(*this);
+  return Importer.ImportAPValue(FromValue);
+}
+
 Error ASTImporter::ImportDefinition(Decl *From) {
   ExpectedDecl ToOrErr = Import(From);
   if (!ToOrErr)
@@ -8940,6 +8940,172 @@ Expected<Selector> ASTImporter::Import(Selector FromSel) {
   return ToContext.Selectors.getSelector(FromSel.getNumArgs(), Idents.data());
 }
 
+llvm::Expected<APValue>
+ASTNodeImporter::ImportAPValue(const APValue &FromValue) {
+  APValue Result;
+  llvm::Error Err = llvm::Error::success();
+  auto ImportLoop = [&](const APValue *From, APValue *To, unsigned Size) {
+    for (unsigned Idx = 0; Idx < Size; Idx++) {
+      APValue Tmp = importChecked(Err, From[Idx]);
+      To[Idx] = Tmp;
+    }
+  };
+  switch (FromValue.getKind()) {
+  case APValue::None:
+  case APValue::Indeterminate:
+  case APValue::Int:
+  case APValue::Float:
+  case APValue::FixedPoint:
+  case APValue::ComplexInt:
+  case APValue::ComplexFloat:
+    Result = FromValue;
+    break;
+  case APValue::Vector: {
+    Result.MakeVector();
+    MutableArrayRef<APValue> Elts =
+        Result.setVectorUninit(FromValue.getVectorLength());
+    ImportLoop(
+        ((const APValue::Vec *)(const char *)FromValue.Data.buffer)->Elts,
+        Elts.data(), FromValue.getVectorLength());
+    break;
+  }
+  case APValue::Array:
+    Result.MakeArray(FromValue.getArrayInitializedElts(),
+                     FromValue.getArraySize());
+    ImportLoop(
+        ((const APValue::Arr *)(const char *)FromValue.Data.buffer)->Elts,
+        ((const APValue::Arr *)(const char *)Result.Data.buffer)->Elts,
+        FromValue.getArrayInitializedElts());
+    break;
+  case APValue::Struct:
+    Result.MakeStruct(FromValue.getStructNumBases(),
+                      FromValue.getStructNumFields());
+    ImportLoop(
+        ((const APValue::StructData *)(const char *)FromValue.Data.buffer)
+            ->Elts,
+        ((const APValue::StructData *)(const char *)Result.Data.buffer)->Elts,
+        FromValue.getStructNumBases() + FromValue.getStructNumFields());
+    break;
+  case APValue::Union: {
+    Result.MakeUnion();
+    const Decl *ImpFDecl = importChecked(Err, FromValue.getUnionField());
+    APValue ImpValue = importChecked(Err, FromValue.getUnionValue());
+    if (Err)
+      return std::move(Err);
+    Result.setUnion(cast<FieldDecl>(ImpFDecl), ImpValue);
+    break;
+  }
+  case APValue::AddrLabelDiff: {
+    Result.MakeAddrLabelDiff();
+    const Expr *ImpLHS = importChecked(Err, FromValue.getAddrLabelDiffLHS());
+    const Expr *ImpRHS = importChecked(Err, FromValue.getAddrLabelDiffRHS());
+    if (Err)
+      return std::move(Err);
+    Result.setAddrLabelDiff(cast<AddrLabelExpr>(ImpLHS),
+                            cast<AddrLabelExpr>(ImpRHS));
+    break;
+  }
+  case APValue::MemberPointer: {
+    const Decl *ImpMemPtrDecl =
+        importChecked(Err, FromValue.getMemberPointerDecl());
+    if (Err)
+      return std::move(Err);
+    MutableArrayRef<const CXXRecordDecl *> ToPath =
+        Result.setMemberPointerUninit(
+            cast<const ValueDecl>(ImpMemPtrDecl),
+            FromValue.isMemberPointerToDerivedMember(),
+            FromValue.getMemberPointerPath().size());
+    llvm::ArrayRef<const CXXRecordDecl *> FromPath =
+        Result.getMemberPointerPath();
+    for (unsigned Idx = 0; Idx < FromValue.getMemberPointerPath().size();
+         Idx++) {
+      const Decl *ImpDecl = importChecked(Err, FromPath[Idx]);
+      if (Err)
+        return std::move(Err);
+      ToPath[Idx] = cast<const CXXRecordDecl>(ImpDecl->getCanonicalDecl());
+    }
+    break;
+  }
+  case APValue::LValue:
+    APValue::LValueBase Base;
+    QualType FromElemTy;
+    if (FromValue.getLValueBase()) {
+      assert(!FromValue.getLValueBase().is<DynamicAllocLValue>() &&
+             "in C++20 dynamic allocation are transient so they shouldn't "
+             "appear in the AST");
+      if (!FromValue.getLValueBase().is<TypeInfoLValue>()) {
+        if (const auto *E =
+                FromValue.getLValueBase().dyn_cast<const Expr *>()) {
+          FromElemTy = E->getType();
+          const Expr *ImpExpr = importChecked(Err, E);
+          if (Err)
+            return std::move(Err);
+          Base = APValue::LValueBase(ImpExpr,
+                                     FromValue.getLValueBase().getCallIndex(),
+                                     FromValue.getLValueBase().getVersion());
+        } else {
+          FromElemTy =
+              FromValue.getLValueBase().get<const ValueDecl *>()->getType();
+          const Decl *ImpDecl = importChecked(
+              Err, FromValue.getLValueBase().get<const ValueDecl *>());
+          if (Err)
+            return std::move(Err);
+          Base = APValue::LValueBase(cast<ValueDecl>(ImpDecl),
+                                     FromValue.getLValueBase().getCallIndex(),
+                                     FromValue.getLValueBase().getVersion());
+        }
+      } else {
+        FromElemTy = FromValue.getLValueBase().getTypeInfoType();
+        QualType ImpTypeInfo = importChecked(
+            Err,
+            QualType(FromValue.getLValueBase().get<TypeInfoLValue>().getType(),
+                     0));
+        QualType ImpType =
+            importChecked(Err, FromValue.getLValueBase().getTypeInfoType());
+        if (Err)
+          return std::move(Err);
+        Base = APValue::LValueBase::getTypeInfo(
+            TypeInfoLValue(ImpTypeInfo.getTypePtr()), ImpType);
+      }
+    }
+    CharUnits Offset = FromValue.getLValueOffset();
+    unsigned PathLength = FromValue.getLValuePath().size();
+    Result.MakeLValue();
+    if (FromValue.hasLValuePath()) {
+      MutableArrayRef<APValue::LValuePathEntry> ToPath = Result.setLValueUninit(
+          Base, Offset, PathLength, FromValue.isLValueOnePastTheEnd(),
+          FromValue.isNullPointer());
+      llvm::ArrayRef<APValue::LValuePathEntry> FromPath =
+          FromValue.getLValuePath();
+      for (unsigned LoopIdx = 0; LoopIdx < PathLength; LoopIdx++) {
+        if (FromElemTy->isRecordType()) {
+          const Decl *FromDecl =
+              FromPath[LoopIdx].getAsBaseOrMember().getPointer();
+          const Decl *ImpDecl = importChecked(Err, FromDecl);
+          if (Err)
+            return std::move(Err);
+          if (auto *RD = dyn_cast<CXXRecordDecl>(FromDecl))
+            FromElemTy = Importer.FromContext.getRecordType(RD);
+          else
+            FromElemTy = cast<ValueDecl>(FromDecl)->getType();
+          ToPath[LoopIdx] = APValue::LValuePathEntry(APValue::BaseOrMemberType(
+              ImpDecl, FromPath[LoopIdx].getAsBaseOrMember().getInt()));
+        } else {
+          FromElemTy =
+              Importer.FromContext.getAsArrayType(FromElemTy)->getElementType();
+          ToPath[LoopIdx] = APValue::LValuePathEntry::ArrayIndex(
+              FromPath[LoopIdx].getAsArrayIndex());
+        }
+      }
+    } else
+      Result.setLValue(Base, Offset, APValue::NoLValuePath{},
+                       FromValue.isNullPointer());
+  }
+  if (Err)
+    return std::move(Err);
+  return Result;
+}
+
 Expected<DeclarationName> ASTImporter::HandleNameConflict(DeclarationName Name,
                                                           DeclContext *DC,
                                                           unsigned IDNS,
diff --git a/clang/lib/AST/DeclBase.cpp b/clang/lib/AST/DeclBase.cpp
index f2502c327a118..0656efae5489f 100644
--- a/clang/lib/AST/DeclBase.cpp
+++ b/clang/lib/AST/DeclBase.cpp
@@ -835,6 +835,7 @@ unsigned Decl::getIdentifierNamespaceForKind(Kind DeclKind) {
     case ExternCContext:
     case Decomposition:
     case MSGuid:
+    case TemplateParamObject:
 
     case UsingDirective:
     case BuiltinTemplate:
diff --git a/clang/lib/AST/DeclTemplate.cpp b/clang/lib/AST/DeclTemplate.cpp
index d99a9c19c506e..9918377070c3a 100644
--- a/clang/lib/AST/DeclTemplate.cpp
+++ b/clang/lib/AST/DeclTemplate.cpp
@@ -1431,3 +1431,36 @@ void TypeConstraint::print(llvm::raw_ostream &OS, PrintingPolicy Policy) const {
     OS << ">";
   }
 }
+
+TemplateParamObjectDecl *TemplateParamObjectDecl::Create(const ASTContext &C,
+                                                         QualType T,
+                                                         const APValue &V) {
+  DeclContext *DC = C.getTranslationUnitDecl();
+  auto *TPOD = new (C, DC) TemplateParamObjectDecl(DC, T, V);
+  C.addDestruction(&TPOD->Value);
+  return TPOD;
+}
+
+TemplateParamObjectDecl *
+TemplateParamObjectDecl::CreateDeserialized(ASTContext &C, unsigned ID) {
+  auto *TPOD = new (C, ID) TemplateParamObjectDecl(nullptr, QualType(), APValue());
+  C.addDestruction(&TPOD->Value);
+  return TPOD;
+}
+
+void TemplateParamObjectDecl::printName(llvm::raw_ostream &OS) const {
+  OS << "<template param ";
+  printAsExpr(OS);
+  OS << ">";
+}
+
+void TemplateParamObjectDecl::printAsExpr(llvm::raw_ostream &OS) const {
+  const ASTContext &Ctx = getASTContext();
+  getType().getUnqualifiedType().print(OS, Ctx.getPrintingPolicy());
+  printAsInit(OS);
+}
+
+void TemplateParamObjectDecl::printAsInit(llvm::raw_ostream &OS) const {
+  const ASTContext &Ctx = getASTContext();
+  getValue().printPretty(OS, Ctx, getType());
+}
diff --git a/clang/lib/AST/Expr.cpp b/clang/lib/AST/Expr.cpp
index 5ab375462cb3a..991f839d2522e 100644
--- a/clang/lib/AST/Expr.cpp
+++ b/clang/lib/AST/Expr.cpp
@@ -360,7 +360,6 @@ llvm::APSInt ConstantExpr::getResultAsAPSInt() const {
 }
 
 APValue ConstantExpr::getAPValueResult() const {
-  assert(hasAPValueResult());
 
   switch (ConstantExprBits.ResultKind) {
   case ConstantExpr::RSK_APValue:
@@ -370,6 +369,8 @@ APValue ConstantExpr::getAPValueResult() const {
         llvm::APSInt(llvm::APInt(ConstantExprBits.BitWidth, Int64Result()),
                      ConstantExprBits.IsUnsigned));
   case ConstantExpr::RSK_None:
+    if (ConstantExprBits.APValueKind == APValue::Indeterminate)
+      return APValue::IndeterminateValue();
     return APValue();
   }
   llvm_unreachable("invalid ResultKind");
diff --git a/clang/lib/AST/ExprCXX.cpp b/clang/lib/AST/ExprCXX.cpp
index 4a421f03e589e..c1ec86075772f 100644
--- a/clang/lib/AST/ExprCXX.cpp
+++ b/clang/lib/AST/ExprCXX.cpp
@@ -1565,6 +1565,15 @@ SizeOfPackExpr *SizeOfPackExpr::CreateDeserialized(ASTContext &Context,
   return new (Storage) SizeOfPackExpr(EmptyShell(), NumPartialArgs);
 }
 
+QualType SubstNonTypeTemplateParmExpr::getParameterType(
+    const ASTContext &Context) const {
+  // Note that, for a class type NTTP, we will have an lvalue of type 'const
+  // T', so we can't just compute this from the type and value category.
+  if (isReferenceParameter())
+    return Context.getLValueReferenceType(getType());
+  return getType().getUnqualifiedType();
+}
+
 SubstNonTypeTemplateParmPackExpr::SubstNonTypeTemplateParmPackExpr(
     QualType T, ExprValueKind ValueKind, NonTypeTemplateParmDecl *Param,
     SourceLocation NameLoc, const TemplateArgument &ArgPack)
diff --git a/clang/lib/AST/ExprClassification.cpp b/clang/lib/AST/ExprClassification.cpp
index 31aa734ffedba..0286c730ce4e3 100644
--- a/clang/lib/AST/ExprClassification.cpp
+++ b/clang/lib/AST/ExprClassification.cpp
@@ -453,12 +453,14 @@ static Cl::Kinds ClassifyDecl(ASTContext &Ctx, const Decl *D) {
 
   bool islvalue;
   if (const auto *NTTParm = dyn_cast<NonTypeTemplateParmDecl>(D))
-    islvalue = NTTParm->getType()->isReferenceType();
+    islvalue = NTTParm->getType()->isReferenceType() ||
+               NTTParm->getType()->isRecordType();
   else
     islvalue = isa<VarDecl>(D) || isa<FieldDecl>(D) ||
                isa<IndirectFieldDecl>(D) ||
                isa<BindingDecl>(D) ||
                isa<MSGuidDecl>(D) ||
+               isa<TemplateParamObjectDecl>(D) ||
                (Ctx.getLangOpts().CPlusPlus &&
                 (isa<FunctionDecl>(D) || isa<MSPropertyDecl>(D) ||
                  isa<FunctionTemplateDecl>(D)));
diff --git a/clang/lib/AST/ExprConstant.cpp b/clang/lib/AST/ExprConstant.cpp
index e736285472579..931a47f9ff737 100644
--- a/clang/lib/AST/ExprConstant.cpp
+++ b/clang/lib/AST/ExprConstant.cpp
@@ -1603,7 +1603,7 @@ namespace {
     }
 
     void setNull(ASTContext &Ctx, QualType PointerTy) {
-      Base = (Expr *)nullptr;
+      Base = (const ValueDecl *)nullptr;
       Offset =
           CharUnits::fromQuantity(Ctx.getTargetNullPointerValue(PointerTy));
       InvalidBase = false;
@@ -1978,6 +1978,8 @@ static bool IsGlobalLValue(APValue::LValueBase B) {
     // ... the address of an object with static storage duration,
     if (const VarDecl *VD = dyn_cast<VarDecl>(D))
       return VD->hasGlobalStorage();
+    if (isa<TemplateParamObjectDecl>(D))
+      return true;
     // ... the address of a function,
     // ... the address of a GUID [MS extension],
     return isa<FunctionDecl>(D) || isa<MSGuidDecl>(D);
@@ -3980,6 +3982,16 @@ static CompleteObject findCompleteObject(EvalInfo &Info, const Expr *E,
       return CompleteObject(LVal.Base, &V, GD->getType());
     }
 
+    // Allow reading from template parameter objects.
+    if (auto *TPO = dyn_cast<TemplateParamObjectDecl>(D)) {
+      if (isModification(AK)) {
+        Info.FFDiag(E, diag::note_constexpr_modify_global);
+        return CompleteObject();
+      }
+      return CompleteObject(LVal.Base, const_cast<APValue *>(&TPO->getValue()),
+                            TPO->getType());
+    }
+
     // In C++98, const, non-volatile integers initialized with ICEs are ICEs.
     // In C++11, constexpr, non-volatile variables initialized with constant
     // expressions are constant expressions too. Inside constexpr functions,
@@ -8026,14 +8038,13 @@ static bool EvaluateLValue(const Expr *E, LValue &Result, EvalInfo &Info,
 }
 
 bool LValueExprEvaluator::VisitDeclRefExpr(const DeclRefExpr *E) {
-  if (const FunctionDecl *FD = dyn_cast<FunctionDecl>(E->getDecl()))
-    return Success(FD);
-  if (const VarDecl *VD = dyn_cast<VarDecl>(E->getDecl()))
+  const NamedDecl *D = E->getDecl();
+  if (isa<FunctionDecl, MSGuidDecl, TemplateParamObjectDecl>(D))
+    return Success(cast<ValueDecl>(D));
+  if (const VarDecl *VD = dyn_cast<VarDecl>(D))
     return VisitVarDecl(E, VD);
-  if (const BindingDecl *BD = dyn_cast<BindingDecl>(E->getDecl()))
+  if (const BindingDecl *BD = dyn_cast<BindingDecl>(D))
     return Visit(BD->getBinding());
-  if (const MSGuidDecl *GD = dyn_cast<MSGuidDecl>(E->getDecl()))
-    return Success(GD);
   return Error(E);
 }
 
@@ -9596,7 +9607,7 @@ static bool HandleClassZeroInitialization(EvalInfo &Info, const Expr *E,
 
   for (const auto *I : RD->fields()) {
     // -- if T is a reference type, no initialization is performed.
-    if (I->getType()->isReferenceType())
+    if (I->isUnnamedBitfield() || I->getType()->isReferenceType())
       continue;
 
     LValue Subobject = This;
diff --git a/clang/lib/AST/ItaniumMangle.cpp b/clang/lib/AST/ItaniumMangle.cpp
index f2f25a669675c..ea53c9fb8fb3d 100644
--- a/clang/lib/AST/ItaniumMangle.cpp
+++ b/clang/lib/AST/ItaniumMangle.cpp
@@ -661,7 +661,12 @@ void CXXNameMangler::mangle(GlobalDecl GD) {
     mangleName(GuidD);
   else if (const BindingDecl *BD = dyn_cast<BindingDecl>(GD.getDecl()))
     mangleName(BD);
-  else
+  else if (isa<TemplateParamObjectDecl>(GD.getDecl())) {
+    DiagnosticsEngine &Diags = Context.getDiags();
+    unsigned DiagID = Diags.getCustomDiagID(DiagnosticsEngine::Error,
+      "cannot mangle template parameter objects yet");
+    Diags.Report(SourceLocation(), DiagID);
+  } else
     llvm_unreachable("unexpected kind of global decl");
 }
 
diff --git a/clang/lib/AST/MicrosoftMangle.cpp b/clang/lib/AST/MicrosoftMangle.cpp
index 814780a611b50..bb89683f91f73 100644
--- a/clang/lib/AST/MicrosoftMangle.cpp
+++ b/clang/lib/AST/MicrosoftMangle.cpp
@@ -503,7 +503,12 @@ void MicrosoftCXXNameMangler::mangle(const NamedDecl *D, StringRef Prefix) {
     // MSVC appears to mangle GUIDs as if they were variables of type
     // 'const struct __s_GUID'.
     Out << "3U__s_GUID@@B";
-  else
+  else if (isa<TemplateParamObjectDecl>(D)) {
+    DiagnosticsEngine &Diags = Context.getDiags();
+    unsigned DiagID = Diags.getCustomDiagID(DiagnosticsEngine::Error,
+      "cannot mangle template parameter objects yet");
+    Diags.Report(SourceLocation(), DiagID);
+  } else
     llvm_unreachable("Tried to mangle unexpected NamedDecl!");
 }
 
diff --git a/clang/lib/AST/StmtPrinter.cpp b/clang/lib/AST/StmtPrinter.cpp
index 1390876dc0df4..55a721194ccf9 100644
--- a/clang/lib/AST/StmtPrinter.cpp
+++ b/clang/lib/AST/StmtPrinter.cpp
@@ -970,6 +970,10 @@ void StmtPrinter::VisitDeclRefExpr(DeclRefExpr *Node) {
     OCED->getInit()->IgnoreImpCasts()->printPretty(OS, nullptr, Policy);
     return;
   }
+  if (const auto *TPOD = dyn_cast<TemplateParamObjectDecl>(Node->getDecl())) {
+    TPOD->printAsExpr(OS);
+    return;
+  }
   if (NestedNameSpecifier *Qualifier = Node->getQualifier())
     Qualifier->print(OS, Policy);
   if (Node->hasTemplateKeyword())
diff --git a/clang/lib/AST/TemplateBase.cpp b/clang/lib/AST/TemplateBase.cpp
index 891c9bd0fd7a1..22af2545779da 100644
--- a/clang/lib/AST/TemplateBase.cpp
+++ b/clang/lib/AST/TemplateBase.cpp
@@ -352,6 +352,13 @@ void TemplateArgument::print(const PrintingPolicy &Policy,
 
   case Declaration: {
     NamedDecl *ND = getAsDecl();
+    if (getParamTypeForDecl()->isRecordType()) {
+      if (auto *TPO = dyn_cast<TemplateParamObjectDecl>(ND)) {
+        // FIXME: Include the type if it's not obvious from the context.
+        TPO->printAsInit(Out);
+        break;
+      }
+    }
     if (!getParamTypeForDecl()->isReferenceType())
       Out << '&';
     ND->printQualifiedName(Out);
diff --git a/clang/lib/Basic/Targets/AMDGPU.h b/clang/lib/Basic/Targets/AMDGPU.h
index a0e4c19f4fea3..2158a8d253d6e 100644
--- a/clang/lib/Basic/Targets/AMDGPU.h
+++ b/clang/lib/Basic/Targets/AMDGPU.h
@@ -285,7 +285,6 @@ class LLVM_LIBRARY_VISIBILITY AMDGPUTargetInfo final : public TargetInfo {
   void setSupportedOpenCLOpts() override {
     auto &Opts = getSupportedOpenCLOpts();
     Opts.support("cl_clang_storage_class_specifiers");
-    Opts.support("cl_khr_icd");
 
     bool IsAMDGCN = isAMDGCN(getTriple());
 
diff --git a/clang/lib/Basic/Targets/NVPTX.h b/clang/lib/Basic/Targets/NVPTX.h
index dd5360ac0b368..8fc25a6744fc8 100644
--- a/clang/lib/Basic/Targets/NVPTX.h
+++ b/clang/lib/Basic/Targets/NVPTX.h
@@ -128,8 +128,6 @@ class LLVM_LIBRARY_VISIBILITY NVPTXTargetInfo : public TargetInfo {
   void setSupportedOpenCLOpts() override {
     auto &Opts = getSupportedOpenCLOpts();
     Opts.support("cl_clang_storage_class_specifiers");
-    Opts.support("cl_khr_gl_sharing");
-    Opts.support("cl_khr_icd");
 
     Opts.support("cl_khr_fp64");
     Opts.support("cl_khr_byte_addressable_store");
diff --git a/clang/lib/Basic/Targets/X86.cpp b/clang/lib/Basic/Targets/X86.cpp
index 9b607a3b39413..c8d96f887e904 100644
--- a/clang/lib/Basic/Targets/X86.cpp
+++ b/clang/lib/Basic/Targets/X86.cpp
@@ -310,6 +310,8 @@ bool X86TargetInfo::handleTargetFeatures(std::vector<std::string> &Features,
       HasSERIALIZE = true;
     } else if (Feature == "+tsxldtrk") {
       HasTSXLDTRK = true;
+    } else if (Feature == "+uintr") {
+      HasUINTR = true;
     }
 
     X86SSEEnum Level = llvm::StringSwitch<X86SSEEnum>(Feature)
@@ -726,6 +728,8 @@ void X86TargetInfo::getTargetDefines(const LangOptions &Opts,
     Builder.defineMacro("__SERIALIZE__");
   if (HasTSXLDTRK)
     Builder.defineMacro("__TSXLDTRK__");
+  if (HasUINTR)
+    Builder.defineMacro("__UINTR__");
 
   // Each case falls through to the previous one here.
   switch (SSELevel) {
@@ -889,6 +893,7 @@ bool X86TargetInfo::isValidFeatureName(StringRef Name) const {
       .Case("sse4a", true)
       .Case("tbm", true)
       .Case("tsxldtrk", true)
+      .Case("uintr", true)
       .Case("vaes", true)
       .Case("vpclmulqdq", true)
       .Case("wbnoinvd", true)
@@ -980,6 +985,7 @@ bool X86TargetInfo::hasFeature(StringRef Feature) const {
       .Case("sse4a", XOPLevel >= SSE4A)
       .Case("tbm", HasTBM)
       .Case("tsxldtrk", HasTSXLDTRK)
+      .Case("uintr", HasUINTR)
       .Case("vaes", HasVAES)
       .Case("vpclmulqdq", HasVPCLMULQDQ)
       .Case("wbnoinvd", HasWBNOINVD)
diff --git a/clang/lib/Basic/Targets/X86.h b/clang/lib/Basic/Targets/X86.h
index 441ab961e2937..5f93385808570 100644
--- a/clang/lib/Basic/Targets/X86.h
+++ b/clang/lib/Basic/Targets/X86.h
@@ -135,6 +135,7 @@ class LLVM_LIBRARY_VISIBILITY X86TargetInfo : public TargetInfo {
   bool HasAMXBF16 = false;
   bool HasSERIALIZE = false;
   bool HasTSXLDTRK = false;
+  bool HasUINTR = false;
 
 protected:
   llvm::X86::CPUKind CPU = llvm::X86::CK_None;
diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
index 52724d31a3608..b4ce791ee748b 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -523,6 +523,14 @@ static void initTargetOptions(DiagnosticsEngine &Diags,
   Options.UniqueSectionNames = CodeGenOpts.UniqueSectionNames;
   Options.UniqueBasicBlockSectionNames =
       CodeGenOpts.UniqueBasicBlockSectionNames;
+  Options.StackProtectorGuard =
+      llvm::StringSwitch<llvm::StackProtectorGuards>(CodeGenOpts
+          .StackProtectorGuard)
+          .Case("tls", llvm::StackProtectorGuards::TLS)
+          .Case("global", llvm::StackProtectorGuards::Global)
+          .Default(llvm::StackProtectorGuards::None);
+  Options.StackProtectorGuardOffset = CodeGenOpts.StackProtectorGuardOffset;
+  Options.StackProtectorGuardReg = CodeGenOpts.StackProtectorGuardReg;
   Options.TLSSize = CodeGenOpts.TLSSize;
   Options.EmulatedTLS = CodeGenOpts.EmulatedTLS;
   Options.ExplicitEmulatedTLS = CodeGenOpts.ExplicitEmulatedTLS;
diff --git a/clang/lib/CodeGen/CGDecl.cpp b/clang/lib/CodeGen/CGDecl.cpp
index 87070a3504a50..168311b066598 100644
--- a/clang/lib/CodeGen/CGDecl.cpp
+++ b/clang/lib/CodeGen/CGDecl.cpp
@@ -118,6 +118,7 @@ void CodeGenFunction::EmitDecl(const Decl &D) {
   case Decl::Label:        // __label__ x;
   case Decl::Import:
   case Decl::MSGuid:    // __declspec(uuid("..."))
+  case Decl::TemplateParamObject:
   case Decl::OMPThreadPrivate:
   case Decl::OMPAllocate:
   case Decl::OMPCapturedExpr:
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 5311a2051b700..81b6d8b55b878 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -2836,6 +2836,10 @@ LValue CodeGenFunction::EmitDeclRefLValue(const DeclRefExpr *E) {
     return MakeAddrLValue(CGM.GetAddrOfMSGuidDecl(GD), T,
                           AlignmentSource::Decl);
 
+  if (const auto *TPO = dyn_cast<TemplateParamObjectDecl>(ND))
+    return MakeAddrLValue(CGM.GetAddrOfTemplateParamObject(TPO), T,
+                          AlignmentSource::Decl);
+
   llvm_unreachable("Unhandled DeclRefExpr");
 }
 
diff --git a/clang/lib/CodeGen/CGExprConstant.cpp b/clang/lib/CodeGen/CGExprConstant.cpp
index 011d39cfa2947..ca1d3a937fa87 100644
--- a/clang/lib/CodeGen/CGExprConstant.cpp
+++ b/clang/lib/CodeGen/CGExprConstant.cpp
@@ -1905,6 +1905,9 @@ ConstantLValueEmitter::tryEmitBase(const APValue::LValueBase &base) {
     if (auto *GD = dyn_cast<MSGuidDecl>(D))
       return CGM.GetAddrOfMSGuidDecl(GD);
 
+    if (auto *TPO = dyn_cast<TemplateParamObjectDecl>(D))
+      return CGM.GetAddrOfTemplateParamObject(TPO);
+
     return nullptr;
   }
 
diff --git a/clang/lib/CodeGen/CGStmt.cpp b/clang/lib/CodeGen/CGStmt.cpp
index 44790a092b4d3..e19eb11df43b2 100644
--- a/clang/lib/CodeGen/CGStmt.cpp
+++ b/clang/lib/CodeGen/CGStmt.cpp
@@ -21,6 +21,7 @@
 #include "clang/Basic/PrettyStackTrace.h"
 #include "clang/Basic/SourceManager.h"
 #include "clang/Basic/TargetInfo.h"
+#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/StringExtras.h"
 #include "llvm/IR/DataLayout.h"
 #include "llvm/IR/InlineAsm.h"
@@ -1976,7 +1977,8 @@ SimplifyConstraint(const char *Constraint, const TargetInfo &Target,
 static std::string
 AddVariableConstraints(const std::string &Constraint, const Expr &AsmExpr,
                        const TargetInfo &Target, CodeGenModule &CGM,
-                       const AsmStmt &Stmt, const bool EarlyClobber) {
+                       const AsmStmt &Stmt, const bool EarlyClobber,
+                       std::string *GCCReg = nullptr) {
   const DeclRefExpr *AsmDeclRef = dyn_cast<DeclRefExpr>(&AsmExpr);
   if (!AsmDeclRef)
     return Constraint;
@@ -2001,6 +2003,8 @@ AddVariableConstraints(const std::string &Constraint, const Expr &AsmExpr,
   }
   // Canonicalize the register here before returning it.
   Register = Target.getNormalizedGCCRegisterName(Register);
+  if (GCCReg != nullptr)
+    *GCCReg = Register.str();
   return (EarlyClobber ? "&{" : "{") + Register.str() + "}";
 }
 
@@ -2199,6 +2203,9 @@ void CodeGenFunction::EmitAsmStmt(const AsmStmt &S) {
   // Keep track of out constraints for tied input operand.
   std::vector<std::string> OutputConstraints;
 
+  // Keep track of defined physregs.
+  llvm::SmallSet<std::string, 8> PhysRegOutputs;
+
   // An inline asm can be marked readonly if it meets the following conditions:
   //  - it doesn't have any sideeffects
   //  - it doesn't clobber memory
@@ -2218,9 +2225,15 @@ void CodeGenFunction::EmitAsmStmt(const AsmStmt &S) {
     const Expr *OutExpr = S.getOutputExpr(i);
     OutExpr = OutExpr->IgnoreParenNoopCasts(getContext());
 
+    std::string GCCReg;
     OutputConstraint = AddVariableConstraints(OutputConstraint, *OutExpr,
                                               getTarget(), CGM, S,
-                                              Info.earlyClobber());
+                                              Info.earlyClobber(),
+                                              &GCCReg);
+    // Give an error on multiple outputs to same physreg.
+    if (!GCCReg.empty() && !PhysRegOutputs.insert(GCCReg).second)
+      CGM.Error(S.getAsmLoc(), "multiple outputs to hard register: " + GCCReg);
+
     OutputConstraints.push_back(OutputConstraint);
     LValue Dest = EmitLValue(OutExpr);
     if (!Constraints.empty())
@@ -2307,7 +2320,8 @@ void CodeGenFunction::EmitAsmStmt(const AsmStmt &S) {
         LargestVectorWidth =
             std::max((uint64_t)LargestVectorWidth,
                      VT->getPrimitiveSizeInBits().getKnownMinSize());
-      if (Info.allowsRegister())
+      // Only tie earlyclobber physregs.
+      if (Info.allowsRegister() && (GCCReg.empty() || Info.earlyClobber()))
         InOutConstraints += llvm::utostr(i);
       else
         InOutConstraints += OutputConstraint;
diff --git a/clang/lib/CodeGen/CodeGenModule.cpp b/clang/lib/CodeGen/CodeGenModule.cpp
index 325dd4cf95f62..9199ea141165f 100644
--- a/clang/lib/CodeGen/CodeGenModule.cpp
+++ b/clang/lib/CodeGen/CodeGenModule.cpp
@@ -2640,6 +2640,12 @@ ConstantAddress CodeGenModule::GetAddrOfMSGuidDecl(const MSGuidDecl *GD) {
   return ConstantAddress(Addr, Alignment);
 }
 
+ConstantAddress CodeGenModule::GetAddrOfTemplateParamObject(
+    const TemplateParamObjectDecl *TPO) {
+  ErrorUnsupported(TPO, "template parameter object");
+  return ConstantAddress::invalid();
+}
+
 ConstantAddress CodeGenModule::GetWeakRefReference(const ValueDecl *VD) {
   const AliasAttr *AA = VD->getAttr<AliasAttr>();
   assert(AA && "No alias?");
diff --git a/clang/lib/CodeGen/CodeGenModule.h b/clang/lib/CodeGen/CodeGenModule.h
index d5b388c12e76c..45f54c72ea08f 100644
--- a/clang/lib/CodeGen/CodeGenModule.h
+++ b/clang/lib/CodeGen/CodeGenModule.h
@@ -873,6 +873,10 @@ class CodeGenModule : public CodeGenTypeCache {
   /// Get the address of a GUID.
   ConstantAddress GetAddrOfMSGuidDecl(const MSGuidDecl *GD);
 
+  /// Get the address of a template parameter object.
+  ConstantAddress
+  GetAddrOfTemplateParamObject(const TemplateParamObjectDecl *TPO);
+
   /// Get the address of the thunk for the given global decl.
   llvm::Constant *GetAddrOfThunk(StringRef Name, llvm::Type *FnTy,
                                  GlobalDecl GD);
diff --git a/clang/lib/Driver/ToolChain.cpp b/clang/lib/Driver/ToolChain.cpp
index e1057c581b19e..4350a0c49ae93 100644
--- a/clang/lib/Driver/ToolChain.cpp
+++ b/clang/lib/Driver/ToolChain.cpp
@@ -845,6 +845,37 @@ std::string ToolChain::ComputeLLVMTriple(const ArgList &Args,
     }
     Triple.setArchName(ArchName + Suffix.str());
 
+    bool isHardFloat =
+        (arm::getARMFloatABI(getDriver(), Triple, Args) == arm::FloatABI::Hard);
+    switch (Triple.getEnvironment()) {
+    case Triple::GNUEABI:
+    case Triple::GNUEABIHF:
+      Triple.setEnvironment(isHardFloat ? Triple::GNUEABIHF : Triple::GNUEABI);
+      break;
+    case Triple::EABI:
+    case Triple::EABIHF:
+      Triple.setEnvironment(isHardFloat ? Triple::EABIHF : Triple::EABI);
+      break;
+    case Triple::MuslEABI:
+    case Triple::MuslEABIHF:
+      Triple.setEnvironment(isHardFloat ? Triple::MuslEABIHF
+                                        : Triple::MuslEABI);
+      break;
+    default: {
+      arm::FloatABI DefaultABI = arm::getDefaultFloatABI(Triple);
+      if (DefaultABI != arm::FloatABI::Invalid &&
+          isHardFloat != (DefaultABI == arm::FloatABI::Hard)) {
+        Arg *ABIArg =
+            Args.getLastArg(options::OPT_msoft_float, options::OPT_mhard_float,
+                            options::OPT_mfloat_abi_EQ);
+        assert(ABIArg && "Non-default float abi expected to be from arg");
+        D.Diag(diag::err_drv_unsupported_opt_for_target)
+            << ABIArg->getAsString(Args) << Triple.getTriple();
+      }
+      break;
+    }
+    }
+
     return Triple.getTriple();
   }
   }
diff --git a/clang/lib/Driver/ToolChains/Arch/AArch64.cpp b/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
index fe742b4bcfcda..06bb705a37214 100644
--- a/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
+++ b/clang/lib/Driver/ToolChains/Arch/AArch64.cpp
@@ -43,6 +43,12 @@ std::string aarch64::getAArch64TargetCPU(const ArgList &Args,
   else if (CPU.size())
     return CPU;
 
+  if (Triple.isTargetMachineMac() &&
+      Triple.getArch() == llvm::Triple::aarch64) {
+    // Apple Silicon macs default to A12 CPUs.
+    return "apple-a12";
+  }
+
   // Make sure we pick the appropriate Apple CPU if -arch is used or when
   // targetting a Darwin OS.
   if (Args.getLastArg(options::OPT_arch) || Triple.isOSDarwin())
diff --git a/clang/lib/Driver/ToolChains/Arch/ARM.cpp b/clang/lib/Driver/ToolChains/Arch/ARM.cpp
index d74d5db0c083d..309a7298300f2 100644
--- a/clang/lib/Driver/ToolChains/Arch/ARM.cpp
+++ b/clang/lib/Driver/ToolChains/Arch/ARM.cpp
@@ -134,6 +134,7 @@ bool arm::useAAPCSForMachO(const llvm::Triple &T) {
   // The backend is hardwired to assume AAPCS for M-class processors, ensure
   // the frontend matches that.
   return T.getEnvironment() == llvm::Triple::EABI ||
+         T.getEnvironment() == llvm::Triple::EABIHF ||
          T.getOS() == llvm::Triple::UnknownOS || isARMMProfile(T);
 }
 
@@ -160,11 +161,73 @@ arm::FloatABI arm::getARMFloatABI(const ToolChain &TC, const ArgList &Args) {
   return arm::getARMFloatABI(TC.getDriver(), TC.getEffectiveTriple(), Args);
 }
 
+arm::FloatABI arm::getDefaultFloatABI(const llvm::Triple &Triple) {
+  auto SubArch = getARMSubArchVersionNumber(Triple);
+  switch (Triple.getOS()) {
+  case llvm::Triple::Darwin:
+  case llvm::Triple::MacOSX:
+  case llvm::Triple::IOS:
+  case llvm::Triple::TvOS:
+    // Darwin defaults to "softfp" for v6 and v7.
+    if (Triple.isWatchABI())
+      return FloatABI::Hard;
+    else
+      return (SubArch == 6 || SubArch == 7) ? FloatABI::SoftFP : FloatABI::Soft;
+
+  case llvm::Triple::WatchOS:
+    return FloatABI::Hard;
+
+  // FIXME: this is invalid for WindowsCE
+  case llvm::Triple::Win32:
+    return FloatABI::Hard;
+
+  case llvm::Triple::NetBSD:
+    switch (Triple.getEnvironment()) {
+    case llvm::Triple::EABIHF:
+    case llvm::Triple::GNUEABIHF:
+      return FloatABI::Hard;
+    default:
+      return FloatABI::Soft;
+    }
+    break;
+
+  case llvm::Triple::FreeBSD:
+    switch (Triple.getEnvironment()) {
+    case llvm::Triple::GNUEABIHF:
+      return FloatABI::Hard;
+    default:
+      // FreeBSD defaults to soft float
+      return FloatABI::Soft;
+    }
+    break;
+
+  case llvm::Triple::OpenBSD:
+    return FloatABI::SoftFP;
+
+  default:
+    switch (Triple.getEnvironment()) {
+    case llvm::Triple::GNUEABIHF:
+    case llvm::Triple::MuslEABIHF:
+    case llvm::Triple::EABIHF:
+      return FloatABI::Hard;
+    case llvm::Triple::GNUEABI:
+    case llvm::Triple::MuslEABI:
+    case llvm::Triple::EABI:
+      // EABI is always AAPCS, and if it was not marked 'hard', it's softfp
+      return FloatABI::SoftFP;
+    case llvm::Triple::Android:
+      return (SubArch >= 7) ? FloatABI::SoftFP : FloatABI::Soft;
+    default:
+      return FloatABI::Invalid;
+    }
+  }
+  return FloatABI::Invalid;
+}
+
 // Select the float ABI as determined by -msoft-float, -mhard-float, and
 // -mfloat-abi=.
 arm::FloatABI arm::getARMFloatABI(const Driver &D, const llvm::Triple &Triple,
                                   const ArgList &Args) {
-  auto SubArch = getARMSubArchVersionNumber(Triple);
   arm::FloatABI ABI = FloatABI::Invalid;
   if (Arg *A =
           Args.getLastArg(options::OPT_msoft_float, options::OPT_mhard_float,
@@ -184,95 +247,23 @@ arm::FloatABI arm::getARMFloatABI(const Driver &D, const llvm::Triple &Triple,
         ABI = FloatABI::Soft;
       }
     }
-
-    // It is incorrect to select hard float ABI on MachO platforms if the ABI is
-    // "apcs-gnu".
-    if (Triple.isOSBinFormatMachO() && !useAAPCSForMachO(Triple) &&
-        ABI == FloatABI::Hard) {
-      D.Diag(diag::err_drv_unsupported_opt_for_target) << A->getAsString(Args)
-                                                       << Triple.getArchName();
-    }
   }
 
   // If unspecified, choose the default based on the platform.
-  if (ABI == FloatABI::Invalid) {
-    switch (Triple.getOS()) {
-    case llvm::Triple::Darwin:
-    case llvm::Triple::MacOSX:
-    case llvm::Triple::IOS:
-    case llvm::Triple::TvOS: {
-      // Darwin defaults to "softfp" for v6 and v7.
-      ABI = (SubArch == 6 || SubArch == 7) ? FloatABI::SoftFP : FloatABI::Soft;
-      ABI = Triple.isWatchABI() ? FloatABI::Hard : ABI;
-      break;
-    }
-    case llvm::Triple::WatchOS:
-      ABI = FloatABI::Hard;
-      break;
+  if (ABI == FloatABI::Invalid)
+    ABI = arm::getDefaultFloatABI(Triple);
 
-    // FIXME: this is invalid for WindowsCE
-    case llvm::Triple::Win32:
+  if (ABI == FloatABI::Invalid) {
+    // Assume "soft", but warn the user we are guessing.
+    if (Triple.isOSBinFormatMachO() &&
+        Triple.getSubArch() == llvm::Triple::ARMSubArch_v7em)
       ABI = FloatABI::Hard;
-      break;
-
-    case llvm::Triple::NetBSD:
-      switch (Triple.getEnvironment()) {
-      case llvm::Triple::EABIHF:
-      case llvm::Triple::GNUEABIHF:
-        ABI = FloatABI::Hard;
-        break;
-      default:
-        ABI = FloatABI::Soft;
-        break;
-      }
-      break;
-
-    case llvm::Triple::FreeBSD:
-      switch (Triple.getEnvironment()) {
-      case llvm::Triple::GNUEABIHF:
-        ABI = FloatABI::Hard;
-        break;
-      default:
-        // FreeBSD defaults to soft float
-        ABI = FloatABI::Soft;
-        break;
-      }
-      break;
-
-    case llvm::Triple::OpenBSD:
-      ABI = FloatABI::SoftFP;
-      break;
+    else
+      ABI = FloatABI::Soft;
 
-    default:
-      switch (Triple.getEnvironment()) {
-      case llvm::Triple::GNUEABIHF:
-      case llvm::Triple::MuslEABIHF:
-      case llvm::Triple::EABIHF:
-        ABI = FloatABI::Hard;
-        break;
-      case llvm::Triple::GNUEABI:
-      case llvm::Triple::MuslEABI:
-      case llvm::Triple::EABI:
-        // EABI is always AAPCS, and if it was not marked 'hard', it's softfp
-        ABI = FloatABI::SoftFP;
-        break;
-      case llvm::Triple::Android:
-        ABI = (SubArch >= 7) ? FloatABI::SoftFP : FloatABI::Soft;
-        break;
-      default:
-        // Assume "soft", but warn the user we are guessing.
-        if (Triple.isOSBinFormatMachO() &&
-            Triple.getSubArch() == llvm::Triple::ARMSubArch_v7em)
-          ABI = FloatABI::Hard;
-        else
-          ABI = FloatABI::Soft;
-
-        if (Triple.getOS() != llvm::Triple::UnknownOS ||
-            !Triple.isOSBinFormatMachO())
-          D.Diag(diag::warn_drv_assuming_mfloat_abi_is) << "soft";
-        break;
-      }
-    }
+    if (Triple.getOS() != llvm::Triple::UnknownOS ||
+        !Triple.isOSBinFormatMachO())
+      D.Diag(diag::warn_drv_assuming_mfloat_abi_is) << "soft";
   }
 
   assert(ABI != FloatABI::Invalid && "must select an ABI");
diff --git a/clang/lib/Driver/ToolChains/Arch/ARM.h b/clang/lib/Driver/ToolChains/Arch/ARM.h
index 0ba1a59852aa0..091c09b160ae8 100644
--- a/clang/lib/Driver/ToolChains/Arch/ARM.h
+++ b/clang/lib/Driver/ToolChains/Arch/ARM.h
@@ -47,6 +47,7 @@ enum class FloatABI {
   Hard,
 };
 
+FloatABI getDefaultFloatABI(const llvm::Triple &Triple);
 FloatABI getARMFloatABI(const ToolChain &TC, const llvm::opt::ArgList &Args);
 FloatABI getARMFloatABI(const Driver &D, const llvm::Triple &Triple,
                         const llvm::opt::ArgList &Args);
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 7c7c1ddf0eeca..2c338655bb547 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -3027,8 +3027,9 @@ static void RenderAnalyzerOptions(const ArgList &Args, ArgStringList &CmdArgs,
   Args.AddAllArgValues(CmdArgs, options::OPT_Xanalyzer);
 }
 
-static void RenderSSPOptions(const ToolChain &TC, const ArgList &Args,
-                             ArgStringList &CmdArgs, bool KernelOrKext) {
+static void RenderSSPOptions(const Driver &D, const ToolChain &TC,
+                             const ArgList &Args, ArgStringList &CmdArgs,
+                             bool KernelOrKext) {
   const llvm::Triple &EffectiveTriple = TC.getEffectiveTriple();
 
   // NVPTX doesn't support stack protectors; from the compiler's perspective, it
@@ -3073,6 +3074,50 @@ static void RenderSSPOptions(const ToolChain &TC, const ArgList &Args,
       A->claim();
     }
   }
+
+  // First support "tls" and "global" for X86 target.
+  // TODO: Support "sysreg" for AArch64.
+  const std::string &TripleStr = EffectiveTriple.getTriple();
+  if (Arg *A = Args.getLastArg(options::OPT_mstack_protector_guard_EQ)) {
+    StringRef Value = A->getValue();
+    if (!EffectiveTriple.isX86() && !EffectiveTriple.isAArch64())
+      D.Diag(diag::err_drv_unsupported_opt_for_target)
+          << A->getAsString(Args) << TripleStr;
+    if (Value != "tls" && Value != "global") {
+      D.Diag(diag::err_drv_invalid_value_with_suggestion)
+      << A->getOption().getName() << Value
+      << "valid arguments to '-mstack-protector-guard=' are:tls global";
+      return;
+    }
+    A->render(Args, CmdArgs);
+  }
+
+  if (Arg *A = Args.getLastArg(options::OPT_mstack_protector_guard_offset_EQ)) {
+    StringRef Value = A->getValue();
+    if (!EffectiveTriple.isX86())
+      D.Diag(diag::err_drv_unsupported_opt_for_target)
+          << A->getAsString(Args) << TripleStr;
+    unsigned Offset;
+    if (Value.getAsInteger(10, Offset)) {
+      D.Diag(diag::err_drv_invalid_value) << A->getOption().getName() << Value;
+      return;
+    }
+    A->render(Args, CmdArgs);
+  }
+
+  if (Arg *A = Args.getLastArg(options::OPT_mstack_protector_guard_reg_EQ)) {
+    StringRef Value = A->getValue();
+    if (!EffectiveTriple.isX86())
+      D.Diag(diag::err_drv_unsupported_opt_for_target)
+          << A->getAsString(Args) << TripleStr;
+    if (EffectiveTriple.isX86() && (Value != "fs" && Value != "gs")) {
+      D.Diag(diag::err_drv_invalid_value_with_suggestion)
+      << A->getOption().getName() << Value
+      << "for X86, valid arguments to '-mstack-protector-guard-reg=' are:fs gs";
+      return;
+    }
+    A->render(Args, CmdArgs);
+  }
 }
 
 static void RenderSCPOptions(const ToolChain &TC, const ArgList &Args,
@@ -5643,7 +5688,7 @@ void Clang::ConstructJob(Compilation &C, const JobAction &JA,
                    options::OPT_mno_speculative_load_hardening, false))
     CmdArgs.push_back(Args.MakeArgString("-mspeculative-load-hardening"));
 
-  RenderSSPOptions(TC, Args, CmdArgs, KernelOrKext);
+  RenderSSPOptions(D, TC, Args, CmdArgs, KernelOrKext);
   RenderSCPOptions(TC, Args, CmdArgs);
   RenderTrivialAutoVarInitOptions(D, TC, Args, CmdArgs);
 
diff --git a/clang/lib/Frontend/CompilerInvocation.cpp b/clang/lib/Frontend/CompilerInvocation.cpp
index fb5094ea940cf..c81dd80d78888 100644
--- a/clang/lib/Frontend/CompilerInvocation.cpp
+++ b/clang/lib/Frontend/CompilerInvocation.cpp
@@ -1258,6 +1258,21 @@ static bool ParseCodeGenArgs(CodeGenOptions &Opts, ArgList &Args, InputKind IK,
   }
   Opts.SSPBufferSize =
       getLastArgIntValue(Args, OPT_stack_protector_buffer_size, 8, Diags);
+
+  Opts.StackProtectorGuard =
+      std::string(Args.getLastArgValue(OPT_mstack_protector_guard_EQ));
+
+  if (Arg *A = Args.getLastArg(OPT_mstack_protector_guard_offset_EQ)) {
+    StringRef Val = A->getValue();
+    unsigned Offset = Opts.StackProtectorGuardOffset;
+    Val.getAsInteger(10, Offset);
+    Opts.StackProtectorGuardOffset = Offset;
+  }
+
+  Opts.StackProtectorGuardReg =
+      std::string(Args.getLastArgValue(OPT_mstack_protector_guard_reg_EQ,
+                                       "none"));
+
   Opts.StackRealignment = Args.hasArg(OPT_mstackrealign);
   if (Arg *A = Args.getLastArg(OPT_mstack_alignment)) {
     StringRef Val = A->getValue();
diff --git a/clang/lib/Headers/CMakeLists.txt b/clang/lib/Headers/CMakeLists.txt
index 7d0b2a0938bad..f9b8bfc21160d 100644
--- a/clang/lib/Headers/CMakeLists.txt
+++ b/clang/lib/Headers/CMakeLists.txt
@@ -113,6 +113,7 @@ set(files
   tgmath.h
   tmmintrin.h
   tsxldtrkintrin.h
+  uintrintrin.h
   unwind.h
   vadefs.h
   vaesintrin.h
diff --git a/clang/lib/Headers/__clang_hip_math.h b/clang/lib/Headers/__clang_hip_math.h
index f2365e8844fed..14d91c66b3522 100644
--- a/clang/lib/Headers/__clang_hip_math.h
+++ b/clang/lib/Headers/__clang_hip_math.h
@@ -547,102 +547,117 @@ float __expf(float __x) { return __ocml_native_exp_f32(__x); }
 #if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fadd_rd(float __x, float __y) { return __ocml_add_rtn_f32(__x, __y); }
-#endif
 __DEVICE__
 float __fadd_rn(float __x, float __y) { return __ocml_add_rte_f32(__x, __y); }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fadd_ru(float __x, float __y) { return __ocml_add_rtp_f32(__x, __y); }
-
 __DEVICE__
 float __fadd_rz(float __x, float __y) { return __ocml_add_rtz_f32(__x, __y); }
+#else
+__DEVICE__
+float __fadd_rn(float __x, float __y) { return __x + __y; }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fdiv_rd(float __x, float __y) { return __ocml_div_rtn_f32(__x, __y); }
-#endif
 __DEVICE__
 float __fdiv_rn(float __x, float __y) { return __ocml_div_rte_f32(__x, __y); }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fdiv_ru(float __x, float __y) { return __ocml_div_rtp_f32(__x, __y); }
-
 __DEVICE__
 float __fdiv_rz(float __x, float __y) { return __ocml_div_rtz_f32(__x, __y); }
+#else
+__DEVICE__
+float __fdiv_rn(float __x, float __y) { return __x / __y; }
 #endif
+
 __DEVICE__
 float __fdividef(float __x, float __y) { return __x / __y; }
+
 #if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fmaf_rd(float __x, float __y, float __z) {
   return __ocml_fma_rtn_f32(__x, __y, __z);
 }
-#endif
 __DEVICE__
 float __fmaf_rn(float __x, float __y, float __z) {
   return __ocml_fma_rte_f32(__x, __y, __z);
 }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fmaf_ru(float __x, float __y, float __z) {
   return __ocml_fma_rtp_f32(__x, __y, __z);
 }
-
 __DEVICE__
 float __fmaf_rz(float __x, float __y, float __z) {
   return __ocml_fma_rtz_f32(__x, __y, __z);
 }
+#else
+__DEVICE__
+float __fmaf_rn(float __x, float __y, float __z) {
+  return __ocml_fma_f32(__x, __y, __z);
+}
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fmul_rd(float __x, float __y) { return __ocml_mul_rtn_f32(__x, __y); }
-#endif
 __DEVICE__
 float __fmul_rn(float __x, float __y) { return __ocml_mul_rte_f32(__x, __y); }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fmul_ru(float __x, float __y) { return __ocml_mul_rtp_f32(__x, __y); }
-
 __DEVICE__
 float __fmul_rz(float __x, float __y) { return __ocml_mul_rtz_f32(__x, __y); }
-
+#else
 __DEVICE__
-float __frcp_rd(float __x) { return __llvm_amdgcn_rcp_f32(__x); }
+float __fmul_rn(float __x, float __y) { return __x * __y; }
 #endif
-__DEVICE__
-float __frcp_rn(float __x) { return __llvm_amdgcn_rcp_f32(__x); }
+
 #if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
-float __frcp_ru(float __x) { return __llvm_amdgcn_rcp_f32(__x); }
-
+float __frcp_rd(float __x) { return __ocml_div_rtn_f32(1.0f, __x); }
+__DEVICE__
+float __frcp_rn(float __x) { return __ocml_div_rte_f32(1.0f, __x); }
 __DEVICE__
-float __frcp_rz(float __x) { return __llvm_amdgcn_rcp_f32(__x); }
+float __frcp_ru(float __x) { return __ocml_div_rtp_f32(1.0f, __x); }
+__DEVICE__
+float __frcp_rz(float __x) { return __ocml_div_rtz_f32(1.0f, __x); }
+#else
+__DEVICE__
+float __frcp_rn(float __x) { return 1.0f / __x; }
 #endif
+
 __DEVICE__
 float __frsqrt_rn(float __x) { return __llvm_amdgcn_rsq_f32(__x); }
+
 #if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fsqrt_rd(float __x) { return __ocml_sqrt_rtn_f32(__x); }
-#endif
 __DEVICE__
 float __fsqrt_rn(float __x) { return __ocml_sqrt_rte_f32(__x); }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fsqrt_ru(float __x) { return __ocml_sqrt_rtp_f32(__x); }
-
 __DEVICE__
 float __fsqrt_rz(float __x) { return __ocml_sqrt_rtz_f32(__x); }
+#else
+__DEVICE__
+float __fsqrt_rn(float __x) { return __ocml_native_sqrt_f32(__x); }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fsub_rd(float __x, float __y) { return __ocml_sub_rtn_f32(__x, __y); }
-#endif
 __DEVICE__
 float __fsub_rn(float __x, float __y) { return __ocml_sub_rte_f32(__x, __y); }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 float __fsub_ru(float __x, float __y) { return __ocml_sub_rtp_f32(__x, __y); }
-
 __DEVICE__
 float __fsub_rz(float __x, float __y) { return __ocml_sub_rtz_f32(__x, __y); }
+#else
+__DEVICE__
+float __fsub_rn(float __x, float __y) { return __x - __y; }
 #endif
+
 __DEVICE__
 float __log10f(float __x) { return __ocml_native_log10_f32(__x); }
 
@@ -1071,125 +1086,139 @@ __DEVICE__
 double __dadd_rd(double __x, double __y) {
   return __ocml_add_rtn_f64(__x, __y);
 }
-#endif
 __DEVICE__
 double __dadd_rn(double __x, double __y) {
   return __ocml_add_rte_f64(__x, __y);
 }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dadd_ru(double __x, double __y) {
   return __ocml_add_rtp_f64(__x, __y);
 }
-
 __DEVICE__
 double __dadd_rz(double __x, double __y) {
   return __ocml_add_rtz_f64(__x, __y);
 }
+#else
+__DEVICE__
+double __dadd_rn(double __x, double __y) { return __x + __y; }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __ddiv_rd(double __x, double __y) {
   return __ocml_div_rtn_f64(__x, __y);
 }
-#endif
 __DEVICE__
 double __ddiv_rn(double __x, double __y) {
   return __ocml_div_rte_f64(__x, __y);
 }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __ddiv_ru(double __x, double __y) {
   return __ocml_div_rtp_f64(__x, __y);
 }
-
 __DEVICE__
 double __ddiv_rz(double __x, double __y) {
   return __ocml_div_rtz_f64(__x, __y);
 }
+#else
+__DEVICE__
+double __ddiv_rn(double __x, double __y) { return __x / __y; }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dmul_rd(double __x, double __y) {
   return __ocml_mul_rtn_f64(__x, __y);
 }
-#endif
 __DEVICE__
 double __dmul_rn(double __x, double __y) {
   return __ocml_mul_rte_f64(__x, __y);
 }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dmul_ru(double __x, double __y) {
   return __ocml_mul_rtp_f64(__x, __y);
 }
-
 __DEVICE__
 double __dmul_rz(double __x, double __y) {
   return __ocml_mul_rtz_f64(__x, __y);
 }
-
+#else
 __DEVICE__
-double __drcp_rd(double __x) { return __llvm_amdgcn_rcp_f64(__x); }
+double __dmul_rn(double __x, double __y) { return __x * __y; }
 #endif
-__DEVICE__
-double __drcp_rn(double __x) { return __llvm_amdgcn_rcp_f64(__x); }
+
 #if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
-double __drcp_ru(double __x) { return __llvm_amdgcn_rcp_f64(__x); }
-
+double __drcp_rd(double __x) { return __ocml_div_rtn_f64(1.0, __x); }
+__DEVICE__
+double __drcp_rn(double __x) { return __ocml_div_rte_f64(1.0, __x); }
+__DEVICE__
+double __drcp_ru(double __x) { return __ocml_div_rtp_f64(1.0, __x); }
+__DEVICE__
+double __drcp_rz(double __x) { return __ocml_div_rtz_f64(1.0, __x); }
+#else
 __DEVICE__
-double __drcp_rz(double __x) { return __llvm_amdgcn_rcp_f64(__x); }
+double __drcp_rn(double __x) { return 1.0 / __x; }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dsqrt_rd(double __x) { return __ocml_sqrt_rtn_f64(__x); }
-#endif
 __DEVICE__
 double __dsqrt_rn(double __x) { return __ocml_sqrt_rte_f64(__x); }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dsqrt_ru(double __x) { return __ocml_sqrt_rtp_f64(__x); }
-
 __DEVICE__
 double __dsqrt_rz(double __x) { return __ocml_sqrt_rtz_f64(__x); }
+#else
+__DEVICE__
+double __dsqrt_rn(double __x) { return __ocml_sqrt_f64(__x); }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dsub_rd(double __x, double __y) {
   return __ocml_sub_rtn_f64(__x, __y);
 }
-#endif
 __DEVICE__
 double __dsub_rn(double __x, double __y) {
   return __ocml_sub_rte_f64(__x, __y);
 }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __dsub_ru(double __x, double __y) {
   return __ocml_sub_rtp_f64(__x, __y);
 }
-
 __DEVICE__
 double __dsub_rz(double __x, double __y) {
   return __ocml_sub_rtz_f64(__x, __y);
 }
+#else
+__DEVICE__
+double __dsub_rn(double __x, double __y) { return __x - __y; }
+#endif
 
+#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __fma_rd(double __x, double __y, double __z) {
   return __ocml_fma_rtn_f64(__x, __y, __z);
 }
-#endif
 __DEVICE__
 double __fma_rn(double __x, double __y, double __z) {
   return __ocml_fma_rte_f64(__x, __y, __z);
 }
-#if defined OCML_BASIC_ROUNDED_OPERATIONS
 __DEVICE__
 double __fma_ru(double __x, double __y, double __z) {
   return __ocml_fma_rtp_f64(__x, __y, __z);
 }
-
 __DEVICE__
 double __fma_rz(double __x, double __y, double __z) {
   return __ocml_fma_rtz_f64(__x, __y, __z);
 }
+#else
+__DEVICE__
+double __fma_rn(double __x, double __y, double __z) {
+  return __ocml_fma_f64(__x, __y, __z);
+}
 #endif
 // END INTRINSICS
 // END DOUBLE
diff --git a/clang/lib/Headers/cpuid.h b/clang/lib/Headers/cpuid.h
index 639aa015fbf2f..6ed7bd04ca0bf 100644
--- a/clang/lib/Headers/cpuid.h
+++ b/clang/lib/Headers/cpuid.h
@@ -186,6 +186,7 @@
 /* Features in %edx for leaf 7 sub-leaf 0 */
 #define bit_AVX5124VNNIW  0x00000004
 #define bit_AVX5124FMAPS  0x00000008
+#define bit_UINTR         0x00000020
 #define bit_SERIALIZE     0x00004000
 #define bit_TSXLDTRK      0x00010000
 #define bit_PCONFIG       0x00040000
diff --git a/clang/lib/Headers/uintrintrin.h b/clang/lib/Headers/uintrintrin.h
new file mode 100644
index 0000000000000..78aa8779c3255
--- /dev/null
+++ b/clang/lib/Headers/uintrintrin.h
@@ -0,0 +1,150 @@
+/*===------------------ uintrintrin.h - UINTR intrinsics -------------------===
+ *
+ * Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+ * See https://llvm.org/LICENSE.txt for license information.
+ * SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+ *
+ *===-----------------------------------------------------------------------===
+ */
+
+#ifndef __X86GPRINTRIN_H
+#error "Never use <uintrintrin.h> directly; include <x86gprintrin.h> instead."
+#endif
+
+#ifndef __UINTRINTRIN_H
+#define __UINTRINTRIN_H
+
+/* Define the default attributes for the functions in this file */
+#define __DEFAULT_FN_ATTRS \
+  __attribute__((__always_inline__, __nodebug__, __target__("uintr")))
+
+#ifdef __x86_64__
+
+/// Clears the user interrupt flag (UIF). Its effect takes place immediately: a
+///    user interrupt cannot be delivered on the instruction boundary following
+///    CLUI. Can be executed only if CR4.UINT = 1, the logical processor is in
+///    64-bit mode, and software is not executing inside an enclave; otherwise,
+///    each causes an invalid-opcode exception. Causes a transactional abort if
+///    executed inside a transactional region; the abort loads EAX as it would
+///    had it been due to an execution of CLI.
+///
+/// \headerfile <x86gprintrin.h>
+///
+/// This intrinsic corresponds to the <c> CLUI </c> instruction.
+///
+/// \operation
+///   UIF := 0
+/// \endoperation
+static __inline__ void __DEFAULT_FN_ATTRS
+_clui (void)
+{
+  __builtin_ia32_clui();
+}
+
+/// Sets the user interrupt flag (UIF). Its effect takes place immediately; a
+///    user interrupt may be delivered on the instruction boundary following
+///    STUI. Can be executed only if CR4.UINT = 1, the logical processor is in
+///    64-bit mode, and software is not executing inside an enclave; otherwise,
+///    each causes an invalid-opcode exception. Causes a transactional abort if
+///    executed inside a transactional region; the abort loads EAX as it would
+///    had it been due to an execution of STI.
+///
+/// \headerfile <x86gprintrin.h>
+///
+/// This intrinsic corresponds to the <c> STUI </c> instruction.
+///
+/// \operation
+///   UIF := 1
+/// \endoperation
+static __inline__ void __DEFAULT_FN_ATTRS
+_stui (void)
+{
+  __builtin_ia32_stui();
+}
+
+/// Get the current value of the user interrupt flag (UIF). Can be executed
+///    regardless of CPL and inside a transactional region. Can be executed only
+///    if CR4.UINT = 1, the logical processor is in 64-bit mode, and software is
+///    not executing inside an enclave; otherwise, it causes an invalid-opcode
+///    exception.
+///
+/// \headerfile <x86gprintrin.h>
+///
+/// This intrinsic corresponds to the <c> TESTUI </c> instruction.
+///
+/// \returns The current value of the user interrupt flag (UIF).
+///
+/// \operation
+///   CF := UIF
+///   ZF := 0
+///   AF := 0
+///   OF := 0
+///   PF := 0
+///   SF := 0
+///   dst := CF
+/// \endoperation
+static __inline__ unsigned char __DEFAULT_FN_ATTRS
+_testui (void)
+{
+  return __builtin_ia32_testui();
+}
+
+/// Send interprocessor user interrupt. Can be executed only if
+///    CR4.UINT = IA32_UINT_TT[0] = 1, the logical processor is in 64-bit mode,
+///    and software is not executing inside an enclave; otherwise, it causes an
+///    invalid-opcode exception. May be executed at any privilege level, all of
+///    its memory accesses are performed with supervisor privilege.
+///
+/// \headerfile <x86gprintrin.h>
+///
+/// This intrinsic corresponds to the <c> SENDUIPI </c> instruction
+///
+/// \param __a
+///    Index of user-interrupt target table entry in user-interrupt target
+///    table.
+///
+/// \operation
+///   IF __a > UITTSZ
+///     GP (0)
+///   FI
+///   tempUITTE := MEM[UITTADDR + (a<<4)]
+///   // tempUITTE must be valid, and can't have any reserved bit set
+///   IF (tempUITTE.V == 0 OR tempUITTE[7:1] != 0)
+///     GP (0)
+///   FI
+///   tempUPID := MEM[tempUITTE.UPIDADDR] // under lock
+///   // tempUPID can't have any reserved bit set
+///   IF (tempUPID[15:2] != 0 OR tempUPID[31:24] != 0)
+///     GP (0) // release lock
+///   FI
+///   tempUPID.PIR[tempUITTE.UV] := 1;
+///   IF (tempUPID.SN == 0 AND tempUPID.ON == 0)
+///     tempUPID.ON := 1
+///     sendNotify := 1
+///   ELSE
+///     sendNotify := 0
+///   FI
+///   MEM[tempUITTE.UPIDADDR] := tempUPID // release lock
+///   IF sendNotify == 1
+///     IF IA32_APIC_BASE[10] == 1 // local APIC is in x2APIC mode
+///       // send ordinary IPI with vector tempUPID.NV to 32-bit physical APIC
+///       // ID tempUPID.NDST
+///       SendOrdinaryIPI(tempUPID.NV, tempUPID.NDST)
+///     ELSE
+///       // send ordinary IPI with vector tempUPID.NV to 8-bit physical APIC
+///       // ID tempUPID.NDST[15:8]
+///       SendOrdinaryIPI(tempUPID.NV, tempUPID.NDST[15:8])
+///     FI
+///   FI
+/// \endoperation
+static __inline__ void __DEFAULT_FN_ATTRS
+_senduipi (unsigned long long __a)
+{
+  __builtin_ia32_senduipi(__a);
+}
+
+#endif /* __x86_64__ */
+
+#undef __DEFAULT_FN_ATTRS
+
+#endif /* __UINTRINTRIN_H */
diff --git a/clang/lib/Headers/x86gprintrin.h b/clang/lib/Headers/x86gprintrin.h
index 6aa0a66e18fc1..1fc6cab4b28fc 100644
--- a/clang/lib/Headers/x86gprintrin.h
+++ b/clang/lib/Headers/x86gprintrin.h
@@ -15,4 +15,9 @@
 #include <hresetintrin.h>
 #endif
 
+#if !(defined(_MSC_VER) || defined(__SCE__)) || __has_feature(modules) ||      \
+    defined(__UINTR__)
+#include <uintrintrin.h>
+#endif
+
 #endif /* __X86GPRINTRIN_H */
diff --git a/clang/lib/Sema/SemaExpr.cpp b/clang/lib/Sema/SemaExpr.cpp
index 4cddb5e434508..bf7abcd313979 100644
--- a/clang/lib/Sema/SemaExpr.cpp
+++ b/clang/lib/Sema/SemaExpr.cpp
@@ -3246,6 +3246,17 @@ ExprResult Sema::BuildDeclarationNameExpr(
         break;
       }
 
+      // [expr.prim.id.unqual]p2:
+      //   If the entity is a template parameter object for a template
+      //   parameter of type T, the type of the expression is const T.
+      //   [...] The expression is an lvalue if the entity is a [...] template
+      //   parameter object.
+      if (type->isRecordType()) {
+        type = type.getUnqualifiedType().withConst();
+        valueKind = VK_LValue;
+        break;
+      }
+
       // For non-references, we need to strip qualifiers just in case
       // the template parameter was declared as 'const int' or whatever.
       valueKind = VK_RValue;
@@ -3345,8 +3356,9 @@ ExprResult Sema::BuildDeclarationNameExpr(
 
     case Decl::MSProperty:
     case Decl::MSGuid:
-      // FIXME: Should MSGuidDecl be subject to capture in OpenMP,
-      // or duplicated between host and device?
+    case Decl::TemplateParamObject:
+      // FIXME: Should MSGuidDecl and template parameter objects be subject to
+      // capture in OpenMP, or duplicated between host and device?
       valueKind = VK_LValue;
       break;
 
diff --git a/clang/lib/Sema/SemaInit.cpp b/clang/lib/Sema/SemaInit.cpp
index 8b11f73cae859..ef3061ee514fd 100644
--- a/clang/lib/Sema/SemaInit.cpp
+++ b/clang/lib/Sema/SemaInit.cpp
@@ -1123,6 +1123,7 @@ static void warnBracedScalarInit(Sema &S, const InitializedEntity &Entity,
   case InitializedEntity::EK_ArrayElement:
   case InitializedEntity::EK_Parameter:
   case InitializedEntity::EK_Parameter_CF_Audited:
+  case InitializedEntity::EK_TemplateParameter:
   case InitializedEntity::EK_Result:
     // Extra braces here are suspicious.
     DiagID = diag::warn_braces_around_init;
@@ -3283,6 +3284,7 @@ DeclarationName InitializedEntity::getName() const {
   case EK_Variable:
   case EK_Member:
   case EK_Binding:
+  case EK_TemplateParameter:
     return Variable.VariableOrMember->getDeclName();
 
   case EK_LambdaCapture:
@@ -3313,6 +3315,7 @@ ValueDecl *InitializedEntity::getDecl() const {
   case EK_Variable:
   case EK_Member:
   case EK_Binding:
+  case EK_TemplateParameter:
     return Variable.VariableOrMember;
 
   case EK_Parameter:
@@ -3350,6 +3353,7 @@ bool InitializedEntity::allowsNRVO() const {
   case EK_Variable:
   case EK_Parameter:
   case EK_Parameter_CF_Audited:
+  case EK_TemplateParameter:
   case EK_Member:
   case EK_Binding:
   case EK_New:
@@ -3381,6 +3385,7 @@ unsigned InitializedEntity::dumpImpl(raw_ostream &OS) const {
   case EK_Parameter: OS << "Parameter"; break;
   case EK_Parameter_CF_Audited: OS << "CF audited function Parameter";
     break;
+  case EK_TemplateParameter: OS << "TemplateParameter"; break;
   case EK_Result: OS << "Result"; break;
   case EK_StmtExprResult: OS << "StmtExprResult"; break;
   case EK_Exception: OS << "Exception"; break;
@@ -6001,6 +6006,11 @@ getAssignmentAction(const InitializedEntity &Entity, bool Diagnose = false) {
     // FIXME: Can we tell apart casting vs. converting?
     return Sema::AA_Casting;
 
+  case InitializedEntity::EK_TemplateParameter:
+    // This is really initialization, but refer to it as conversion for
+    // consistency with CheckConvertedConstantExpression.
+    return Sema::AA_Converting;
+
   case InitializedEntity::EK_Member:
   case InitializedEntity::EK_Binding:
   case InitializedEntity::EK_ArrayElement:
@@ -6035,6 +6045,7 @@ static bool shouldBindAsTemporary(const InitializedEntity &Entity) {
   case InitializedEntity::EK_LambdaToBlockConversionBlockElement:
   case InitializedEntity::EK_LambdaCapture:
   case InitializedEntity::EK_CompoundLiteralInit:
+  case InitializedEntity::EK_TemplateParameter:
     return false;
 
   case InitializedEntity::EK_Parameter:
@@ -6069,6 +6080,7 @@ static bool shouldDestroyEntity(const InitializedEntity &Entity) {
     case InitializedEntity::EK_Variable:
     case InitializedEntity::EK_Parameter:
     case InitializedEntity::EK_Parameter_CF_Audited:
+    case InitializedEntity::EK_TemplateParameter:
     case InitializedEntity::EK_Temporary:
     case InitializedEntity::EK_ArrayElement:
     case InitializedEntity::EK_Exception:
@@ -6102,6 +6114,7 @@ static SourceLocation getInitializationLoc(const InitializedEntity &Entity,
   case InitializedEntity::EK_Member:
   case InitializedEntity::EK_Parameter:
   case InitializedEntity::EK_Parameter_CF_Audited:
+  case InitializedEntity::EK_TemplateParameter:
   case InitializedEntity::EK_Temporary:
   case InitializedEntity::EK_New:
   case InitializedEntity::EK_Base:
@@ -6345,7 +6358,7 @@ static void CheckCXX98CompatAccessibleCopy(Sema &S,
 
 void InitializationSequence::PrintInitLocationNote(Sema &S,
                                               const InitializedEntity &Entity) {
-  if (Entity.isParameterKind() && Entity.getDecl()) {
+  if (Entity.isParamOrTemplateParamKind() && Entity.getDecl()) {
     if (Entity.getDecl()->getLocation().isInvalid())
       return;
 
@@ -6611,6 +6624,10 @@ static LifetimeResult getEntityLifetime(
     //      the call.
     return {nullptr, LK_FullExpression};
 
+  case InitializedEntity::EK_TemplateParameter:
+    // FIXME: This will always be ill-formed; should we eagerly diagnose it here?
+    return {nullptr, LK_FullExpression};
+
   case InitializedEntity::EK_Result:
     //   -- The lifetime of a temporary bound to the returned value in a
     //      function return statement is not extended; the temporary is
@@ -7882,7 +7899,7 @@ ExprResult InitializationSequence::Perform(Sema &S,
 
   if (S.getLangOpts().CPlusPlus11 && Entity.getType()->isReferenceType() &&
       Args.size() == 1 && isa<InitListExpr>(Args[0]) &&
-      !Entity.isParameterKind()) {
+      !Entity.isParamOrTemplateParamKind()) {
     // Produce a C++98 compatibility warning if we are initializing a reference
     // from an initializer list. For parameters, we produce a better warning
     // elsewhere.
diff --git a/clang/lib/Sema/SemaOverload.cpp b/clang/lib/Sema/SemaOverload.cpp
index d2eabd0c05121..7da2a8c2d84f9 100644
--- a/clang/lib/Sema/SemaOverload.cpp
+++ b/clang/lib/Sema/SemaOverload.cpp
@@ -5571,7 +5571,8 @@ static bool CheckConvertedConstantConversions(Sema &S,
 static ExprResult CheckConvertedConstantExpression(Sema &S, Expr *From,
                                                    QualType T, APValue &Value,
                                                    Sema::CCEKind CCE,
-                                                   bool RequireInt) {
+                                                   bool RequireInt,
+                                                   NamedDecl *Dest) {
   assert(S.getLangOpts().CPlusPlus11 &&
          "converted constant expression outside C++11");
 
@@ -5601,9 +5602,10 @@ static ExprResult CheckConvertedConstantExpression(Sema &S, Expr *From,
     SCS = &ICS.Standard;
     break;
   case ImplicitConversionSequence::UserDefinedConversion:
-    // We are converting to a non-class type, so the Before sequence
-    // must be trivial.
-    SCS = &ICS.UserDefined.After;
+    if (T->isRecordType())
+      SCS = &ICS.UserDefined.Before;
+    else
+      SCS = &ICS.UserDefined.After;
     break;
   case ImplicitConversionSequence::AmbiguousConversion:
   case ImplicitConversionSequence::BadConversion:
@@ -5630,8 +5632,20 @@ static ExprResult CheckConvertedConstantExpression(Sema &S, Expr *From,
            << From->getType() << From->getSourceRange() << T;
   }
 
-  ExprResult Result =
-      S.PerformImplicitConversion(From, T, ICS, Sema::AA_Converting);
+  // Usually we can simply apply the ImplicitConversionSequence we formed
+  // earlier, but that's not guaranteed to work when initializing an object of
+  // class type.
+  ExprResult Result;
+  if (T->isRecordType()) {
+    assert(CCE == Sema::CCEK_TemplateArg &&
+           "unexpected class type converted constant expr");
+    Result = S.PerformCopyInitialization(
+        InitializedEntity::InitializeTemplateParameter(
+            T, cast<NonTypeTemplateParmDecl>(Dest)),
+        SourceLocation(), From);
+  } else {
+    Result = S.PerformImplicitConversion(From, T, ICS, Sema::AA_Converting);
+  }
   if (Result.isInvalid())
     return Result;
 
@@ -5724,8 +5738,10 @@ static ExprResult CheckConvertedConstantExpression(Sema &S, Expr *From,
 }
 
 ExprResult Sema::CheckConvertedConstantExpression(Expr *From, QualType T,
-                                                  APValue &Value, CCEKind CCE) {
-  return ::CheckConvertedConstantExpression(*this, From, T, Value, CCE, false);
+                                                  APValue &Value, CCEKind CCE,
+                                                  NamedDecl *Dest) {
+  return ::CheckConvertedConstantExpression(*this, From, T, Value, CCE, false,
+                                            Dest);
 }
 
 ExprResult Sema::CheckConvertedConstantExpression(Expr *From, QualType T,
@@ -5734,7 +5750,8 @@ ExprResult Sema::CheckConvertedConstantExpression(Expr *From, QualType T,
   assert(T->isIntegralOrEnumerationType() && "unexpected converted const type");
 
   APValue V;
-  auto R = ::CheckConvertedConstantExpression(*this, From, T, V, CCE, true);
+  auto R = ::CheckConvertedConstantExpression(*this, From, T, V, CCE, true,
+                                              /*Dest=*/nullptr);
   if (!R.isInvalid() && !R.get()->isValueDependent())
     Value = V.getInt();
   return R;
diff --git a/clang/lib/Sema/SemaTemplate.cpp b/clang/lib/Sema/SemaTemplate.cpp
index 660964c8f92b2..d23ad9f7d91d5 100644
--- a/clang/lib/Sema/SemaTemplate.cpp
+++ b/clang/lib/Sema/SemaTemplate.cpp
@@ -23,6 +23,7 @@
 #include "clang/Basic/Stack.h"
 #include "clang/Basic/TargetInfo.h"
 #include "clang/Sema/DeclSpec.h"
+#include "clang/Sema/Initialization.h"
 #include "clang/Sema/Lookup.h"
 #include "clang/Sema/Overload.h"
 #include "clang/Sema/ParsedTemplate.h"
@@ -1441,6 +1442,7 @@ QualType Sema::CheckNonTypeTemplateParameterType(QualType T,
     return QualType();
   }
 
+  Diag(Loc, diag::warn_cxx17_compat_template_nontype_parm_type) << T;
   return T.getUnqualifiedType();
 }
 
@@ -6806,14 +6808,15 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
   SourceLocation StartLoc = Arg->getBeginLoc();
 
   // If the parameter type somehow involves auto, deduce the type now.
-  if (getLangOpts().CPlusPlus17 && ParamType->isUndeducedType()) {
+  DeducedType *DeducedT = ParamType->getContainedDeducedType();
+  if (getLangOpts().CPlusPlus17 && DeducedT && !DeducedT->isDeduced()) {
     // During template argument deduction, we allow 'decltype(auto)' to
     // match an arbitrary dependent argument.
     // FIXME: The language rules don't say what happens in this case.
     // FIXME: We get an opaque dependent type out of decltype(auto) if the
     // expression is merely instantiation-dependent; is this enough?
     if (CTAK == CTAK_Deduced && Arg->isTypeDependent()) {
-      auto *AT = dyn_cast<AutoType>(ParamType);
+      auto *AT = dyn_cast<AutoType>(DeducedT);
       if (AT && AT->isDecltypeAuto()) {
         Converted = TemplateArgument(Arg);
         return Arg;
@@ -6827,14 +6830,26 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
     Expr *DeductionArg = Arg;
     if (auto *PE = dyn_cast<PackExpansionExpr>(DeductionArg))
       DeductionArg = PE->getPattern();
-    if (DeduceAutoType(
-            Context.getTrivialTypeSourceInfo(ParamType, Param->getLocation()),
-            DeductionArg, ParamType, Depth,
-            // We do not check constraints right now because the
-            // immediately-declared constraint of the auto type is also an
-            // associated constraint, and will be checked along with the other
-            // associated constraints after checking the template argument list.
-            /*IgnoreConstraints=*/true) == DAR_Failed) {
+    TypeSourceInfo *TSI =
+        Context.getTrivialTypeSourceInfo(ParamType, Param->getLocation());
+    if (auto *DTST = dyn_cast<DeducedTemplateSpecializationType>(DeducedT)) {
+      InitializedEntity Entity =
+          InitializedEntity::InitializeTemplateParameter(ParamType, Param);
+      InitializationKind Kind = InitializationKind::CreateForInit(
+          DeductionArg->getBeginLoc(), /*DirectInit*/false, DeductionArg);
+      Expr *Inits[1] = {DeductionArg};
+      ParamType =
+          DeduceTemplateSpecializationFromInitializer(TSI, Entity, Kind, Inits);
+      if (ParamType.isNull())
+        return ExprError();
+    } else if (DeduceAutoType(
+                   TSI, DeductionArg, ParamType, Depth,
+                   // We do not check constraints right now because the
+                   // immediately-declared constraint of the auto type is also
+                   // an associated constraint, and will be checked along with
+                   // the other associated constraints after checking the
+                   // template argument list.
+                   /*IgnoreConstraints=*/true) == DAR_Failed) {
       Diag(Arg->getExprLoc(),
            diag::err_non_type_template_parm_type_deduction_failure)
         << Param->getDeclName() << Param->getType() << Arg->getType()
@@ -6857,18 +6872,21 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
   assert(!ParamType.hasQualifiers() &&
          "non-type template parameter type cannot be qualified");
 
+  // FIXME: When Param is a reference, should we check that Arg is an lvalue?
   if (CTAK == CTAK_Deduced &&
-      !Context.hasSameType(ParamType.getNonLValueExprType(Context),
-                           Arg->getType())) {
+      (ParamType->isReferenceType()
+           ? !Context.hasSameType(ParamType.getNonReferenceType(),
+                                  Arg->getType())
+           : !Context.hasSameUnqualifiedType(ParamType, Arg->getType()))) {
     // FIXME: If either type is dependent, we skip the check. This isn't
     // correct, since during deduction we're supposed to have replaced each
     // template parameter with some unique (non-dependent) placeholder.
     // FIXME: If the argument type contains 'auto', we carry on and fail the
     // type check in order to force specific types to be more specialized than
     // 'auto'. It's not clear how partial ordering with 'auto' is supposed to
-    // work.
+    // work. Similarly for CTAD, when comparing 'A<x>' against 'A'.
     if ((ParamType->isDependentType() || Arg->isTypeDependent()) &&
-        !Arg->getType()->getContainedAutoType()) {
+        !Arg->getType()->getContainedDeducedType()) {
       Converted = TemplateArgument(Arg);
       return Arg;
     }
@@ -6915,12 +6933,36 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
       *this, Sema::ExpressionEvaluationContext::ConstantEvaluated);
 
   if (getLangOpts().CPlusPlus17) {
+    QualType CanonParamType = Context.getCanonicalType(ParamType);
+
+    // Avoid making a copy when initializing a template parameter of class type
+    // from a template parameter object of the same type. This is going beyond
+    // the standard, but is required for soundness: in
+    //   template<A a> struct X { X *p; X<a> *q; };
+    // ... we need p and q to have the same type.
+    //
+    // Similarly, don't inject a call to a copy constructor when initializing
+    // from a template parameter of the same type.
+    Expr *InnerArg = Arg->IgnoreParenImpCasts();
+    if (ParamType->isRecordType() && isa<DeclRefExpr>(InnerArg) &&
+        Context.hasSameUnqualifiedType(ParamType, InnerArg->getType())) {
+      NamedDecl *ND = cast<DeclRefExpr>(InnerArg)->getDecl();
+      if (auto *TPO = dyn_cast<TemplateParamObjectDecl>(ND)) {
+        Converted = TemplateArgument(TPO, CanonParamType);
+        return Arg;
+      }
+      if (isa<NonTypeTemplateParmDecl>(ND)) {
+        Converted = TemplateArgument(Arg);
+        return Arg;
+      }
+    }
+
     // C++17 [temp.arg.nontype]p1:
     //   A template-argument for a non-type template parameter shall be
     //   a converted constant expression of the type of the template-parameter.
     APValue Value;
     ExprResult ArgResult = CheckConvertedConstantExpression(
-        Arg, ParamType, Value, CCEK_TemplateArg);
+        Arg, ParamType, Value, CCEK_TemplateArg, Param);
     if (ArgResult.isInvalid())
       return ExprError();
 
@@ -6931,8 +6973,6 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
       return ArgResult;
     }
 
-    QualType CanonParamType = Context.getCanonicalType(ParamType);
-
     // Convert the APValue to a TemplateArgument.
     switch (Value.getKind()) {
     case APValue::None:
@@ -7002,6 +7042,13 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
                      : TemplateArgument(CanonParamType, /*isNullPtr*/true);
       break;
     }
+    case APValue::Struct:
+    case APValue::Union:
+      // Get or create the corresponding template parameter object.
+      Converted = TemplateArgument(
+          Context.getTemplateParamObjectDecl(CanonParamType, Value),
+          CanonParamType);
+      break;
     case APValue::AddrLabelDiff:
       return Diag(StartLoc, diag::err_non_type_template_arg_addr_label_diff);
     case APValue::FixedPoint:
@@ -7010,8 +7057,6 @@ ExprResult Sema::CheckTemplateArgument(NonTypeTemplateParmDecl *Param,
     case APValue::ComplexFloat:
     case APValue::Vector:
     case APValue::Array:
-    case APValue::Struct:
-    case APValue::Union:
       return Diag(StartLoc, diag::err_non_type_template_arg_unsupported)
              << ParamType;
     }
@@ -7505,6 +7550,11 @@ Sema::BuildExpressionFromDeclTemplateArgument(const TemplateArgument &Arg,
     RefExpr = CreateBuiltinUnaryOp(Loc, UO_AddrOf, RefExpr.get());
     if (RefExpr.isInvalid())
       return ExprError();
+  } else if (ParamType->isRecordType()) {
+    assert(isa<TemplateParamObjectDecl>(VD) &&
+           "arg for class template param not a template parameter object");
+    // No conversions apply in this case.
+    return RefExpr;
   } else {
     assert(ParamType->isReferenceType() &&
            "unexpected type for decl template argument");
diff --git a/clang/lib/Sema/SemaTemplateDeduction.cpp b/clang/lib/Sema/SemaTemplateDeduction.cpp
index b266f360a1387..886377108e3f7 100644
--- a/clang/lib/Sema/SemaTemplateDeduction.cpp
+++ b/clang/lib/Sema/SemaTemplateDeduction.cpp
@@ -171,30 +171,41 @@ static void MarkUsedTemplateParameters(ASTContext &Ctx, QualType T,
 /// If the given expression is of a form that permits the deduction
 /// of a non-type template parameter, return the declaration of that
 /// non-type template parameter.
-static NonTypeTemplateParmDecl *
-getDeducedParameterFromExpr(TemplateDeductionInfo &Info, Expr *E) {
+static const NonTypeTemplateParmDecl *
+getDeducedParameterFromExpr(const Expr *E, unsigned Depth) {
   // If we are within an alias template, the expression may have undergone
   // any number of parameter substitutions already.
   while (true) {
-    if (ImplicitCastExpr *IC = dyn_cast<ImplicitCastExpr>(E))
+    if (const auto *IC = dyn_cast<ImplicitCastExpr>(E))
       E = IC->getSubExpr();
-    else if (ConstantExpr *CE = dyn_cast<ConstantExpr>(E))
+    else if (const auto *CE = dyn_cast<ConstantExpr>(E))
       E = CE->getSubExpr();
-    else if (SubstNonTypeTemplateParmExpr *Subst =
-               dyn_cast<SubstNonTypeTemplateParmExpr>(E))
+    else if (const auto *Subst = dyn_cast<SubstNonTypeTemplateParmExpr>(E))
       E = Subst->getReplacement();
-    else
+    else if (const auto *CCE = dyn_cast<CXXConstructExpr>(E)) {
+      // Look through implicit copy construction from an lvalue of the same type.
+      if (CCE->getParenOrBraceRange().isValid())
+        break;
+      // Note, there could be default arguments.
+      assert(CCE->getNumArgs() >= 1 && "implicit construct expr should have 1 arg");
+      E = CCE->getArg(0);
+    } else
       break;
   }
 
-  if (DeclRefExpr *DRE = dyn_cast<DeclRefExpr>(E))
-    if (auto *NTTP = dyn_cast<NonTypeTemplateParmDecl>(DRE->getDecl()))
-      if (NTTP->getDepth() == Info.getDeducedDepth())
+  if (const auto *DRE = dyn_cast<DeclRefExpr>(E))
+    if (const auto *NTTP = dyn_cast<NonTypeTemplateParmDecl>(DRE->getDecl()))
+      if (NTTP->getDepth() == Depth)
         return NTTP;
 
   return nullptr;
 }
 
+static const NonTypeTemplateParmDecl *
+getDeducedParameterFromExpr(TemplateDeductionInfo &Info, Expr *E) {
+  return getDeducedParameterFromExpr(E, Info.getDeducedDepth());
+}
+
 /// Determine whether two declaration pointers refer to the same
 /// declaration.
 static bool isSameDeclaration(Decl *X, Decl *Y) {
@@ -374,7 +385,7 @@ checkDeducedTemplateArguments(ASTContext &Context,
 /// deduction is funneled through here.
 static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
     Sema &S, TemplateParameterList *TemplateParams,
-    NonTypeTemplateParmDecl *NTTP, const DeducedTemplateArgument &NewDeduced,
+    const NonTypeTemplateParmDecl *NTTP, const DeducedTemplateArgument &NewDeduced,
     QualType ValueType, TemplateDeductionInfo &Info,
     SmallVectorImpl<DeducedTemplateArgument> &Deduced) {
   assert(NTTP->getDepth() == Info.getDeducedDepth() &&
@@ -383,7 +394,7 @@ static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
   DeducedTemplateArgument Result = checkDeducedTemplateArguments(
       S.Context, Deduced[NTTP->getIndex()], NewDeduced);
   if (Result.isNull()) {
-    Info.Param = NTTP;
+    Info.Param = const_cast<NonTypeTemplateParmDecl*>(NTTP);
     Info.FirstArg = Deduced[NTTP->getIndex()];
     Info.SecondArg = NewDeduced;
     return Sema::TDK_Inconsistent;
@@ -410,10 +421,16 @@ static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
   // type from an argument (of non-reference type) should be performed.
   // For now, we just remove reference types from both sides and let
   // the final check for matching types sort out the mess.
+  ValueType = ValueType.getNonReferenceType();
+  if (ParamType->isReferenceType())
+    ParamType = ParamType.getNonReferenceType();
+  else
+    // Top-level cv-qualifiers are irrelevant for a non-reference type.
+    ValueType = ValueType.getUnqualifiedType();
+
   return DeduceTemplateArgumentsByTypeMatch(
-      S, TemplateParams, ParamType.getNonReferenceType(),
-      ValueType.getNonReferenceType(), Info, Deduced, TDF_SkipNonDependent,
-      /*PartialOrdering=*/false,
+      S, TemplateParams, ParamType, ValueType, Info, Deduced,
+      TDF_SkipNonDependent, /*PartialOrdering=*/false,
       /*ArrayBound=*/NewDeduced.wasDeducedFromArrayBound());
 }
 
@@ -421,7 +438,7 @@ static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
 /// from the given integral constant.
 static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
     Sema &S, TemplateParameterList *TemplateParams,
-    NonTypeTemplateParmDecl *NTTP, const llvm::APSInt &Value,
+    const NonTypeTemplateParmDecl *NTTP, const llvm::APSInt &Value,
     QualType ValueType, bool DeducedFromArrayBound, TemplateDeductionInfo &Info,
     SmallVectorImpl<DeducedTemplateArgument> &Deduced) {
   return DeduceNonTypeTemplateArgument(
@@ -435,7 +452,7 @@ static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
 /// from the given null pointer template argument type.
 static Sema::TemplateDeductionResult DeduceNullPtrTemplateArgument(
     Sema &S, TemplateParameterList *TemplateParams,
-    NonTypeTemplateParmDecl *NTTP, QualType NullPtrType,
+    const NonTypeTemplateParmDecl *NTTP, QualType NullPtrType,
     TemplateDeductionInfo &Info,
     SmallVectorImpl<DeducedTemplateArgument> &Deduced) {
   Expr *Value =
@@ -454,7 +471,7 @@ static Sema::TemplateDeductionResult DeduceNullPtrTemplateArgument(
 /// \returns true if deduction succeeded, false otherwise.
 static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
     Sema &S, TemplateParameterList *TemplateParams,
-    NonTypeTemplateParmDecl *NTTP, Expr *Value, TemplateDeductionInfo &Info,
+    const NonTypeTemplateParmDecl *NTTP, Expr *Value, TemplateDeductionInfo &Info,
     SmallVectorImpl<DeducedTemplateArgument> &Deduced) {
   return DeduceNonTypeTemplateArgument(S, TemplateParams, NTTP,
                                        DeducedTemplateArgument(Value),
@@ -467,7 +484,7 @@ static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
 /// \returns true if deduction succeeded, false otherwise.
 static Sema::TemplateDeductionResult DeduceNonTypeTemplateArgument(
     Sema &S, TemplateParameterList *TemplateParams,
-    NonTypeTemplateParmDecl *NTTP, ValueDecl *D, QualType T,
+    const NonTypeTemplateParmDecl *NTTP, ValueDecl *D, QualType T,
     TemplateDeductionInfo &Info,
     SmallVectorImpl<DeducedTemplateArgument> &Deduced) {
   D = D ? cast<ValueDecl>(D->getCanonicalDecl()) : nullptr;
@@ -1598,14 +1615,18 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
       return Sema::TDK_Success;
     }
   } else if (!Param->isDependentType()) {
-    CanQualType ParamUnqualType = CanParam.getUnqualifiedType(),
-                ArgUnqualType = CanArg.getUnqualifiedType();
-    bool Success =
-        (TDF & TDF_AllowCompatibleFunctionType)
-            ? S.isSameOrCompatibleFunctionType(ParamUnqualType, ArgUnqualType)
-            : ParamUnqualType == ArgUnqualType;
-    if (Success)
+    if (!(TDF & TDF_SkipNonDependent)) {
+      CanQualType ParamUnqualType = CanParam.getUnqualifiedType(),
+                  ArgUnqualType = CanArg.getUnqualifiedType();
+      bool Success =
+          (TDF & TDF_AllowCompatibleFunctionType)
+              ? S.isSameOrCompatibleFunctionType(ParamUnqualType, ArgUnqualType)
+              : ParamUnqualType == ArgUnqualType;
+      if (Success)
+        return Sema::TDK_Success;
+    } else {
       return Sema::TDK_Success;
+    }
   }
 
   switch (Param->getTypeClass()) {
@@ -1757,7 +1778,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
         return Result;
 
       // Determine the array bound is something we can deduce.
-      NonTypeTemplateParmDecl *NTTP
+      const NonTypeTemplateParmDecl *NTTP
         = getDeducedParameterFromExpr(Info, DependentArrayParm->getSizeExpr());
       if (!NTTP)
         return Sema::TDK_Success;
@@ -1826,7 +1847,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
       // deducing through the noexcept-specifier if it's part of the canonical
       // type. libstdc++ relies on this.
       Expr *NoexceptExpr = FunctionProtoParam->getNoexceptExpr();
-      if (NonTypeTemplateParmDecl *NTTP =
+      if (const NonTypeTemplateParmDecl *NTTP =
           NoexceptExpr ? getDeducedParameterFromExpr(Info, NoexceptExpr)
                        : nullptr) {
         assert(NTTP->getDepth() == Info.getDeducedDepth() &&
@@ -2015,7 +2036,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
           return Result;
 
         // Perform deduction on the vector size, if we can.
-        NonTypeTemplateParmDecl *NTTP =
+        const NonTypeTemplateParmDecl *NTTP =
             getDeducedParameterFromExpr(Info, VectorParam->getSizeExpr());
         if (!NTTP)
           return Sema::TDK_Success;
@@ -2039,7 +2060,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
           return Result;
 
         // Perform deduction on the vector size, if we can.
-        NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(
+        const NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(
             Info, VectorParam->getSizeExpr());
         if (!NTTP)
           return Sema::TDK_Success;
@@ -2068,8 +2089,8 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
           return Result;
 
         // Perform deduction on the vector size, if we can.
-        NonTypeTemplateParmDecl *NTTP
-          = getDeducedParameterFromExpr(Info, VectorParam->getSizeExpr());
+        const NonTypeTemplateParmDecl *NTTP =
+            getDeducedParameterFromExpr(Info, VectorParam->getSizeExpr());
         if (!NTTP)
           return Sema::TDK_Success;
 
@@ -2094,8 +2115,8 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
           return Result;
 
         // Perform deduction on the vector size, if we can.
-        NonTypeTemplateParmDecl *NTTP
-          = getDeducedParameterFromExpr(Info, VectorParam->getSizeExpr());
+        const NonTypeTemplateParmDecl *NTTP =
+            getDeducedParameterFromExpr(Info, VectorParam->getSizeExpr());
         if (!NTTP)
           return Sema::TDK_Success;
 
@@ -2171,7 +2192,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
               return Sema::TDK_NonDeducedMismatch;
             }
 
-            NonTypeTemplateParmDecl *NTTP =
+            const NonTypeTemplateParmDecl *NTTP =
                 getDeducedParameterFromExpr(Info, ParamExpr);
             if (!NTTP)
               return Sema::TDK_Success;
@@ -2218,7 +2239,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
           return Result;
 
         // Perform deduction on the address space, if we can.
-        NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(
+        const NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(
             Info, AddressSpaceParam->getAddrSpaceExpr());
         if (!NTTP)
           return Sema::TDK_Success;
@@ -2241,7 +2262,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
           return Result;
 
         // Perform deduction on the address space, if we can.
-        NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(
+        const NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(
             Info, AddressSpaceParam->getAddrSpaceExpr());
         if (!NTTP)
           return Sema::TDK_Success;
@@ -2260,7 +2281,7 @@ DeduceTemplateArgumentsByTypeMatch(Sema &S,
         if (IntParam->isUnsigned() != IntArg->isUnsigned())
           return Sema::TDK_NonDeducedMismatch;
 
-        NonTypeTemplateParmDecl *NTTP =
+        const NonTypeTemplateParmDecl *NTTP =
             getDeducedParameterFromExpr(Info, IntParam->getNumBitsExpr());
         if (!NTTP)
           return Sema::TDK_Success;
@@ -2377,8 +2398,8 @@ DeduceTemplateArguments(Sema &S,
     return Sema::TDK_NonDeducedMismatch;
 
   case TemplateArgument::Expression:
-    if (NonTypeTemplateParmDecl *NTTP
-          = getDeducedParameterFromExpr(Info, Param.getAsExpr())) {
+    if (const NonTypeTemplateParmDecl *NTTP =
+            getDeducedParameterFromExpr(Info, Param.getAsExpr())) {
       if (Arg.getKind() == TemplateArgument::Integral)
         return DeduceNonTypeTemplateArgument(S, TemplateParams, NTTP,
                                              Arg.getAsIntegral(),
@@ -3982,7 +4003,7 @@ static Sema::TemplateDeductionResult DeduceFromInitializerList(
   //   from the length of the initializer list.
   if (auto *DependentArrTy = dyn_cast_or_null<DependentSizedArrayType>(ArrTy)) {
     // Determine the array bound is something we can deduce.
-    if (NonTypeTemplateParmDecl *NTTP =
+    if (const NonTypeTemplateParmDecl *NTTP =
             getDeducedParameterFromExpr(Info, DependentArrTy->getSizeExpr())) {
       // We can perform template argument deduction for the given non-type
       // template parameter.
@@ -5728,26 +5749,7 @@ MarkUsedTemplateParameters(ASTContext &Ctx,
   if (const PackExpansionExpr *Expansion = dyn_cast<PackExpansionExpr>(E))
     E = Expansion->getPattern();
 
-  // Skip through any implicit casts we added while type-checking, and any
-  // substitutions performed by template alias expansion.
-  while (true) {
-    if (const ImplicitCastExpr *ICE = dyn_cast<ImplicitCastExpr>(E))
-      E = ICE->getSubExpr();
-    else if (const ConstantExpr *CE = dyn_cast<ConstantExpr>(E))
-      E = CE->getSubExpr();
-    else if (const SubstNonTypeTemplateParmExpr *Subst =
-               dyn_cast<SubstNonTypeTemplateParmExpr>(E))
-      E = Subst->getReplacement();
-    else
-      break;
-  }
-
-  const DeclRefExpr *DRE = dyn_cast<DeclRefExpr>(E);
-  if (!DRE)
-    return;
-
-  const NonTypeTemplateParmDecl *NTTP
-    = dyn_cast<NonTypeTemplateParmDecl>(DRE->getDecl());
+  const NonTypeTemplateParmDecl *NTTP = getDeducedParameterFromExpr(E, Depth);
   if (!NTTP)
     return;
 
diff --git a/clang/lib/Sema/SemaTemplateInstantiate.cpp b/clang/lib/Sema/SemaTemplateInstantiate.cpp
index 6e5b214687a0f..57114e4cab815 100644
--- a/clang/lib/Sema/SemaTemplateInstantiate.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiate.cpp
@@ -1517,9 +1517,12 @@ TemplateInstantiator::TransformTemplateParmRefExpr(DeclRefExpr *E,
       if (TargetType.isNull())
         return ExprError();
 
+      QualType ExprType = TargetType.getNonLValueExprType(SemaRef.Context);
+      if (TargetType->isRecordType())
+        ExprType.addConst();
+
       return new (SemaRef.Context) SubstNonTypeTemplateParmPackExpr(
-          TargetType.getNonLValueExprType(SemaRef.Context),
-          TargetType->isReferenceType() ? VK_LValue : VK_RValue, NTTP,
+          ExprType, TargetType->isReferenceType() ? VK_LValue : VK_RValue, NTTP,
           E->getLocation(), Arg);
     }
 
@@ -1614,15 +1617,39 @@ ExprResult TemplateInstantiator::transformNonTypeTemplateParmRef(
                                                  SourceLocation loc,
                                                  TemplateArgument arg) {
   ExprResult result;
-  QualType type;
 
-  // The template argument itself might be an expression, in which
-  // case we just return that expression.
+  // Determine the substituted parameter type. We can usually infer this from
+  // the template argument, but not always.
+  auto SubstParamType = [&] {
+    QualType T;
+    if (parm->isExpandedParameterPack())
+      T = parm->getExpansionType(SemaRef.ArgumentPackSubstitutionIndex);
+    else
+      T = parm->getType();
+    if (parm->isParameterPack() && isa<PackExpansionType>(T))
+      T = cast<PackExpansionType>(T)->getPattern();
+    return SemaRef.SubstType(T, TemplateArgs, loc, parm->getDeclName());
+  };
+
+  bool refParam = false;
+
+  // The template argument itself might be an expression, in which case we just
+  // return that expression. This happens when substituting into an alias
+  // template.
   if (arg.getKind() == TemplateArgument::Expression) {
     Expr *argExpr = arg.getAsExpr();
     result = argExpr;
-    type = argExpr->getType();
-
+    if (argExpr->isLValue()) {
+      if (argExpr->getType()->isRecordType()) {
+        // Check whether the parameter was actually a reference.
+        QualType paramType = SubstParamType();
+        if (paramType.isNull())
+          return ExprError();
+        refParam = paramType->isReferenceType();
+      } else {
+        refParam = true;
+      }
+    }
   } else if (arg.getKind() == TemplateArgument::Declaration ||
              arg.getKind() == TemplateArgument::NullPtr) {
     ValueDecl *VD;
@@ -1640,36 +1667,25 @@ ExprResult TemplateInstantiator::transformNonTypeTemplateParmRef(
       VD = nullptr;
     }
 
-    // Derive the type we want the substituted decl to have.  This had
-    // better be non-dependent, or these checks will have serious problems.
-    if (parm->isExpandedParameterPack()) {
-      type = parm->getExpansionType(SemaRef.ArgumentPackSubstitutionIndex);
-    } else if (parm->isParameterPack() &&
-               isa<PackExpansionType>(parm->getType())) {
-      type = SemaRef.SubstType(
-                        cast<PackExpansionType>(parm->getType())->getPattern(),
-                                     TemplateArgs, loc, parm->getDeclName());
-    } else {
-      type = SemaRef.SubstType(VD ? arg.getParamTypeForDecl() : arg.getNullPtrType(),
-                               TemplateArgs, loc, parm->getDeclName());
-    }
-    assert(!type.isNull() && "type substitution failed for param type");
-    assert(!type->isDependentType() && "param type still dependent");
-    result = SemaRef.BuildExpressionFromDeclTemplateArgument(arg, type, loc);
-
-    if (!result.isInvalid()) type = result.get()->getType();
+    QualType paramType = VD ? arg.getParamTypeForDecl() : arg.getNullPtrType();
+    assert(!paramType.isNull() && "type substitution failed for param type");
+    assert(!paramType->isDependentType() && "param type still dependent");
+    result = SemaRef.BuildExpressionFromDeclTemplateArgument(arg, paramType, loc);
+    refParam = paramType->isReferenceType();
   } else {
     result = SemaRef.BuildExpressionFromIntegralTemplateArgument(arg, loc);
-
-    // Note that this type can be different from the type of 'result',
-    // e.g. if it's an enum type.
-    type = arg.getIntegralType();
+    assert(result.isInvalid() ||
+           SemaRef.Context.hasSameType(result.get()->getType(),
+                                       arg.getIntegralType()));
   }
-  if (result.isInvalid()) return ExprError();
+
+  if (result.isInvalid())
+    return ExprError();
 
   Expr *resultExpr = result.get();
   return new (SemaRef.Context) SubstNonTypeTemplateParmExpr(
-      type, resultExpr->getValueKind(), loc, parm, resultExpr);
+      resultExpr->getType(), resultExpr->getValueKind(), loc, parm, refParam,
+      resultExpr);
 }
 
 ExprResult
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index bacd990475738..092f5b83fd4e4 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -846,6 +846,11 @@ Decl *TemplateDeclInstantiator::VisitMSGuidDecl(MSGuidDecl *D) {
   llvm_unreachable("GUID declaration cannot be instantiated");
 }
 
+Decl *TemplateDeclInstantiator::VisitTemplateParamObjectDecl(
+    TemplateParamObjectDecl *D) {
+  llvm_unreachable("template parameter objects cannot be instantiated");
+}
+
 Decl *
 TemplateDeclInstantiator::VisitLabelDecl(LabelDecl *D) {
   LabelDecl *Inst = LabelDecl::Create(SemaRef.Context, Owner, D->getLocation(),
diff --git a/clang/lib/Sema/SemaType.cpp b/clang/lib/Sema/SemaType.cpp
index d887e11895b8f..1eb94aeeeae3b 100644
--- a/clang/lib/Sema/SemaType.cpp
+++ b/clang/lib/Sema/SemaType.cpp
@@ -3386,8 +3386,9 @@ static QualType GetDeclSpecTypeForDeclarator(TypeProcessingState &state,
       Error = 7; // Exception declaration
       break;
     case DeclaratorContext::TemplateParamContext:
-      if (isa<DeducedTemplateSpecializationType>(Deduced))
-        Error = 19; // Template parameter
+      if (isa<DeducedTemplateSpecializationType>(Deduced) &&
+          !SemaRef.getLangOpts().CPlusPlus20)
+        Error = 19; // Template parameter (until C++20)
       else if (!SemaRef.getLangOpts().CPlusPlus17)
         Error = 8; // Template parameter (until C++17)
       break;
@@ -8978,7 +8979,17 @@ static QualType getDecltypeForExpr(Sema &S, Expr *E) {
 
   // C++11 [dcl.type.simple]p4:
   //   The type denoted by decltype(e) is defined as follows:
-  //
+
+  // C++20:
+  //     - if E is an unparenthesized id-expression naming a non-type
+  //       template-parameter (13.2), decltype(E) is the type of the
+  //       template-parameter after performing any necessary type deduction
+  // Note that this does not pick up the implicit 'const' for a template
+  // parameter object. This rule makes no difference before C++20 so we apply
+  // it unconditionally.
+  if (const auto *SNTTPE = dyn_cast<SubstNonTypeTemplateParmExpr>(E))
+    return SNTTPE->getParameterType(S.Context);
+
   //     - if e is an unparenthesized id-expression or an unparenthesized class
   //       member access (5.2.5), decltype(e) is the type of the entity named
   //       by e. If there is no such entity, or if e names a set of overloaded
@@ -8987,6 +8998,8 @@ static QualType getDecltypeForExpr(Sema &S, Expr *E) {
   // We apply the same rules for Objective-C ivar and property references.
   if (const DeclRefExpr *DRE = dyn_cast<DeclRefExpr>(E)) {
     const ValueDecl *VD = DRE->getDecl();
+    if (auto *TPO = dyn_cast<TemplateParamObjectDecl>(VD))
+      return TPO->getType().getUnqualifiedType();
     return VD->getType();
   } else if (const MemberExpr *ME = dyn_cast<MemberExpr>(E)) {
     if (const ValueDecl *VD = ME->getMemberDecl())
diff --git a/clang/lib/Sema/TreeTransform.h b/clang/lib/Sema/TreeTransform.h
index a6e7c2666c023..daf6d222610a9 100644
--- a/clang/lib/Sema/TreeTransform.h
+++ b/clang/lib/Sema/TreeTransform.h
@@ -1307,7 +1307,7 @@ class TreeTransform {
     return SemaRef.ActOnLabelStmt(IdentLoc, L, ColonLoc, SubStmt);
   }
 
-  /// Build a new label statement.
+  /// Build a new attributed statement.
   ///
   /// By default, performs semantic analysis to build the new statement.
   /// Subclasses may override this routine to provide different behavior.
diff --git a/clang/lib/Serialization/ASTCommon.cpp b/clang/lib/Serialization/ASTCommon.cpp
index 75c510d7aa570..9a202692ec980 100644
--- a/clang/lib/Serialization/ASTCommon.cpp
+++ b/clang/lib/Serialization/ASTCommon.cpp
@@ -385,6 +385,7 @@ bool serialization::isRedeclarableDeclKind(unsigned Kind) {
   case Decl::Field:
   case Decl::MSProperty:
   case Decl::MSGuid:
+  case Decl::TemplateParamObject:
   case Decl::ObjCIvar:
   case Decl::ObjCAtDefsField:
   case Decl::NonTypeTemplateParm:
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 94a4c48063d2b..4ab6d4a3b42c7 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -8970,48 +8970,148 @@ ReadFixedPointSemantics(const SmallVectorImpl<uint64_t> &Record,
                                    HasUnsignedPadding);
 }
 
-static const llvm::fltSemantics &
-readAPFloatSemantics(ASTRecordReader &reader) {
-  return llvm::APFloatBase::EnumToSemantics(
-    static_cast<llvm::APFloatBase::Semantics>(reader.readInt()));
-}
-
 APValue ASTRecordReader::readAPValue() {
-  unsigned Kind = readInt();
-  switch ((APValue::ValueKind) Kind) {
+  auto Kind = static_cast<APValue::ValueKind>(asImpl().readUInt32());
+  switch (Kind) {
   case APValue::None:
     return APValue();
   case APValue::Indeterminate:
     return APValue::IndeterminateValue();
   case APValue::Int:
-    return APValue(readAPSInt());
+    return APValue(asImpl().readAPSInt());
   case APValue::Float: {
-    const llvm::fltSemantics &FloatSema = readAPFloatSemantics(*this);
-    return APValue(readAPFloat(FloatSema));
+    const llvm::fltSemantics &FloatSema = llvm::APFloatBase::EnumToSemantics(
+        static_cast<llvm::APFloatBase::Semantics>(asImpl().readUInt32()));
+    return APValue(asImpl().readAPFloat(FloatSema));
   }
   case APValue::FixedPoint: {
     llvm::FixedPointSemantics FPSema = ReadFixedPointSemantics(Record, Idx);
     return APValue(llvm::APFixedPoint(readAPInt(), FPSema));
   }
   case APValue::ComplexInt: {
-    llvm::APSInt First = readAPSInt();
-    return APValue(std::move(First), readAPSInt());
+    llvm::APSInt First = asImpl().readAPSInt();
+    return APValue(std::move(First), asImpl().readAPSInt());
   }
   case APValue::ComplexFloat: {
-    const llvm::fltSemantics &FloatSema1 = readAPFloatSemantics(*this);
-    llvm::APFloat First = readAPFloat(FloatSema1);
-    const llvm::fltSemantics &FloatSema2 = readAPFloatSemantics(*this);
-    return APValue(std::move(First), readAPFloat(FloatSema2));
-  }
-  case APValue::LValue:
-  case APValue::Vector:
-  case APValue::Array:
-  case APValue::Struct:
-  case APValue::Union:
-  case APValue::MemberPointer:
-  case APValue::AddrLabelDiff:
-    // TODO : Handle all these APValue::ValueKind.
-    return APValue();
+    const llvm::fltSemantics &FloatSema = llvm::APFloatBase::EnumToSemantics(
+        static_cast<llvm::APFloatBase::Semantics>(asImpl().readUInt32()));
+    llvm::APFloat First = readAPFloat(FloatSema);
+    return APValue(std::move(First), asImpl().readAPFloat(FloatSema));
+  }
+  case APValue::Vector: {
+    APValue Result;
+    Result.MakeVector();
+    unsigned Length = asImpl().readUInt32();
+    (void)Result.setVectorUninit(Length);
+    for (unsigned LoopIdx = 0; LoopIdx < Length; LoopIdx++)
+      Result.getVectorElt(LoopIdx) = asImpl().readAPValue();
+    return Result;
+  }
+  case APValue::Array: {
+    APValue Result;
+    unsigned InitLength = asImpl().readUInt32();
+    unsigned TotalLength = asImpl().readUInt32();
+    Result.MakeArray(InitLength, TotalLength);
+    for (unsigned LoopIdx = 0; LoopIdx < InitLength; LoopIdx++)
+      Result.getArrayInitializedElt(LoopIdx) = asImpl().readAPValue();
+    if (Result.hasArrayFiller())
+      Result.getArrayFiller() = asImpl().readAPValue();
+    return Result;
+  }
+  case APValue::Struct: {
+    APValue Result;
+    unsigned BasesLength = asImpl().readUInt32();
+    unsigned FieldsLength = asImpl().readUInt32();
+    Result.MakeStruct(BasesLength, FieldsLength);
+    for (unsigned LoopIdx = 0; LoopIdx < BasesLength; LoopIdx++)
+      Result.getStructBase(LoopIdx) = asImpl().readAPValue();
+    for (unsigned LoopIdx = 0; LoopIdx < FieldsLength; LoopIdx++)
+      Result.getStructField(LoopIdx) = asImpl().readAPValue();
+    return Result;
+  }
+  case APValue::Union: {
+    auto *FDecl = asImpl().readDeclAs<FieldDecl>();
+    APValue Value = asImpl().readAPValue();
+    return APValue(FDecl, std::move(Value));
+  }
+  case APValue::AddrLabelDiff: {
+    auto *LHS = cast<AddrLabelExpr>(asImpl().readExpr());
+    auto *RHS = cast<AddrLabelExpr>(asImpl().readExpr());
+    return APValue(LHS, RHS);
+  }
+  case APValue::MemberPointer: {
+    APValue Result;
+    bool IsDerived = asImpl().readUInt32();
+    auto *Member = asImpl().readDeclAs<ValueDecl>();
+    unsigned PathSize = asImpl().readUInt32();
+    const CXXRecordDecl **PathArray =
+        Result.setMemberPointerUninit(Member, IsDerived, PathSize).data();
+    for (unsigned LoopIdx = 0; LoopIdx < PathSize; LoopIdx++)
+      PathArray[LoopIdx] =
+          asImpl().readDeclAs<const CXXRecordDecl>()->getCanonicalDecl();
+    return Result;
+  }
+  case APValue::LValue: {
+    uint64_t Bits = asImpl().readUInt32();
+    bool HasLValuePath = Bits & 0x1;
+    bool IsLValueOnePastTheEnd = Bits & 0x2;
+    bool IsExpr = Bits & 0x4;
+    bool IsTypeInfo = Bits & 0x8;
+    bool IsNullPtr = Bits & 0x10;
+    bool HasBase = Bits & 0x20;
+    APValue::LValueBase Base;
+    QualType ElemTy;
+    assert((!IsExpr || !IsTypeInfo) && "LValueBase cannot be both");
+    if (HasBase) {
+      if (!IsTypeInfo) {
+        unsigned CallIndex = asImpl().readUInt32();
+        unsigned Version = asImpl().readUInt32();
+        if (IsExpr) {
+          Base = APValue::LValueBase(asImpl().readExpr(), CallIndex, Version);
+          ElemTy = Base.get<const Expr *>()->getType();
+        } else {
+          Base = APValue::LValueBase(asImpl().readDeclAs<const ValueDecl>(),
+                                     CallIndex, Version);
+          ElemTy = Base.get<const ValueDecl *>()->getType();
+        }
+      } else {
+        QualType TypeInfo = asImpl().readType();
+        QualType Type = asImpl().readType();
+        Base = APValue::LValueBase::getTypeInfo(
+            TypeInfoLValue(TypeInfo.getTypePtr()), Type);
+        Base.getTypeInfoType();
+      }
+    }
+    CharUnits Offset = CharUnits::fromQuantity(asImpl().readUInt32());
+    unsigned PathLength = asImpl().readUInt32();
+    APValue Result;
+    Result.MakeLValue();
+    if (HasLValuePath) {
+      APValue::LValuePathEntry *Path =
+          Result
+              .setLValueUninit(Base, Offset, PathLength, IsLValueOnePastTheEnd,
+                               IsNullPtr)
+              .data();
+      for (unsigned LoopIdx = 0; LoopIdx < PathLength; LoopIdx++) {
+        if (ElemTy->getAs<RecordType>()) {
+          unsigned Int = asImpl().readUInt32();
+          Decl *D = asImpl().readDeclAs<Decl>();
+          if (auto *RD = dyn_cast<CXXRecordDecl>(D))
+            ElemTy = getASTContext().getRecordType(RD);
+          else
+            ElemTy = cast<ValueDecl>(D)->getType();
+          Path[LoopIdx] =
+              APValue::LValuePathEntry(APValue::BaseOrMemberType(D, Int));
+        } else {
+          ElemTy = getASTContext().getAsArrayType(ElemTy)->getElementType();
+          Path[LoopIdx] =
+              APValue::LValuePathEntry::ArrayIndex(asImpl().readUInt32());
+        }
+      }
+    } else
+      Result.setLValue(Base, Offset, APValue::NoLValuePath{}, IsNullPtr);
+    return Result;
+  }
   }
   llvm_unreachable("Invalid APValue::ValueKind");
 }
diff --git a/clang/lib/Serialization/ASTReaderDecl.cpp b/clang/lib/Serialization/ASTReaderDecl.cpp
index 4d50e4af31f7b..fd3b981150a0e 100644
--- a/clang/lib/Serialization/ASTReaderDecl.cpp
+++ b/clang/lib/Serialization/ASTReaderDecl.cpp
@@ -369,6 +369,7 @@ namespace clang {
     void VisitFieldDecl(FieldDecl *FD);
     void VisitMSPropertyDecl(MSPropertyDecl *FD);
     void VisitMSGuidDecl(MSGuidDecl *D);
+    void VisitTemplateParamObjectDecl(TemplateParamObjectDecl *D);
     void VisitIndirectFieldDecl(IndirectFieldDecl *FD);
     RedeclarableResult VisitVarDeclImpl(VarDecl *D);
     void VisitVarDecl(VarDecl *VD) { VisitVarDeclImpl(VD); }
@@ -1377,6 +1378,17 @@ void ASTDeclReader::VisitMSGuidDecl(MSGuidDecl *D) {
     Reader.getContext().setPrimaryMergedDecl(D, Existing->getCanonicalDecl());
 }
 
+void ASTDeclReader::VisitTemplateParamObjectDecl(TemplateParamObjectDecl *D) {
+  VisitValueDecl(D);
+  D->Value = Record.readAPValue();
+
+  // Add this template parameter object to the AST context's lookup structure,
+  // and merge if needed.
+  if (TemplateParamObjectDecl *Existing =
+          Reader.getContext().TemplateParamObjectDecls.GetOrInsertNode(D))
+    Reader.getContext().setPrimaryMergedDecl(D, Existing->getCanonicalDecl());
+}
+
 void ASTDeclReader::VisitIndirectFieldDecl(IndirectFieldDecl *FD) {
   VisitValueDecl(FD);
 
@@ -2409,8 +2421,10 @@ void ASTDeclReader::VisitLifetimeExtendedTemporaryDecl(
   VisitDecl(D);
   D->ExtendingDecl = readDeclAs<ValueDecl>();
   D->ExprWithTemporary = Record.readStmt();
-  if (Record.readInt())
+  if (Record.readInt()) {
     D->Value = new (D->getASTContext()) APValue(Record.readAPValue());
+    D->getASTContext().addDestruction(D->Value);
+  }
   D->ManglingNumber = Record.readInt();
   mergeMergeable(D);
 }
@@ -3983,6 +3997,9 @@ Decl *ASTReader::ReadDeclRecord(DeclID ID) {
   case DECL_MS_GUID:
     D = MSGuidDecl::CreateDeserialized(Context, ID);
     break;
+  case DECL_TEMPLATE_PARAM_OBJECT:
+    D = TemplateParamObjectDecl::CreateDeserialized(Context, ID);
+    break;
   case DECL_CAPTURED:
     D = CapturedDecl::CreateDeserialized(Context, ID, Record.readInt());
     break;
diff --git a/clang/lib/Serialization/ASTReaderStmt.cpp b/clang/lib/Serialization/ASTReaderStmt.cpp
index 363527f884b33..cf335302f7f08 100644
--- a/clang/lib/Serialization/ASTReaderStmt.cpp
+++ b/clang/lib/Serialization/ASTReaderStmt.cpp
@@ -2121,7 +2121,8 @@ void ASTStmtReader::VisitSizeOfPackExpr(SizeOfPackExpr *E) {
 void ASTStmtReader::VisitSubstNonTypeTemplateParmExpr(
                                               SubstNonTypeTemplateParmExpr *E) {
   VisitExpr(E);
-  E->Param = readDeclAs<NonTypeTemplateParmDecl>();
+  E->ParamAndRef.setPointer(readDeclAs<NonTypeTemplateParmDecl>());
+  E->ParamAndRef.setInt(Record.readInt());
   E->SubstNonTypeTemplateParmExprBits.NameLoc = readSourceLocation();
   E->Replacement = Record.readSubExpr();
 }
diff --git a/clang/lib/Serialization/ASTWriter.cpp b/clang/lib/Serialization/ASTWriter.cpp
index ed00a3bc62812..b23e04279d29e 100644
--- a/clang/lib/Serialization/ASTWriter.cpp
+++ b/clang/lib/Serialization/ASTWriter.cpp
@@ -5155,22 +5155,105 @@ void ASTRecordWriter::AddAPValue(const APValue &Value) {
     return;
   }
   case APValue::ComplexFloat: {
+    assert(llvm::APFloatBase::SemanticsToEnum(
+               Value.getComplexFloatImag().getSemantics()) ==
+           llvm::APFloatBase::SemanticsToEnum(
+               Value.getComplexFloatReal().getSemantics()));
     push_back(static_cast<uint64_t>(llvm::APFloatBase::SemanticsToEnum(
         Value.getComplexFloatReal().getSemantics())));
     AddAPFloat(Value.getComplexFloatReal());
-    push_back(static_cast<uint64_t>(llvm::APFloatBase::SemanticsToEnum(
-        Value.getComplexFloatImag().getSemantics())));
     AddAPFloat(Value.getComplexFloatImag());
     return;
   }
-  case APValue::LValue:
   case APValue::Vector:
+    push_back(Value.getVectorLength());
+    for (unsigned Idx = 0; Idx < Value.getVectorLength(); Idx++)
+      AddAPValue(Value.getVectorElt(Idx));
+    return;
   case APValue::Array:
+    push_back(Value.getArrayInitializedElts());
+    push_back(Value.getArraySize());
+    for (unsigned Idx = 0; Idx < Value.getArrayInitializedElts(); Idx++)
+      AddAPValue(Value.getArrayInitializedElt(Idx));
+    if (Value.hasArrayFiller())
+      AddAPValue(Value.getArrayFiller());
+    return;
   case APValue::Struct:
+    push_back(Value.getStructNumBases());
+    push_back(Value.getStructNumFields());
+    for (unsigned Idx = 0; Idx < Value.getStructNumBases(); Idx++)
+      AddAPValue(Value.getStructBase(Idx));
+    for (unsigned Idx = 0; Idx < Value.getStructNumFields(); Idx++)
+      AddAPValue(Value.getStructField(Idx));
+    return;
   case APValue::Union:
-  case APValue::MemberPointer:
+    AddDeclRef(Value.getUnionField());
+    AddAPValue(Value.getUnionValue());
+    return;
   case APValue::AddrLabelDiff:
-    // TODO : Handle all these APValue::ValueKind.
+    AddStmt(const_cast<AddrLabelExpr *>(Value.getAddrLabelDiffLHS()));
+    AddStmt(const_cast<AddrLabelExpr *>(Value.getAddrLabelDiffRHS()));
+    return;
+  case APValue::MemberPointer: {
+    push_back(Value.isMemberPointerToDerivedMember());
+    AddDeclRef(Value.getMemberPointerDecl());
+    ArrayRef<const CXXRecordDecl *> RecordPath = Value.getMemberPointerPath();
+    push_back(RecordPath.size());
+    for (auto Elem : RecordPath)
+      AddDeclRef(Elem);
+    return;
+  }
+  case APValue::LValue: {
+    push_back(Value.hasLValuePath() | Value.isLValueOnePastTheEnd() << 1 |
+              Value.getLValueBase().is<const Expr *>() << 2 |
+              Value.getLValueBase().is<TypeInfoLValue>() << 3 |
+              Value.isNullPointer() << 4 |
+              static_cast<bool>(Value.getLValueBase()) << 5);
+    QualType ElemTy;
+    if (Value.getLValueBase()) {
+      assert(!Value.getLValueBase().is<DynamicAllocLValue>() &&
+             "in C++20 dynamic allocation are transient so they shouldn't "
+             "appear in the AST");
+      if (!Value.getLValueBase().is<TypeInfoLValue>()) {
+        push_back(Value.getLValueBase().getCallIndex());
+        push_back(Value.getLValueBase().getVersion());
+        if (const auto *E = Value.getLValueBase().dyn_cast<const Expr *>()) {
+          AddStmt(const_cast<Expr *>(E));
+          ElemTy = E->getType();
+        } else {
+          AddDeclRef(Value.getLValueBase().get<const ValueDecl *>());
+          ElemTy = Value.getLValueBase().get<const ValueDecl *>()->getType();
+        }
+      } else {
+        AddTypeRef(
+            QualType(Value.getLValueBase().get<TypeInfoLValue>().getType(), 0));
+        AddTypeRef(Value.getLValueBase().getTypeInfoType());
+        ElemTy = Value.getLValueBase().getTypeInfoType();
+      }
+    }
+    push_back(Value.getLValueOffset().getQuantity());
+    push_back(Value.getLValuePath().size());
+    if (Value.hasLValuePath()) {
+      ArrayRef<APValue::LValuePathEntry> Path = Value.getLValuePath();
+      for (auto Elem : Path) {
+        if (ElemTy->getAs<RecordType>()) {
+          push_back(Elem.getAsBaseOrMember().getInt());
+          const Decl *BaseOrMember = Elem.getAsBaseOrMember().getPointer();
+          if (const auto *RD = dyn_cast<CXXRecordDecl>(BaseOrMember)) {
+            AddDeclRef(RD);
+            ElemTy = Writer->Context->getRecordType(RD);
+          } else {
+            const auto *VD = cast<ValueDecl>(BaseOrMember);
+            AddDeclRef(VD);
+            ElemTy = VD->getType();
+          }
+        } else {
+          push_back(Elem.getAsArrayIndex());
+          ElemTy = Writer->Context->getAsArrayType(ElemTy)->getElementType();
+        }
+      }
+    }
+  }
     return;
   }
   llvm_unreachable("Invalid APValue::ValueKind");
diff --git a/clang/lib/Serialization/ASTWriterDecl.cpp b/clang/lib/Serialization/ASTWriterDecl.cpp
index 5b0a6ffb9566c..df73044574194 100644
--- a/clang/lib/Serialization/ASTWriterDecl.cpp
+++ b/clang/lib/Serialization/ASTWriterDecl.cpp
@@ -96,6 +96,7 @@ namespace clang {
     void VisitFieldDecl(FieldDecl *D);
     void VisitMSPropertyDecl(MSPropertyDecl *D);
     void VisitMSGuidDecl(MSGuidDecl *D);
+    void VisitTemplateParamObjectDecl(TemplateParamObjectDecl *D);
     void VisitIndirectFieldDecl(IndirectFieldDecl *D);
     void VisitVarDecl(VarDecl *D);
     void VisitImplicitParamDecl(ImplicitParamDecl *D);
@@ -965,6 +966,12 @@ void ASTDeclWriter::VisitMSGuidDecl(MSGuidDecl *D) {
   Code = serialization::DECL_MS_GUID;
 }
 
+void ASTDeclWriter::VisitTemplateParamObjectDecl(TemplateParamObjectDecl *D) {
+  VisitValueDecl(D);
+  Record.AddAPValue(D->getValue());
+  Code = serialization::DECL_TEMPLATE_PARAM_OBJECT;
+}
+
 void ASTDeclWriter::VisitIndirectFieldDecl(IndirectFieldDecl *D) {
   VisitValueDecl(D);
   Record.push_back(D->getChainingSize());
diff --git a/clang/lib/Serialization/ASTWriterStmt.cpp b/clang/lib/Serialization/ASTWriterStmt.cpp
index 0121f25832073..522cd3247f342 100644
--- a/clang/lib/Serialization/ASTWriterStmt.cpp
+++ b/clang/lib/Serialization/ASTWriterStmt.cpp
@@ -2018,6 +2018,7 @@ void ASTStmtWriter::VisitSubstNonTypeTemplateParmExpr(
                                               SubstNonTypeTemplateParmExpr *E) {
   VisitExpr(E);
   Record.AddDeclRef(E->getParameter());
+  Record.push_back(E->isReferenceParameter());
   Record.AddSourceLocation(E->getNameLoc());
   Record.AddStmt(E->getReplacement());
   Code = serialization::EXPR_SUBST_NON_TYPE_TEMPLATE_PARM;
diff --git a/clang/test/ASTMerge/APValue/APValue.cpp b/clang/test/ASTMerge/APValue/APValue.cpp
new file mode 100644
index 0000000000000..c8695fb6074bb
--- /dev/null
+++ b/clang/test/ASTMerge/APValue/APValue.cpp
@@ -0,0 +1,462 @@
+// RUN: %clang_cc1 -std=gnu++2a -emit-pch %s -o %t.pch
+// RUN: %clang_cc1 -std=gnu++2a %s -DEMIT -ast-merge %t.pch -ast-dump-all | FileCheck %s
+
+// XFAIL: *
+
+#ifndef EMIT
+#define EMIT
+
+namespace Integer {
+
+consteval int fint() {
+  return 6789;
+}
+
+int Unique_Int = fint();
+//CHECK:      VarDecl {{.*}} Unique_Int
+//CHECK-NEXT: ConstantExpr {{.*}} 'int'
+//CHECK-NEXT: value: Int 6789
+
+consteval __uint128_t fint128() {
+  return ((__uint128_t)0x75f17d6b3588f843 << 64) | 0xb13dea7c9c324e51;
+}
+
+constexpr __uint128_t Unique_Int128 = fint128();
+//CHECK:      VarDecl {{.*}} Unique_Int128
+//CHECK-NEXT: value: Int 156773562844924187900898496343692168785
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Int 156773562844924187900898496343692168785
+
+} // namespace Integer
+
+namespace FloatingPoint {
+
+consteval double fdouble() {
+  return double(567890.67890);
+}
+
+double Unique_Double = fdouble();
+//CHECK:      VarDecl {{.*}} Unique_Double
+//CHECK-NEXT: ConstantExpr {{.*}}
+//CHECK-NEXT: value: Float 5.678907e+05
+
+} // namespace FloatingPoint
+
+// FIXME: Add test for FixedPoint, ComplexInt, ComplexFloat, AddrLabelDiff.
+
+namespace Struct {
+
+struct B {
+  int i;
+  double d;
+};
+
+consteval B fB() {
+  return B{1, 0.7};
+}
+
+constexpr B Basic_Struct = fB();
+//CHECK:      VarDecl {{.*}} Basic_Struct
+//CHECK-NEXT: value: Struct
+//CHECK-NEXT: fields: Int 1, Float 7.000000e-01
+//CHECK-NEXT: ImplicitCastExpr
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Struct
+//CHECK-NEXT: fields: Int 1, Float 7.000000e-01
+
+struct C {
+  int i = 9;
+};
+
+struct A : B {
+  constexpr A(B b, int I, double D, C _c) : B(b), i(I), d(D), c(_c) {}
+  int i;
+  double d;
+  C c;
+};
+
+consteval A fA() {
+  return A(Basic_Struct, 1, 79.789, {});
+}
+
+A Advanced_Struct = fA();
+//CHECK:      VarDecl {{.*}} Advanced_Struct
+//CHECK-NEXT: ConstantExpr {{.*}}
+//CHECK-NEXT: value: Struct
+//CHECK-NEXT: base: Struct
+//CHECK-NEXT: fields: Int 1, Float 7.000000e-01
+//CHECK-NEXT: fields: Int 1, Float 7.978900e+01
+//CHECK-NEXT: field: Struct
+//CHECK-NEXT: field: Int 9
+
+} // namespace Struct
+
+namespace Vector {
+
+using v4si = int __attribute__((__vector_size__(16)));
+
+consteval v4si fv4si() {
+  return (v4si){8, 2, 3};
+}
+
+v4si Vector_Int = fv4si();
+//CHECK:      VarDecl {{.*}} Vector_Int
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Vector length=4
+//CHECK-NEXT: elements: Int 8, Int 2, Int 3, Int 0
+
+} // namespace Vector
+
+namespace Array {
+
+struct B {
+  int arr[6];
+};
+
+consteval B fint() {
+  return B{1, 2, 3, 4, 5, 6};
+}
+
+B Array_Int = fint();
+//CHECK:      VarDecl {{.*}} Array_Int
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Struct
+//CHECK-NEXT: field: Array size=6
+//CHECK-NEXT: elements: Int 1, Int 2, Int 3, Int 4
+//CHECK-NEXT: elements: Int 5, Int 6
+
+struct A {
+  int i = 789;
+  double d = 67890.09876;
+};
+
+struct C {
+  A arr[3];
+};
+
+consteval C fA() {
+  return {{A{}, A{-45678, 9.8}, A{9}}};
+}
+
+C Array2_Struct = fA();
+//CHECK:      VarDecl {{.*}} Array2_Struct
+//CHECK-NEXT: ConstantExpr {{.*}}
+
+using v4si = int __attribute__((__vector_size__(16)));
+
+struct D {
+  v4si arr[2];
+};
+
+consteval D fv4si() {
+  return {{{1, 2, 3, 4}, {4, 5, 6, 7}}};
+}
+
+D Array_Vector = fv4si();
+//CHECK:      VarDecl {{.*}} Array_Vector
+//CHECK-NEXT: ConstantExpr {{.*}}
+//CHECK-NEXT: value: Struct
+//CHECK-NEXT: field: Array size=2
+//CHECK-NEXT: element: Vector length=4
+//CHECK-NEXT: elements: Int 1, Int 2, Int 3, Int 4
+//CHECK-NEXT: element: Vector length=4
+//CHECK-NEXT: elements: Int 4, Int 5, Int 6, Int 7
+
+} // namespace Array
+
+namespace Union {
+
+struct A {
+  int i = 6789;
+  float f = 987.9876;
+};
+
+union U {
+  int i;
+  A a{567890, 9876.5678f};
+};
+
+consteval U fU1() {
+  return U{0};
+}
+
+U Unique_Union1 = fU1();
+//CHECK:      VarDecl {{.*}} Unique_Union
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Union .i Int 0
+
+consteval U fU() {
+  return U{};
+}
+
+U Unique_Union2 = fU();
+//CHECK:      VarDecl {{.*}} Unique_Union
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Union .a
+//CHECK-NEXT: Struct
+//CHECK-NEXT: fields: Int 567890, Float 9.876567e+03
+
+} // namespace Union
+
+namespace MemberPointer {
+
+struct A {
+  struct B {
+    struct C {
+      struct D {
+        struct E {
+          struct F {
+            struct G {
+              int i;
+            };
+          };
+        };
+      };
+    };
+  };
+};
+
+consteval auto fmem_ptr() -> decltype(&A::B::C::D::E::F::G::i) {
+  return &A::B::C::D::E::F::G::i;
+}
+
+auto MemberPointer1 = fmem_ptr();
+//CHECK:      VarDecl {{.*}} MemberPointer1
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: MemberPointer &G::i
+
+struct A1 {
+  struct B1 {
+    int f() const {
+      return 0;
+    }
+  };
+};
+
+consteval auto fmem_ptr2() {
+  return &A1::B1::f;
+}
+
+auto MemberPointer2 = fmem_ptr2();
+//CHECK:      VarDecl {{.*}} MemberPointer2
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: MemberPointer &B1::f
+
+} // namespace MemberPointer
+
+namespace std {
+struct type_info;
+};
+
+namespace LValue {
+
+constexpr int g = 0;
+
+consteval const int &fg_ref() {
+  return g;
+}
+
+const int &g_ref = fg_ref();
+//CHECK:      VarDecl {{.*}} g_ref
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &g
+
+consteval const int *fint_ptr() {
+  return &g;
+}
+
+const int *g_ptr = fint_ptr();
+//CHECK:      VarDecl {{.*}} g_ptr
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &g
+
+consteval const int *fnull_ptr() {
+  return nullptr;
+}
+
+const int *ptr2 = fnull_ptr();
+//CHECK:      VarDecl {{.*}} ptr2
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue nullptr
+
+int fconst();
+
+consteval auto ffunc_ptr() {
+  return &fconst;
+}
+
+int (*func_ptr)() = ffunc_ptr();
+//CHECK:      VarDecl {{.*}} func_ptr
+//CHECK-NEXT: ConstantExpr {{.*}}
+//CHECK-NEXT: value: LValue &fconst
+
+struct A {
+  int Arr[6] = {0, 1, 3, 4, 5, 9};
+  int i = 0;
+};
+
+struct D {
+  A arr[6] = {};
+};
+
+consteval D fA() {
+  return {};
+}
+
+constexpr D Arr = fA();
+// CHECK:      VarDecl {{.*}} Arr
+// CHECK-NEXT: value: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: ImplicitCastExpr
+// CHECK-NEXT: ConstantExpr
+// CHECK-NEXT: value: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+// CHECK-NEXT: element: Struct
+// CHECK-NEXT: field: Array size=6
+// CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+// CHECK-NEXT: elements: Int 5, Int 9
+// CHECK-NEXT: field: Int 0
+
+consteval const int &fconstintref() {
+  return Arr.arr[0].i;
+}
+
+const int &ArrayStructRef1 = fconstintref();
+//CHECK:      VarDecl {{.*}} ArrayStructRef1
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &Arr.arr[0].i
+
+consteval const int &fconstintref2() {
+  return Arr.arr[1].Arr[5];
+}
+
+const int &ArrayStructRef2 = fconstintref2();
+//CHECK:      VarDecl {{.*}} ArrayStructRef2
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &Arr.arr[1].Arr[5]
+
+consteval const int *fconststar() {
+  return &ArrayStructRef2;
+}
+
+const int *ArrayStructRef3 = fconststar();
+//CHECK:      VarDecl {{.*}} ArrayStructRef3
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue  &Arr.arr[1].Arr[5]
+
+struct B : A {
+};
+
+struct C {
+  B b;
+};
+
+consteval C fC() {
+  return {};
+}
+
+C c = fC();
+//CHECK:      VarDecl {{.*}} c
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: Struct
+//CHECK-NEXT: field: Struct
+//CHECK-NEXT: base: Struct
+//CHECK-NEXT: field: Array size=6
+//CHECK-NEXT: elements: Int 0, Int 1, Int 3, Int 4
+//CHECK-NEXT: elements: Int 5, Int 9
+//CHECK-NEXT: field: Int 0
+
+consteval const int &f2constintref() {
+  return c.b.i;
+}
+
+const int &StructPathRef = f2constintref();
+//CHECK:      VarDecl {{.*}} StructPathRef
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &c.b.A::i
+
+consteval const std::type_info *ftype_info() {
+  return &typeid(c);
+}
+
+const std::type_info *T1 = ftype_info();
+//CHECK:      VarDecl {{.*}} T1
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT:value: LValue &typeid(LValue::C)
+
+consteval const std::type_info *ftype_info2() {
+  return &typeid(Arr.arr[1].Arr[2]);
+}
+
+const std::type_info *T2 = ftype_info2();
+//CHECK:      VarDecl {{.*}} T2
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &typeid(int)
+
+consteval const char *fstring() {
+  return "test";
+}
+
+const char *cptr = fstring();
+//CHECK:      VarDecl {{.*}} cptr
+//CHECK-NEXT: ConstantExpr
+//CHECK-NEXT: value: LValue &"test"[0]
+
+} // namespace LValue
+
+#endif
diff --git a/clang/test/CXX/dcl.dcl/dcl.spec/dcl.type/dcl.type.class.deduct/p2.cpp b/clang/test/CXX/dcl.dcl/dcl.spec/dcl.type/dcl.type.class.deduct/p2.cpp
new file mode 100644
index 0000000000000..8468731a3cb03
--- /dev/null
+++ b/clang/test/CXX/dcl.dcl/dcl.spec/dcl.type/dcl.type.class.deduct/p2.cpp
@@ -0,0 +1,23 @@
+// RUN: %clang_cc1 -std=c++17 -verify=expected,cxx17 %s
+// RUN: %clang_cc1 -std=c++20 -verify=expected,cxx20 %s
+
+template<typename T> struct A { constexpr A(T) {} }; // expected-note 1+{{here}}
+
+A a = 0;
+A b(0);
+A c = A(0);
+A d = A{0};
+auto *p = new A(0);
+A *q = new A(0); // expected-error {{cannot form pointer to deduced class template specialization type}}
+
+struct B {
+  operator A() { // expected-error {{argument deduction not allowed in conversion function type}}
+    return A(0);
+  }
+};
+
+void f(A a); // expected-error {{argument deduction not allowed in function prototype}}
+A f(); // expected-error {{argument deduction not allowed in function return type}}
+
+template<A a> // cxx17-error {{argument deduction not allowed in template parameter}}
+void f();
diff --git a/clang/test/CXX/dcl.dcl/dcl.spec/dcl.type/dcl.type.decltype/p1.cpp b/clang/test/CXX/dcl.dcl/dcl.spec/dcl.type/dcl.type.decltype/p1.cpp
new file mode 100644
index 0000000000000..a62b73d9915ea
--- /dev/null
+++ b/clang/test/CXX/dcl.dcl/dcl.spec/dcl.type/dcl.type.decltype/p1.cpp
@@ -0,0 +1,29 @@
+// RUN: %clang_cc1 -std=c++20 -verify %s
+// expected-no-diagnostics
+
+// bullet 2: decltype(x) where x is a non-type template parameter gives the
+// type of X, after deduction, if any.
+namespace ClassNTTP {
+  template<decltype(auto) v, typename ParamT, typename ExprT> void f() {
+    using U = decltype(v);
+    using U = ParamT;
+
+    using V = decltype((v));
+    using V = ExprT;
+  }
+
+  // The names of most non-reference NTTPs are prvalues.
+  template void f<0, int, int>();
+
+  // The name of a class NTTP of type T is an lvalue of type 'const T'.
+  struct X {};
+  template void f<X{}, X, const X&>();
+
+  // Ensure we get this right for references to classes too.
+  template<auto x> auto &TempParamObject = x;
+  template void f<TempParamObject<X{}>, const X&, const X&>();
+
+  struct Y {} y;
+  template void f<(y), Y&, Y&>();
+  template void f<y, Y, const Y&>();
+}
diff --git a/clang/test/CXX/drs/dr3xx.cpp b/clang/test/CXX/drs/dr3xx.cpp
index be3169ceab761..cec9acf63a9e4 100644
--- a/clang/test/CXX/drs/dr3xx.cpp
+++ b/clang/test/CXX/drs/dr3xx.cpp
@@ -906,8 +906,8 @@ namespace dr367 { // dr367: yes
 namespace dr368 { // dr368: yes
   template<typename T, T> struct S {}; // expected-note {{here}}
   template<typename T> int f(S<T, T()> *); // expected-error {{function type}}
-  //template<typename T> int g(S<T, (T())> *); // FIXME: crashes clang
-  template<typename T> int g(S<T, true ? T() : T()> *); // expected-note {{cannot have type 'dr368::X'}}
+  template<typename T> int g(S<T, (T())> *); // expected-note {{type 'dr368::X'}}
+  template<typename T> int g(S<T, true ? T() : T()> *); // expected-note {{type 'dr368::X'}}
   struct X {};
   int n = g<X>(0); // expected-error {{no matching}}
 }
diff --git a/clang/test/CXX/temp/temp.arg/temp.arg.nontype/p1-cxx20.cpp b/clang/test/CXX/temp/temp.arg/temp.arg.nontype/p1-cxx20.cpp
new file mode 100644
index 0000000000000..d6d2ad742d0ec
--- /dev/null
+++ b/clang/test/CXX/temp/temp.arg/temp.arg.nontype/p1-cxx20.cpp
@@ -0,0 +1,23 @@
+// RUN: %clang_cc1 -std=c++20 %s -verify
+
+using size_t = __SIZE_TYPE__;
+
+namespace CTAD {
+  template<typename T> struct A { constexpr A(T) {} };
+  template<A a> using DeducedA = decltype(a);
+
+  using ATest1 = DeducedA<A(0)>;
+  using ATest1 = A<int>; // expected-note {{previous}}
+  using ATest1 = void; // expected-error {{different}}
+
+  using ATest2 = DeducedA<A(0.0)>;
+  using ATest2 = A<double>;
+
+  template <size_t N> struct B {
+    constexpr B(const char (&r)[N]) { __builtin_memcpy(text, r, N); }
+    char text[N];
+  };
+
+  template<B b> constexpr const char *str() { return b.text; }
+  static_assert(__builtin_strcmp("hello world", str<"hello world">()) == 0);
+}
diff --git a/clang/test/CXX/temp/temp.param/p8-cxx20.cpp b/clang/test/CXX/temp/temp.param/p8-cxx20.cpp
new file mode 100644
index 0000000000000..7a31846756c31
--- /dev/null
+++ b/clang/test/CXX/temp/temp.param/p8-cxx20.cpp
@@ -0,0 +1,28 @@
+// RUN: %clang_cc1 -std=c++20 -verify %s
+
+struct A { int n; };
+
+template<A a> struct B {
+  static constexpr A &v = a; // expected-error {{binding reference of type 'A' to value of type 'const A' drops 'const' qualifier}}
+};
+
+template<A a> struct C {
+  static constexpr const A &v = a;
+};
+
+// All such template parameters in the program of the same type with the same
+// value denote the same template parameter object.
+template<A a, typename T> void check() {
+  static_assert(&a == &T::v); // expected-error {{failed}}
+}
+
+using T = C<A{1}>;
+template void check<A{1}, T>();
+template void check<A{2}, T>(); // expected-note {{instantiation of}}
+
+// Different types with the same value are unequal.
+struct A2 { int n; };
+template<A2 a2> struct C2 {
+  static constexpr const A2 &v = a2;
+};
+static_assert((void*)&C<A{}>::v != (void*)&C2<A2{}>::v);
diff --git a/clang/test/CodeGen/X86/x86-uintr-builtins.c b/clang/test/CodeGen/X86/x86-uintr-builtins.c
new file mode 100644
index 0000000000000..ea937995a1e92
--- /dev/null
+++ b/clang/test/CodeGen/X86/x86-uintr-builtins.c
@@ -0,0 +1,31 @@
+// RUN: %clang_cc1 %s -ffreestanding -triple x86_64-unknown-unknown -target-feature +uintr -emit-llvm -o - | FileCheck %s
+
+#include <x86gprintrin.h>
+
+void test_clui() {
+// CHECK-LABEL: @test_clui
+// CHECK: call void @llvm.x86.clui()
+// CHECK: ret
+  _clui();
+}
+
+void test_stui() {
+// CHECK-LABEL: @test_stui
+// CHECK: call void @llvm.x86.stui()
+// CHECK: ret
+  _stui();
+}
+
+unsigned char test_testui() {
+// CHECK-LABEL: @test_testui
+// CHECK: %[[TMP0:.+]] = call i8 @llvm.x86.testui()
+// CHECK: ret i8 %[[TMP0]]
+  return _testui();
+}
+
+void test_senduipi(unsigned long long a) {
+// CHECK-LABEL: @test_senduipi
+// CHECK: call void @llvm.x86.senduipi(i64 %{{.+}})
+// CHECK: ret
+  _senduipi(a);
+}
diff --git a/clang/test/CodeGen/aarch64-inline-asm.c b/clang/test/CodeGen/aarch64-inline-asm.c
index 0889a7157f0b1..a6e8faef8b9e2 100644
--- a/clang/test/CodeGen/aarch64-inline-asm.c
+++ b/clang/test/CodeGen/aarch64-inline-asm.c
@@ -74,3 +74,9 @@ void test_gcc_registers(void) {
     asm volatile("mov r0, r1\n");
     // CHECK: call void asm sideeffect "mov r0, r1\0A", ""()
 }
+
+void test_tied_earlyclobber(void) {
+  register int a asm("x1");
+  asm("" : "+&r"(a));
+  // CHECK: call i32 asm "", "=&{x1},0"(i32 %0)
+}
diff --git a/clang/test/CodeGen/basic-block-sections.c b/clang/test/CodeGen/basic-block-sections.c
index 3d68394bb78eb..0381bc78c6a96 100644
--- a/clang/test/CodeGen/basic-block-sections.c
+++ b/clang/test/CodeGen/basic-block-sections.c
@@ -6,7 +6,7 @@
 // RUN: %clang_cc1 -triple x86_64 -S -fbasic-block-sections=all -o - < %s | FileCheck %s --check-prefix=BB_WORLD --check-prefix=BB_ALL
 // RUN: %clang_cc1 -triple x86_64 -S -fbasic-block-sections=list=%S/Inputs/basic-block-sections.funcnames -o - < %s | FileCheck %s --check-prefix=BB_WORLD --check-prefix=BB_LIST
 // RUN: %clang_cc1 -triple x86_64 -S -fbasic-block-sections=all -funique-basic-block-section-names -o - < %s | FileCheck %s --check-prefix=UNIQUE
-// RUN: not %clang_cc1 -fbasic-block-sections=list= -emit-obj %s 2>&1 | FileCheck %s --check-prefix=ERROR
+// RUN: not %clang_cc1 -fbasic-block-sections=list= -emit-llvm -o - %s 2>&1 | FileCheck %s --check-prefix=ERROR
 
 int world(int a) {
   if (a > 10)
diff --git a/clang/test/CodeGen/systemz-inline-asm-02.c b/clang/test/CodeGen/systemz-inline-asm-02.c
new file mode 100644
index 0000000000000..754d7e66f04b2
--- /dev/null
+++ b/clang/test/CodeGen/systemz-inline-asm-02.c
@@ -0,0 +1,13 @@
+// RUN: not %clang_cc1 -triple s390x-linux-gnu -O2 -emit-llvm -o - %s 2>&1 \
+// RUN:  | FileCheck %s
+// REQUIRES: systemz-registered-target
+
+// Test that an error is given if a physreg is defined by multiple operands.
+int test_physreg_defs(void) {
+  register int l __asm__("r7") = 0;
+
+  // CHECK: error: multiple outputs to hard register: r7
+  __asm__("" : "+r"(l), "=r"(l));
+
+  return l;
+}
diff --git a/clang/test/CodeGen/systemz-inline-asm.c b/clang/test/CodeGen/systemz-inline-asm.c
index c5497655f505f..357fd4c184e70 100644
--- a/clang/test/CodeGen/systemz-inline-asm.c
+++ b/clang/test/CodeGen/systemz-inline-asm.c
@@ -129,3 +129,17 @@ long double test_f128(long double f, long double g) {
 // CHECK: [[RESULT:%.*]] = tail call fp128 asm "axbr $0, $2", "=f,0,f"(fp128 %f, fp128 %g)
 // CHECK: store fp128 [[RESULT]], fp128* [[DEST]]
 }
+
+// Test that there are no tied physreg uses. TwoAddress pass cannot deal with them.
+int test_physregs(void) {
+  // CHECK-LABEL: define signext i32 @test_physregs()
+  register int l __asm__("r7") = 0;
+
+  // CHECK: call i32 asm "lr $0, $1", "={r7},{r7}"
+  __asm__("lr %0, %1" : "+r"(l));
+
+  // CHECK: call i32 asm "$0 $1 $2", "={r7},{r7},{r7}"
+  __asm__("%0 %1 %2" : "+r"(l) : "r"(l));
+
+  return l;
+}
diff --git a/clang/test/CodeGen/thinlto-distributed-newpm.ll b/clang/test/CodeGen/thinlto-distributed-newpm.ll
index de758e3758b69..3056bf45877ac 100644
--- a/clang/test/CodeGen/thinlto-distributed-newpm.ll
+++ b/clang/test/CodeGen/thinlto-distributed-newpm.ll
@@ -129,13 +129,13 @@
 ; CHECK-O: Running pass: InstCombinePass on main
 ; CHECK-O: Running pass: JumpThreadingPass on main
 ; CHECK-O: Running pass: CorrelatedValuePropagationPass on main
-; CHECK-O: Running pass: DSEPass on main
+; CHECK-O: Running pass: ADCEPass on main
 ; CHECK-O: Running analysis: PostDominatorTreeAnalysis on main
+; CHECK-O: Running pass: DSEPass on main
 ; CHECK-O: Starting {{.*}}Function pass manager run.
 ; CHECK-O: Running pass: LoopSimplifyPass on main
 ; CHECK-O: Running pass: LCSSAPass on main
 ; CHECK-O: Finished {{.*}}Function pass manager run.
-; CHECK-O: Running pass: ADCEPass on main
 ; CHECK-O: Running pass: SimplifyCFGPass on main
 ; CHECK-O: Running pass: InstCombinePass on main
 ; CHECK-O: Finished {{.*}}Function pass manager run.
diff --git a/clang/test/Driver/aarch64-mac-cpus.c b/clang/test/Driver/aarch64-mac-cpus.c
new file mode 100644
index 0000000000000..437c383765453
--- /dev/null
+++ b/clang/test/Driver/aarch64-mac-cpus.c
@@ -0,0 +1,20 @@
+// arm64 Mac-based targets default to Apple A12.
+
+// RUN: %clang -target arm64-apple-macos             -### -c %s 2>&1 | FileCheck %s
+// RUN: %clang -target arm64-apple-ios-macabi        -### -c %s 2>&1 | FileCheck %s
+// RUN: %clang -target arm64-apple-ios-simulator     -### -c %s 2>&1 | FileCheck %s
+// RUN: %clang -target arm64-apple-watchos-simulator -### -c %s 2>&1 | FileCheck %s
+// RUN: %clang -target arm64-apple-tvos-simulator    -### -c %s 2>&1 | FileCheck %s
+
+// RUN: %clang -target arm64-apple-macos -arch arm64 -### -c %s 2>&1 | FileCheck %s
+
+// RUN: %clang -target arm64-apple-macos -mcpu=apple-a11 -### -c %s 2>&1 | FileCheck --check-prefix=EXPLICIT-A11 %s
+// RUN: %clang -target arm64-apple-macos -mcpu=apple-a7  -### -c %s 2>&1 | FileCheck --check-prefix=EXPLICIT-A7 %s
+// RUN: %clang -target arm64-apple-macos -mcpu=apple-a13 -### -c %s 2>&1 | FileCheck --check-prefix=EXPLICIT-A13 %s
+
+// CHECK: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "apple-a12"
+// CHECK-SAME: "-target-feature" "+v8.3a"
+
+// EXPLICIT-A11: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "apple-a11"
+// EXPLICIT-A7: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "apple-a7"
+// EXPLICIT-A13: "-cc1"{{.*}} "-triple" "arm64{{.*}}" "-target-cpu" "apple-a13"
diff --git a/clang/test/Driver/arm-float-abi-lto.c b/clang/test/Driver/arm-float-abi-lto.c
new file mode 100644
index 0000000000000..83c2435d97a4d
--- /dev/null
+++ b/clang/test/Driver/arm-float-abi-lto.c
@@ -0,0 +1,63 @@
+// REQUIRES: arm-registered-target
+
+// RUN: %clang --target=arm-none-eabi -mcpu=cortex-m33 -mfloat-abi=hard -O1 %s -S -o - -emit-llvm -DCALL_LIB -DDEFINE_LIB | FileCheck %s
+
+// RUN: %clang --target=arm-none-eabi -mcpu=cortex-m33 -mfloat-abi=hard -O1 %s -flto=full -c -o %t.call_full.bc -DCALL_LIB
+// RUN: %clang --target=arm-none-eabi -mcpu=cortex-m33 -mfloat-abi=hard -O1 %s -flto=full -c -o %t.define_full.bc -DDEFINE_LIB
+// RUN: llvm-lto2 run -o %t.lto_full -save-temps %t.call_full.bc %t.define_full.bc \
+// RUN:  -r %t.call_full.bc,fn,px \
+// RUN:  -r %t.call_full.bc,fwrite,l \
+// RUN:  -r %t.call_full.bc,putchar,l \
+// RUN:  -r %t.call_full.bc,stdout,px \
+// RUN:  -r %t.define_full.bc,fwrite,px \
+// RUN:  -r %t.define_full.bc,putchar,px \
+// RUN:  -r %t.define_full.bc,otherfn,px
+// RUN: llvm-dis %t.lto_full.0.4.opt.bc -o - | FileCheck %s
+
+// RUN: %clang --target=arm-none-eabi -mcpu=cortex-m33 -mfloat-abi=hard -O1 %s -flto=thin -c -o %t.call_thin.bc -DCALL_LIB
+// RUN: %clang --target=arm-none-eabi -mcpu=cortex-m33 -mfloat-abi=hard -O1 %s -flto=thin -c -o %t.define_thin.bc -DDEFINE_LIB
+// RUN: llvm-lto2 run -o %t.lto_thin -save-temps %t.call_thin.bc %t.define_thin.bc \
+// RUN:  -r %t.call_thin.bc,fn,px \
+// RUN:  -r %t.call_thin.bc,fwrite,l \
+// RUN:  -r %t.call_thin.bc,putchar,l \
+// RUN:  -r %t.call_thin.bc,stdout,px \
+// RUN:  -r %t.define_thin.bc,fwrite,px \
+// RUN:  -r %t.define_thin.bc,putchar,px \
+// RUN:  -r %t.define_thin.bc,otherfn,px
+// RUN: llvm-dis %t.lto_thin.1.4.opt.bc -o - | FileCheck %s
+
+// We expect that the fprintf is optimised to fwrite, and the printf is
+// optimised to putchar. Check that we don't have a mismatch in calling
+// conventions causing the call to be replaced by a trap.
+// CHECK-LABEL: define{{.*}}void @fn()
+// CHECK-NOT: call void @llvm.trap()
+
+typedef struct FILE FILE;
+typedef unsigned int size_t;
+extern FILE *stdout;
+extern int fprintf(FILE *, const char *, ...);
+extern int printf(const char *, ...);
+extern void otherfn(const void *);
+
+#ifdef CALL_LIB
+
+void fn() {
+  fprintf(stdout, "hello world");
+  printf("a");
+}
+
+#endif
+
+#ifdef DEFINE_LIB
+
+size_t fwrite(const void *ptr, size_t size, size_t nmemb, FILE *stream) {
+  otherfn(ptr);
+  return 0;
+}
+
+int putchar(int c) {
+  otherfn(&c);
+  return 0;
+}
+
+#endif
diff --git a/clang/test/Driver/arm-float-abi.c b/clang/test/Driver/arm-float-abi.c
index 9a76d1ee39c45..806ef0fbd93e6 100644
--- a/clang/test/Driver/arm-float-abi.c
+++ b/clang/test/Driver/arm-float-abi.c
@@ -1,8 +1,9 @@
+// REQUIRES: arm-registered-target
 // RUN: not %clang %s -target armv7-apple-ios -mfloat-abi=hard 2>&1 | FileCheck -check-prefix=ARMV7-ERROR %s
 // RUN: %clang %s -target armv7-apple-ios -mfloat-abi=softfp -### 2>&1 | FileCheck -check-prefix=NOERROR %s
 // RUN: %clang %s -arch armv7 -target thumbv7-apple-darwin-eabi -mfloat-abi=hard -### 2>&1 | FileCheck -check-prefix=NOERROR %s
 
-// ARMV7-ERROR: unsupported option '-mfloat-abi=hard' for target 'thumbv7'
+// ARMV7-ERROR: unsupported option '-mfloat-abi=hard' for target 'thumbv7-apple-ios'
 // NOERROR-NOT: unsupported option
 
 // RUN: %clang -target armv7-linux-androideabi21 %s -### -c 2>&1 \
@@ -14,3 +15,25 @@
 // RUN:   | FileCheck --check-prefix=CHECK-ARM8-ANDROID %s
 // CHECK-ARM8-ANDROID-NOT: "-target-feature" "+soft-float"
 // CHECK-ARM8-ANDROID: "-target-feature" "+soft-float-abi"
+
+// RUN: not %clang -target armv7-linux-androideabi21 %s -S -o - -mfloat-abi=hard 2>&1 \
+// RUN:   | FileCheck --check-prefix=CHECK-ANDROID-ERROR %s
+// CHECK-ANDROID-ERROR: unsupported option '-mfloat-abi=hard' for target 'armv7-unknown-linux-android21'
+
+// RUN: %clang -target armv7-linux-androideabi21 %s -S -o - -mfloat-abi=soft -### 2>&1 \
+// RUN:   | FileCheck --check-prefix=CHECK-ANDROID-NOERROR %s
+// RUN: %clang -target armv7-linux-androideabi21 %s -S -o - -mfloat-abi=softfp -### 2>&1 \
+// RUN:   | FileCheck --check-prefix=CHECK-ANDROID-NOERROR %s
+// CHECK-ANDROID-NOERROR-NOT: unsupported option
+
+// RUN: not %clang -target armv7-apple-watchos4 %s -S -o - -mfloat-abi=soft 2>&1 \
+// RUN:   | FileCheck --check-prefix=CHECK-WATCHOS-ERROR1 %s
+// CHECK-WATCHOS-ERROR1: unsupported option '-mfloat-abi=soft' for target 'thumbv7-apple-watchos4'
+
+// RUN: not %clang -target armv7-apple-watchos4 %s -S -o - -mfloat-abi=softfp 2>&1 \
+// RUN:   | FileCheck --check-prefix=CHECK-WATCHOS-ERROR2 %s
+// CHECK-WATCHOS-ERROR2: unsupported option '-mfloat-abi=softfp' for target 'thumbv7-apple-watchos4'
+
+// RUN: %clang -target armv7-apple-watchos4 %s -S -o - -mfloat-abi=hard -### 2>&1 \
+// RUN:   | FileCheck --check-prefix=CHECK-WATCHOS-NOERROR %s
+// CHECK-WATCHOS-NOERROR-NOT: unsupported option
diff --git a/clang/test/Driver/arm-triple.c b/clang/test/Driver/arm-triple.c
new file mode 100644
index 0000000000000..fa9f7b189c827
--- /dev/null
+++ b/clang/test/Driver/arm-triple.c
@@ -0,0 +1,48 @@
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabi \
+// RUN:   | FileCheck %s --check-prefix=CHECK-DEFAULT
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=armeb-none-eabi -mlittle-endian \
+// RUN:   | FileCheck %s --check-prefix=CHECK-DEFAULT
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabihf -march=armv4t -mfloat-abi=softfp \
+// RUN:   | FileCheck %s --check-prefix=CHECK-DEFAULT
+// CHECK-DEFAULT: armv4t-none-unknown-eabi
+
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=armeb-none-eabi \
+// RUN:   | FileCheck %s --check-prefix=CHECK-EB
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabi -mbig-endian \
+// RUN:   | FileCheck %s --check-prefix=CHECK-EB
+// CHECK-EB: armebv4t-none-unknown-eabi
+
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabihf -march=armv4t \
+// RUN:   | FileCheck %s --check-prefix=CHECK-HF
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabi -mfloat-abi=hard \
+// RUN:   | FileCheck %s --check-prefix=CHECK-HF
+// CHECK-HF: armv4t-none-unknown-eabihf
+
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=armeb-none-eabihf -march=armv4t \
+// RUN:   | FileCheck %s --check-prefix=CHECK-EB-HF
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=armeb-none-eabi -mfloat-abi=hard \
+// RUN:   | FileCheck %s --check-prefix=CHECK-EB-HF
+// RUN: %clang -print-effective-triple -march=armv4t \
+// RUN:   --target=arm-none-eabihf -mbig-endian \
+// RUN:   | FileCheck %s --check-prefix=CHECK-EB-HF
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabi -mbig-endian -mfloat-abi=hard \
+// RUN:   | FileCheck %s --check-prefix=CHECK-EB-HF
+// CHECK-EB-HF: armebv4t-none-unknown-eabihf
+
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabi -march=armv8m.main -mbig-endian -mfloat-abi=hard \
+// RUN:   | FileCheck %s --check-prefix=CHECK-V8M-EB-HF
+// RUN: %clang -print-effective-triple \
+// RUN:   --target=arm-none-eabi -mcpu=cortex-m33 -mbig-endian -mfloat-abi=hard \
+// RUN:   | FileCheck %s --check-prefix=CHECK-V8M-EB-HF
+// CHECK-V8M-EB-HF: thumbebv8m.main-none-unknown-eabihf
diff --git a/clang/test/Driver/stack-protector-guard.c b/clang/test/Driver/stack-protector-guard.c
new file mode 100644
index 0000000000000..f2bb8cd8555d6
--- /dev/null
+++ b/clang/test/Driver/stack-protector-guard.c
@@ -0,0 +1,43 @@
+// RUN: %clang -### -target x86_64-unknown-unknown -mstack-protector-guard=tls %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=CHECK-TLS %s
+// RUN: %clang -### -target x86_64-unknown-unknown -mstack-protector-guard=global %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=CHECK-GLOBAL %s
+// RUN: %clang -### -target x86_64-unknown-unknown -mstack-protector-guard=local %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=INVALID-VALUE %s
+
+// CHECK-TLS: "-cc1" {{.*}}"-mstack-protector-guard=tls"
+// CHECK-GLOBAL: "-cc1" {{.*}}"-mstack-protector-guard=global"
+// INVALID-VALUE: error: invalid value 'local' in 'mstack-protector-guard=','valid arguments to '-mstack-protector-guard=' are:tls global'
+
+// RUN: %clang -### -target x86_64-unknown-unknown -mstack-protector-guard-reg=fs %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=CHECK-FS %s
+// RUN: %clang -### -target x86_64-unknown-unknown -mstack-protector-guard-reg=gs %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=CHECK-GS %s
+
+// Invalid arch
+// RUN: not %clang -target arm-eabi-c -mstack-protector-guard=tls %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=INVALID-ARCH %s
+// INVALID-ARCH: unsupported option '-mstack-protector-guard=tls' for target
+
+// RUN: not %clang -target powerpc64le-linux-gnu -mstack-protector-guard-reg=fs %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=INVALID-ARCH2 %s
+// INVALID-ARCH2: unsupported option '-mstack-protector-guard-reg=fs' for target
+
+// RUN: not %clang -target aarch64-linux-gnu -mstack-protector-guard-offset=10 %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=INVALID-ARCH3 %s
+// INVALID-ARCH3: unsupported option '-mstack-protector-guard-offset=10' for target
+
+// Invalid option value
+// RUN: not %clang -target x86_64-unknown-unknown -c -mstack-protector-guard-reg=cs %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=INVALID-REG %s
+// RUN: not %clang -target x86_64-unknown-unknown -c -mstack-protector-guard-reg=ds %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=INVALID-REG %s
+
+// CHECK-FS: "-cc1" {{.*}}"-mstack-protector-guard-reg=fs"
+// CHECK-GS: "-cc1" {{.*}}"-mstack-protector-guard-reg=gs"
+// INVALID-REG: error: invalid value {{.*}} in 'mstack-protector-guard-reg=','for X86, valid arguments to '-mstack-protector-guard-reg=' are:fs gs'
+
+// RUN: %clang -### -target x86_64-unknown-unknown -mstack-protector-guard-offset=30 %s 2>&1 | \
+// RUN:   FileCheck -check-prefix=CHECK-OFFSET %s
+
+// CHECK-OFFSET: "-cc1" {{.*}}"-mstack-protector-guard-offset=30"
diff --git a/clang/test/Driver/windows-thumbv7em.cpp b/clang/test/Driver/windows-thumbv7em.cpp
index 5d7c00b31fd16..94d18e4a9b83b 100644
--- a/clang/test/Driver/windows-thumbv7em.cpp
+++ b/clang/test/Driver/windows-thumbv7em.cpp
@@ -1,8 +1,8 @@
 // RUN: %clang -target thumb-none-windows-eabi-coff -mcpu=cortex-m7 -### -c %s 2>&1 \
 // RUN: | FileCheck %s --check-prefix CHECK-V7
-// CHECK-V7-NOT: error: the target architecture 'thumbv7em' is not supported by the target 'thumbv7em-none-windows-eabi'
+// CHECK-V7-NOT: error: the target architecture 'thumbv7em' is not supported by the target 'thumbv7em-none-windows-eabihf'
 
 // RUN: %clang -target thumb-none-windows-eabi-coff -mcpu=cortex-m1 -### -c %s 2>&1 \
 // RUN: | FileCheck %s --check-prefix CHECK-V6
-// CHECK-V6: error: the target architecture 'thumbv6m' is not supported by the target 'thumbv6m-none-windows-eabi'
+// CHECK-V6: error: the target architecture 'thumbv6m' is not supported by the target 'thumbv6m-none-windows-eabihf'
 
diff --git a/clang/test/Driver/x86-target-features.c b/clang/test/Driver/x86-target-features.c
index 3cd4748c942fb..cd79129a887eb 100644
--- a/clang/test/Driver/x86-target-features.c
+++ b/clang/test/Driver/x86-target-features.c
@@ -283,3 +283,8 @@
 // RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-hreset %s -### -o %t.o 2>&1 | FileCheck -check-prefix=NO-HRESET %s
 // HRESET: "-target-feature" "+hreset"
 // NO-HRESET: "-target-feature" "-hreset"
+
+// RUN: %clang -target i386-unknown-linux-gnu -march=i386 -muintr %s -### -o %t.o 2>&1 | FileCheck -check-prefix=UINTR %s
+// RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-uintr %s -### -o %t.o 2>&1 | FileCheck -check-prefix=NO-UINTR %s
+// UINTR: "-target-feature" "+uintr"
+// NO-UINTR: "-target-feature" "-uintr"
diff --git a/clang/test/Misc/amdgcn.languageOptsOpenCL.cl b/clang/test/Misc/amdgcn.languageOptsOpenCL.cl
index 6a07fc98069ca..e7cb8d0513211 100644
--- a/clang/test/Misc/amdgcn.languageOptsOpenCL.cl
+++ b/clang/test/Misc/amdgcn.languageOptsOpenCL.cl
@@ -29,17 +29,6 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_int64_extended_atomics: enable
 
-#ifdef cl_khr_gl_sharing
-#error "Incorrect cl_khr_gl_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_sharing' - ignoring}}
-
-#ifndef cl_khr_icd
-#error "Missing cl_khr_icd define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_icd: enable
-
 // Core features in CL 1.1
 
 #ifndef cl_khr_byte_addressable_store
@@ -82,13 +71,6 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_local_int32_extended_atomics' is core feature or supported optional core feature - ignoring}}
 #endif
 
-#ifdef cl_khr_select_fprounding_mode
-#error "Incorrect cl_khr_select_fprounding_mode define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_select_fprounding_mode: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_select_fprounding_mode' - ignoring}}
-
-
 // Core feature in CL 1.2
 #ifndef cl_khr_fp64
 #error "Missing cl_khr_fp64 define"
@@ -107,80 +89,12 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_3d_image_writes' is core feature or supported optional core feature - ignoring}}
 #endif
 
-
-
-#ifdef cl_khr_gl_event
-#error "Incorrect cl_khr_gl_event define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_event: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_event' - ignoring}}
-
-#ifdef cl_khr_d3d10_sharing
-#error "Incorrect cl_khr_d3d10_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d10_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_d3d10_sharing' - ignoring}}
-
-#ifdef cl_khr_context_abort
-#error "Incorrect cl_context_abort define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_context_abort: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_context_abort' - ignoring}}
-
-#ifdef cl_khr_d3d11_sharing
-#error "Incorrect cl_khr_d3d11_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d11_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_d3d11_sharing' - ignoring}}
-
-#ifdef cl_khr_dx9_media_sharing
-#error "Incorrect cl_khr_dx9_media_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_dx9_media_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_dx9_media_sharing' - ignoring}}
-
-#ifdef cl_khr_image2d_from_buffer
-#error "Incorrect cl_khr_image2d_from_buffer define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_image2d_from_buffer: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_image2d_from_buffer' - ignoring}}
-
-#ifdef cl_khr_initialize_memory
-#error "Incorrect cl_khr_initialize_memory define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_initialize_memory: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_initialize_memory' - ignoring}}
-
-#ifdef cl_khr_gl_depth_images
-#error "Incorrect cl_khr_gl_depth_images define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_depth_images: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_depth_images' - ignoring}}
-
 #ifdef cl_khr_gl_msaa_sharing
 #error "Incorrect cl_khr_gl_msaa_sharing define"
 #endif
 #pragma OPENCL EXTENSION cl_khr_gl_msaa_sharing: enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_msaa_sharing' - ignoring}}
 
-#ifdef cl_khr_spir
-#error "Incorrect cl_khr_spir define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_spir: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_spir' - ignoring}}
-
-#ifdef cl_khr_egl_event
-#error "Incorrect cl_khr_egl_event define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_event: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_egl_event' - ignoring}}
-
-#ifdef cl_khr_egl_image
-#error "Incorrect cl_khr_egl_image define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_image: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_egl_image' - ignoring}}
-
 #if (__OPENCL_C_VERSION__ >= 200)
 #ifndef cl_khr_mipmap_image
 #error "Missing cl_khr_mipmap_image define"
@@ -211,12 +125,6 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_subgroups: enable
 
-#ifdef cl_khr_terminate_context
-#error "Incorrect cl_khr_terminate_context define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_terminate_context: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_terminate_context' - ignoring}}
-
 #ifndef cl_amd_media_ops
 #error "Missing cl_amd_media_ops define"
 #endif
diff --git a/clang/test/Misc/nvptx.languageOptsOpenCL.cl b/clang/test/Misc/nvptx.languageOptsOpenCL.cl
index 62672487b02d5..732b18203d32b 100644
--- a/clang/test/Misc/nvptx.languageOptsOpenCL.cl
+++ b/clang/test/Misc/nvptx.languageOptsOpenCL.cl
@@ -47,16 +47,6 @@
 #pragma OPENCL EXTENSION cl_khr_int64_extended_atomics : enable
 // expectedwarning@-1{{unsupported OpenCL extension 'cl_khr_int64_extended_atomics' - ignoring}}
 
-#ifndef cl_khr_gl_sharing
-#error "Missing cl_khr_gl_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_sharing : enable
-
-#ifndef cl_khr_icd
-#error "Missing cl_khr_icd define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_icd : enable
-
 // Core features in CL 1.1
 
 #ifndef cl_khr_byte_addressable_store
@@ -99,15 +89,6 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_local_int32_extended_atomics' is core feature or supported optional core feature - ignoring}}
 #endif
 
-#if (__OPENCL_C_VERSION__ < 110)
-// Deprecated above 1.0
-#ifdef cl_khr_select_fprounding_mode
-#error "Incorrect cl_khr_select_fprounding_mode define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_select_fprounding_mode : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_select_fprounding_mode' - ignoring}}
-#endif
-
 // Core feature in CL 1.2
 #ifndef cl_khr_fp64
 #error "Missing cl_khr_fp64 define"
@@ -130,78 +111,12 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_3d_image_writes' is core feature or supported optional core feature - ignoring}}
 #endif
 
-#ifdef cl_khr_gl_event
-#error "Incorrect cl_khr_gl_event define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_event : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_event' - ignoring}}
-
-#ifdef cl_khr_d3d10_sharing
-#error "Incorrect cl_khr_d3d10_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d10_sharing : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_d3d10_sharing' - ignoring}}
-
-#ifdef cl_khr_context_abort
-#error "Incorrect cl_context_abort define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_context_abort : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_context_abort' - ignoring}}
-
-#ifdef cl_khr_d3d11_sharing
-#error "Incorrect cl_khr_d3d11_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d11_sharing : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_d3d11_sharing' - ignoring}}
-
-#ifdef cl_khr_dx9_media_sharing
-#error "Incorrect cl_khr_dx9_media_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_dx9_media_sharing : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_dx9_media_sharing' - ignoring}}
-
-#ifdef cl_khr_image2d_from_buffer
-#error "Incorrect cl_khr_image2d_from_buffer define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_image2d_from_buffer : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_image2d_from_buffer' - ignoring}}
-
-#ifdef cl_khr_initialize_memory
-#error "Incorrect cl_khr_initialize_memory define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_initialize_memory : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_initialize_memory' - ignoring}}
-
-#ifdef cl_khr_gl_depth_images
-#error "Incorrect cl_khr_gl_depth_images define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_depth_images : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_depth_images' - ignoring}}
-
 #ifdef cl_khr_gl_msaa_sharing
 #error "Incorrect cl_khr_gl_msaa_sharing define"
 #endif
 #pragma OPENCL EXTENSION cl_khr_gl_msaa_sharing : enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_msaa_sharing' - ignoring}}
 
-#ifdef cl_khr_spir
-#error "Incorrect cl_khr_spir define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_spir : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_spir' - ignoring}}
-
-#ifdef cl_khr_egl_event
-#error "Incorrect cl_khr_egl_event define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_event : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_egl_event' - ignoring}}
-
-#ifdef cl_khr_egl_image
-#error "Missing cl_khr_egl_image define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_image : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_egl_image' - ignoring}}
-
 #ifdef cl_khr_srgb_image_writes
 #error "Incorrect cl_khr_srgb_image_writes define"
 #endif
@@ -213,9 +128,3 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_subgroups : enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_subgroups' - ignoring}}
-
-#ifdef cl_khr_terminate_context
-#error "Incorrect cl_khr_terminate_context define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_terminate_context : enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_terminate_context' - ignoring}}
diff --git a/clang/test/Misc/r600.languageOptsOpenCL.cl b/clang/test/Misc/r600.languageOptsOpenCL.cl
index 58444cf7688a6..e0d3022f3e4a0 100644
--- a/clang/test/Misc/r600.languageOptsOpenCL.cl
+++ b/clang/test/Misc/r600.languageOptsOpenCL.cl
@@ -48,17 +48,6 @@
 #pragma OPENCL EXTENSION cl_khr_int64_extended_atomics: enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_int64_extended_atomics' - ignoring}}
 
-#ifdef cl_khr_gl_sharing
-#error "Incorrect cl_khr_gl_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_sharing' - ignoring}}
-
-#ifndef cl_khr_icd
-#error "Missing cl_khr_icd define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_icd: enable
-
 // Core features in CL 1.1
 
 #ifndef cl_khr_byte_addressable_store
@@ -101,14 +90,6 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_local_int32_extended_atomics' is core feature or supported optional core feature - ignoring}}
 #endif
 
-// Deprecated abvoe 1.0
-#ifdef cl_khr_select_fprounding_mode
-#error "Incorrect cl_khr_select_fprounding_mode define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_select_fprounding_mode: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_select_fprounding_mode' - ignoring}}
-
-
 // Core feature in CL 1.2
 #ifdef __HAS_FP64__
 #ifndef cl_khr_fp64
@@ -133,79 +114,12 @@
 #pragma OPENCL EXTENSION cl_khr_3d_image_writes: enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_3d_image_writes' - ignoring}}
 
-
-#ifdef cl_khr_gl_event
-#error "Incorrect cl_khr_gl_event define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_event: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_event' - ignoring}}
-
-#ifdef cl_khr_d3d10_sharing
-#error "Incorrect cl_khr_d3d10_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d10_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_d3d10_sharing' - ignoring}}
-
-#ifdef cl_khr_context_abort
-#error "Incorrect cl_context_abort define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_context_abort: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_context_abort' - ignoring}}
-
-#ifdef cl_khr_d3d11_sharing
-#error "Incorrect cl_khr_d3d11_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d11_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_d3d11_sharing' - ignoring}}
-
-#ifdef cl_khr_dx9_media_sharing
-#error "Incorrect cl_khr_dx9_media_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_dx9_media_sharing: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_dx9_media_sharing' - ignoring}}
-
-#ifdef cl_khr_image2d_from_buffer
-#error "Incorrect cl_khr_image2d_from_buffer define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_image2d_from_buffer: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_image2d_from_buffer' - ignoring}}
-
-#ifdef cl_khr_initialize_memory
-#error "Incorrect cl_khr_initialize_memory define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_initialize_memory: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_initialize_memory' - ignoring}}
-
-#ifdef cl_khr_gl_depth_images
-#error "Incorrect cl_khr_gl_depth_images define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_depth_images: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_depth_images' - ignoring}}
-
 #ifdef cl_khr_gl_msaa_sharing
 #error "Incorrect cl_khr_gl_msaa_sharing define"
 #endif
 #pragma OPENCL EXTENSION cl_khr_gl_msaa_sharing: enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_gl_msaa_sharing' - ignoring}}
 
-#ifdef cl_khr_spir
-#error "Incorrect cl_khr_spir define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_spir: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_spir' - ignoring}}
-
-#ifdef cl_khr_egl_event
-#error "Incorrect cl_khr_egl_event define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_event: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_egl_event' - ignoring}}
-
-#ifdef cl_khr_egl_image
-#error "Incorrect cl_khr_egl_image define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_image: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_egl_image' - ignoring}}
-
 #ifdef cl_khr_srgb_image_writes
 #error "Incorrect cl_khr_srgb_image_writes define"
 #endif
@@ -218,8 +132,3 @@
 #pragma OPENCL EXTENSION cl_khr_subgroups: enable
 // expected-warning@-1{{unsupported OpenCL extension 'cl_khr_subgroups' - ignoring}}
 
-#ifdef cl_khr_terminate_context
-#error "Incorrect cl_khr_terminate_context define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_terminate_context: enable
-// expected-warning@-1{{unsupported OpenCL extension 'cl_khr_terminate_context' - ignoring}}
diff --git a/clang/test/OpenMP/driver-openmp-target.c b/clang/test/OpenMP/driver-openmp-target.c
index bb58b386c06af..2261cbaa0d561 100644
--- a/clang/test/OpenMP/driver-openmp-target.c
+++ b/clang/test/OpenMP/driver-openmp-target.c
@@ -1,5 +1,5 @@
 // REQUIRES: x86-registered-target
-// REQUIRES: clang-64-bits
+// REQUIRES: clang-target-64-bits
 // RUN: %clang %s -c -E -dM -fopenmp=libomp -fopenmp-version=45 -fopenmp-targets=x86_64-unknown-unknown -o - | FileCheck --check-prefix=CHECK-45-VERSION --check-prefix=CHECK-45-VERSION2 %s
 // CHECK-45-VERSION: #define _OPENMP 201511
 // CHECK-45-VERSION2: #define _OPENMP 201511
diff --git a/clang/test/PCH/cxx20-template-args.cpp b/clang/test/PCH/cxx20-template-args.cpp
new file mode 100644
index 0000000000000..f9ac35ba53a45
--- /dev/null
+++ b/clang/test/PCH/cxx20-template-args.cpp
@@ -0,0 +1,40 @@
+// RUN: %clang_cc1 -std=c++20 -include %s %s -o %t
+
+// RUN: %clang_cc1 -std=c++20 -emit-pch %s -o %t
+// RUN: %clang_cc1 -std=c++20 -include-pch %t -verify %s
+
+// expected-no-diagnostics
+
+#ifndef HEADER
+#define HEADER
+
+int g;
+struct A { union { int n, m; }; int *p; int A::*q; char buffer[32]; };
+
+template<A a> constexpr const A &get = a;
+
+constexpr const A &v = get<A{}>;
+constexpr const A &w = get<A{1, &g, &A::n, "hello"}>;
+
+#else /*included pch*/
+
+template<A a> constexpr const A &get2 = a;
+
+constexpr const A &v2 = get2<A{}>;
+constexpr const A &w2 = get2<A{1, &g, &A::n, "hello\0\0\0\0\0"}>;
+
+static_assert(&v == &v2);
+static_assert(&w == &w2);
+
+static_assert(&v != &w);
+static_assert(&v2 != &w);
+static_assert(&v != &w2);
+static_assert(&v2 != &w2);
+
+constexpr const A &v3 = get2<A{.n = 0}>;
+constexpr const A &x = get2<A{.m = 0}>;
+
+static_assert(&v == &v3);
+static_assert(&v != &x);
+
+#endif // HEADER
diff --git a/clang/test/Preprocessor/aarch64-target-features.c b/clang/test/Preprocessor/aarch64-target-features.c
index ad84ba93ccf3c..0529a3a3d534b 100644
--- a/clang/test/Preprocessor/aarch64-target-features.c
+++ b/clang/test/Preprocessor/aarch64-target-features.c
@@ -247,7 +247,7 @@
 // CHECK-MCPU-CARMEL: "-cc1"{{.*}} "-triple" "aarch64{{.*}}" "-target-feature" "+v8.2a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-target-feature" "+crc" "-target-feature" "+crypto" "-target-feature" "+fullfp16" "-target-feature" "+ras" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+sha2" "-target-feature" "+aes"
 
 // RUN: %clang -target x86_64-apple-macosx -arch arm64 -### -c %s 2>&1 | FileCheck --check-prefix=CHECK-ARCH-ARM64 %s
-// CHECK-ARCH-ARM64: "-target-cpu" "apple-a7" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-target-feature" "+crypto" "-target-feature" "+zcm" "-target-feature" "+zcz"
+// CHECK-ARCH-ARM64: "-target-cpu" "apple-a12" "-target-feature" "+v8.3a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-target-feature" "+crc" "-target-feature" "+crypto" "-target-feature" "+fullfp16" "-target-feature" "+ras" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+rcpc" "-target-feature" "+zcm" "-target-feature" "+zcz"
 
 // RUN: %clang -target x86_64-apple-macosx -arch arm64_32 -### -c %s 2>&1 | FileCheck --check-prefix=CHECK-ARCH-ARM64_32 %s
 // CHECK-ARCH-ARM64_32: "-target-cpu" "apple-s4" "-target-feature" "+v8.3a" "-target-feature" "+fp-armv8" "-target-feature" "+neon" "-target-feature" "+crc" "-target-feature" "+crypto" "-target-feature" "+fullfp16" "-target-feature" "+ras" "-target-feature" "+lse" "-target-feature" "+rdm" "-target-feature" "+rcpc" "-target-feature" "+zcm" "-target-feature" "+zcz" "-target-feature" "+sha2" "-target-feature" "+aes"
diff --git a/clang/test/Preprocessor/predefined-arch-macros.c b/clang/test/Preprocessor/predefined-arch-macros.c
index 820a242955c2d..f8ee187b15f3a 100644
--- a/clang/test/Preprocessor/predefined-arch-macros.c
+++ b/clang/test/Preprocessor/predefined-arch-macros.c
@@ -1688,6 +1688,7 @@
 // CHECK_SPR_M32: #define __SSE__ 1
 // CHECK_SPR_M32: #define __SSSE3__ 1
 // CHECK_SPR_M32: #define __TSXLDTRK__ 1
+// CHECK_SPR_M32: #define __UINTR__ 1
 // CHECK_SPR_M32: #define __VAES__ 1
 // CHECK_SPR_M32: #define __VPCLMULQDQ__ 1
 // CHECK_SPR_M32: #define __WAITPKG__ 1
@@ -1757,6 +1758,7 @@
 // CHECK_SPR_M64: #define __SSE__ 1
 // CHECK_SPR_M64: #define __SSSE3__ 1
 // CHECK_SPR_M64: #define __TSXLDTRK__ 1
+// CHECK_SPR_M64: #define __UINTR__ 1
 // CHECK_SPR_M64: #define __VAES__ 1
 // CHECK_SPR_M64: #define __VPCLMULQDQ__ 1
 // CHECK_SPR_M64: #define __WAITPKG__ 1
diff --git a/clang/test/Preprocessor/x86_target_features.c b/clang/test/Preprocessor/x86_target_features.c
index 1d8e893282da4..69078c2ff2bed 100644
--- a/clang/test/Preprocessor/x86_target_features.c
+++ b/clang/test/Preprocessor/x86_target_features.c
@@ -536,3 +536,11 @@
 // RUN: %clang -target i386-unknown-unknown -march=atom -mno-hreset -x c -E -dM -o - %s | FileCheck -match-full-lines --check-prefix=NOHRESET %s
 
 // NOHRESET-NOT: #define __HRESET__ 1
+
+// RUN: %clang -target i386-unknown-linux-gnu -march=i386 -muintr -x c -E -dM -o - %s | FileCheck -check-prefix=UINTR %s
+
+// UINTR: #define __UINTR__ 1
+
+// RUN: %clang -target i386-unknown-linux-gnu -march=i386 -mno-uintr -x c -E -dM -o - %s | FileCheck -check-prefix=NOUINTR %s
+
+// NOUINTR-NOT: #define __UINTR__ 1
diff --git a/clang/test/SemaCXX/cxx17-compat.cpp b/clang/test/SemaCXX/cxx17-compat.cpp
index b65ed3ea340b7..41a465ac613ef 100644
--- a/clang/test/SemaCXX/cxx17-compat.cpp
+++ b/clang/test/SemaCXX/cxx17-compat.cpp
@@ -121,3 +121,13 @@ struct DefaultedComparisons {
   // expected-warning@-10 {{defaulted comparison operators are incompatible with C++ standards before C++20}}
 #endif
 };
+
+namespace NTTP {
+  struct A {};
+  template<A> struct Class {};
+#if __cplusplus <= 201703L
+  // expected-error@-2 {{non-type template parameter cannot have type 'NTTP::A' before C++20}}
+#else
+  // expected-warning@-4 {{non-type template parameter of type 'NTTP::A' is incompatible with C++ standards before C++20}}
+#endif
+}
diff --git a/clang/test/SemaOpenCL/extension-version.cl b/clang/test/SemaOpenCL/extension-version.cl
index 4d0e79cd39c5d..c4bf936f18ff7 100644
--- a/clang/test/SemaOpenCL/extension-version.cl
+++ b/clang/test/SemaOpenCL/extension-version.cl
@@ -34,16 +34,6 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_int64_extended_atomics: enable
 
-#ifndef cl_khr_gl_sharing
-#error "Missing cl_khr_gl_sharing define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_sharing: enable
-
-#ifndef cl_khr_icd
-#error "Missing cl_khr_icd define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_icd: enable
-
 // Core features in CL 1.1
 
 #ifndef cl_khr_byte_addressable_store
@@ -86,15 +76,6 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_local_int32_extended_atomics' is core feature or supported optional core feature - ignoring}}
 #endif
 
-#if (defined(__OPENCL_C_VERSION__) && __OPENCL_C_VERSION__ < 110)
-// Deprecated abvoe 1.0
-#ifndef cl_khr_select_fprounding_mode
-#error "Missing cl_khr_select_fp_rounding_mode define"
-#endif
-#pragma OPENCL EXTENSION cl_khr_select_fprounding_mode: enable
-#endif
-
-
 // Core feature in CL 1.2
 #ifndef cl_khr_fp64
 #error "Missing cl_khr_fp64 define"
@@ -113,24 +94,6 @@
 // expected-warning@-2{{OpenCL extension 'cl_khr_3d_image_writes' is core feature or supported optional core feature - ignoring}}
 #endif
 
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 110)
-#ifndef cl_khr_gl_event
-#error "Missing cl_khr_gl_event define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_gl_event' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_event : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 110)
-#ifndef cl_khr_d3d10_sharing
-#error "Missing cl_khr_d3d10_sharing define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_d3d10_sharing' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d10_sharing : enable
-
 #if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 110)
 #ifndef cles_khr_int64
 #error "Missing cles_khr_int64 define"
@@ -140,60 +103,6 @@
 #endif
 #pragma OPENCL EXTENSION cles_khr_int64 : enable
 
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_context_abort
-#error "Missing cl_context_abort define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_context_abort' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_context_abort : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_d3d11_sharing
-#error "Missing cl_khr_d3d11_sharing define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_d3d11_sharing' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_d3d11_sharing : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_dx9_media_sharing
-#error "Missing cl_khr_dx9_media_sharing define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_dx9_media_sharing' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_dx9_media_sharing : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_image2d_from_buffer
-#error "Missing cl_khr_image2d_from_buffer define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_image2d_from_buffer' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_image2d_from_buffer : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_initialize_memory
-#error "Missing cl_khr_initialize_memory define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_initialize_memory' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_initialize_memory : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_gl_depth_images
-#error "Missing cl_khr_gl_depth_images define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_gl_depth_images' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_gl_depth_images : enable
-
 #if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
 #ifndef cl_khr_gl_msaa_sharing
 #error "Missing cl_khr_gl_msaa_sharing define"
@@ -203,33 +112,6 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_gl_msaa_sharing : enable
 
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 120)
-#ifndef cl_khr_spir
-#error "Missing cl_khr_spir define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_spir' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_spir : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 200)
-#ifndef cl_khr_egl_event
-#error "Missing cl_khr_egl_event define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_egl_event' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_event : enable
-
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 200)
-#ifndef cl_khr_egl_image
-#error "Missing cl_khr_egl_image define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_egl_image' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_egl_image : enable
-
 #if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 200)
 #ifndef cl_khr_mipmap_image
 #error "Missing cl_khr_mipmap_image define"
@@ -275,15 +157,6 @@
 #endif
 #pragma OPENCL EXTENSION cl_khr_subgroups : enable
 
-#if (defined(__OPENCL_CPP_VERSION__) || __OPENCL_C_VERSION__ >= 200)
-#ifndef cl_khr_terminate_context
-#error "Missing cl_khr_terminate_context define"
-#endif
-#else
-// expected-warning@+2{{unsupported OpenCL extension 'cl_khr_terminate_context' - ignoring}}
-#endif
-#pragma OPENCL EXTENSION cl_khr_terminate_context: enable
-
 #ifndef cl_amd_media_ops
 #error "Missing cl_amd_media_ops define"
 #endif
diff --git a/clang/test/SemaTemplate/deduction.cpp b/clang/test/SemaTemplate/deduction.cpp
index a068bcaea0483..b9a1f0dccb243 100644
--- a/clang/test/SemaTemplate/deduction.cpp
+++ b/clang/test/SemaTemplate/deduction.cpp
@@ -312,6 +312,13 @@ namespace nullptr_deduction {
     f(X<nullptr_t, nullptr>()); // expected-note {{instantiation of}}
   }
 
+  template<template<typename T, T> class X, typename T, int *P>
+    void f0(X<T, P>) {} // expected-note {{deduced non-type template argument does not have the same type as the corresponding template parameter ('nullptr_t' vs 'int *')}}
+  void h0() {
+    f0(X<int*, nullptr>());
+    f0(X<nullptr_t, nullptr>()); // expected-error {{no matching function}}
+  }
+
   template<template<typename T, T> class X, typename T, typename U, int *P>
     void f1(X<T, P>, X<U, P>) {} // expected-note 2{{values of conflicting types}}
   void h() {
diff --git a/clang/test/SemaTemplate/temp_arg_nontype_cxx20.cpp b/clang/test/SemaTemplate/temp_arg_nontype_cxx20.cpp
index aa9e71ff1f6e2..9ad0bad36e2a9 100644
--- a/clang/test/SemaTemplate/temp_arg_nontype_cxx20.cpp
+++ b/clang/test/SemaTemplate/temp_arg_nontype_cxx20.cpp
@@ -1,11 +1,14 @@
 // RUN: %clang_cc1 -fsyntax-only -verify -std=c++20 %s
 
+using size_t = __SIZE_TYPE__;
+
 // floating-point arguments
 template<float> struct Float {};
 using F1 = Float<1.0f>; // FIXME expected-error {{sorry}}
 using F1 = Float<2.0f / 2>; // FIXME expected-error {{sorry}}
 
 struct S { int n[3]; } s; // expected-note 1+{{here}}
+union U { int a, b; } u;
 int n; // expected-note 1+{{here}}
 
 // pointers to subobjects
@@ -31,8 +34,13 @@ using IP3 = IntRef<*(s.n + 3)>; // expected-error {{not a constant expression}}
 
 // classes
 template<S> struct Struct {};
-using S123 = Struct<S{1, 2, 3}>; // FIXME: expected-error {{sorry}}
-using S123 = Struct<S{1, 2, 3}>; // FIXME: expected-error {{sorry}}
+using S123 = Struct<S{1, 2, 3}>;
+using S123 = Struct<S{1, 2, 3}>; // expected-note {{previous}}
+using S123 = Struct<S{1, 2, 4}>; // expected-error {{different types}}
+template<U> struct Union {};
+using U1 = Union<U{1}>;
+using U1 = Union<U{.a = 1}>; // expected-note {{previous}}
+using U1 = Union<U{.b = 1}>; // expected-error {{different types}}
 
 // miscellaneous scalar types
 template<_Complex int> struct ComplexInt {};
@@ -42,3 +50,173 @@ using CI = ComplexInt<1 + 3i>; // FIXME: expected-error {{sorry}}
 template<_Complex float> struct ComplexFloat {};
 using CF = ComplexFloat<1.0f + 3.0fi>; // FIXME: expected-error {{sorry}}
 using CF = ComplexFloat<1.0f + 3.0fi>; // FIXME: expected-error {{sorry}}
+
+namespace ClassNTTP {
+  struct A { // expected-note 2{{candidate}}
+    int x, y;
+  };
+  template<A a> constexpr int f() { return a.y; }
+  static_assert(f<A{1,2}>() == 2);
+
+  template<A a> int id;
+  constexpr A a = {1, 2};
+  static_assert(&id<A{1,2}> == &id<a>);
+  static_assert(&id<A{1,3}> != &id<a>);
+
+  int k = id<1>; // expected-error {{no viable conversion from 'int' to 'ClassNTTP::A'}}
+
+  struct B {
+    constexpr B() {}
+    constexpr B(int) = delete; // expected-note {{here}}
+  };
+  template<B> struct Q {}; // expected-note {{passing argument to parameter here}}
+  Q<1> q; // expected-error {{conversion function from 'int' to 'ClassNTTP::B' invokes a deleted function}}
+
+  struct C {
+    constexpr C() {}
+    C(const C&) = delete; // expected-note {{here}}
+  };
+  template<C> struct R {}; // expected-note {{passing argument to parameter here}}
+  constexpr C c;
+  R<c> r; // expected-error {{call to deleted constructor}}
+}
+
+namespace ConvertedConstant {
+  struct A {
+    constexpr A(float) {}
+  };
+  template <A> struct X {};
+  void f(X<1.0f>) {} // OK, user-defined conversion
+  void f(X<2>) {} // expected-error {{conversion from 'int' to 'ConvertedConstant::A' is not allowed in a converted constant expression}}
+}
+
+namespace CopyCounting {
+  // Make sure we don't use the copy constructor when transferring the "same"
+  // template parameter object around.
+  struct A { int n; constexpr A(int n = 0) : n(n) {} constexpr A(const A &a) : n(a.n+1) {} };
+  template<A a> struct X {};
+  template<A a> constexpr int f(X<a> x) { return a.n; }
+
+  static_assert(f(X<A{}>()) == 0);
+
+  template<A a> struct Y { void f(); };
+  template<A a> void g(Y<a> y) { y.Y<a>::f(); }
+  void h() { constexpr A a; g<a>(Y<a>{}); }
+
+  template<A a> struct Z {
+    constexpr int f() {
+      constexpr A v = a; // this is {a.n+1}
+      return Z<v>().f() + 1; // this is Z<{a.n+2}>
+    }
+  };
+  template<> struct Z<A{20}> {
+    constexpr int f() {
+      return 32;
+    }
+  };
+  static_assert(Z<A{}>().f() == 42);
+}
+
+namespace StableAddress {
+  template<size_t N> struct str {
+    char arr[N];
+  };
+  // FIXME: Deduction guide not needed with P1816R0.
+  template<size_t N> str(const char (&)[N]) -> str<N>;
+
+  template<str s> constexpr int sum() {
+    int n = 0;
+    for (char c : s.arr)
+      n += c;
+    return n;
+  }
+  static_assert(sum<str{"$hello $world."}>() == 1234);
+}
+
+namespace TemplateSpecializations {
+  struct A { int arr[10]; };
+  template<A> struct X; // expected-note {{here}}
+
+  using T = X<A{1, 2, 3}>;
+  using T = X<A{1, 2, 3, 0}>;
+  using T = X<A{1, 2, 3, 0, 0}>;
+  using T = X<A{1, 2, 3, 0, 0, 0}>;
+
+  template<> struct X<A{1, 2, 3, 4}> {};
+  X<A{1, 2, 3, 4, 0}> x;
+
+  template<auto V, auto W> constexpr bool Same = false;
+  template<auto V> constexpr bool Same<V, V> = true;
+  static_assert(Same<A{}, A{0, 0}>);
+  static_assert(Same<A{1}, A{1, 0}>);
+  static_assert(!Same<A{1}, A{1, 1}>);
+
+  // We can't directly specialize on member values...
+  template<int N> // expected-note {{parameter 'N'}}
+    struct X<A{N, N}> {}; // expected-error {{cannot be deduced}}
+
+  // ... but we can fake it up.
+  // template<int N> struct X<A{N, N}>
+  template <A V> requires Same<V, A{V.arr[0], V.arr[0]}>
+  struct X<V> {
+    static constexpr bool match = true;
+  };
+  static_assert(X<A{1, 1}>::match);
+  static_assert(X<A{2, 2}>::match);
+  static_assert(X<A{1, 2}>::match); // expected-error {{undefined}}
+
+  template<int, A> struct Y; // expected-note {{here}}
+  template<int N> struct Y<N, A{N, N, N}> {};
+  Y<1, A{1, 1, 1, 0}> y1;
+  Y<1, A{1, 1, 1, 1}> y2; // expected-error {{undefined}}
+
+  template<A, A> struct Z; // expected-note {{here}}
+  template<A V> struct Z<V, V> {};
+  Z<A{1, 2}, A{1, 2, 0}> z1;
+  Z<A{1, 2}, A{1, 3}> z2; // expected-error {{undefined}}
+
+  template struct Z<A{1}, A{1, 0}>;
+}
+
+namespace Diags {
+  struct A { int n, m; };
+  template<A a> struct X { static_assert(a.n == a.m); }; // expected-error {{static_assert failed due to requirement 'Diags::A{1, 2}.n == Diags::A{1, 2}.m'}}
+  template struct X<A{1, 2}>; // expected-note {{in instantiation of template class 'Diags::X<{1, 2}>' requested here}}
+}
+
+namespace CTADPartialOrder {
+  template<int> struct A {};
+  template<typename T, typename U, A a> struct X; // expected-note {{declared here}}
+  template<typename T, A a> struct X<T, int, a> { static constexpr int n = 1; }; // expected-note {{matches}}
+  template<typename T, A a> struct X<T *, int, a> { static constexpr int n = 2; };
+  template<typename T, A a> struct X<T, T, a> { static constexpr int n = 3; }; // expected-note {{matches}}
+
+  A<0> a;
+  static_assert(X<void, int, a>::n == 1);
+  static_assert(X<int*, int, a>::n == 2);
+  static_assert(X<void, void, a>::n == 3);
+  static_assert(X<int, int, a>::n == -1); // expected-error {{ambiguous}}
+  static_assert(X<int*, void, a>::n == 2); // expected-error {{undefined}}
+
+  template<typename T, A<0> a> struct X<T, T, a> { static constexpr int n = 4; };
+  static_assert(X<float, float, a>::n == 4);
+}
+
+namespace UnnamedBitfield {
+  struct A {
+    __INT32_TYPE__ : 32;
+  };
+  // Make sure we don't distinguish between the unnamed bit-field being
+  // uninitialized and it being zeroed. Those are not distinct states
+  // according to [temp.type]p2.
+  //
+  // FIXME: We shouldn't track a value for unnamed bit-fields, nor number
+  // them when computing field indexes.
+  template <A> struct X {};
+  constexpr A a;
+  using T = X<a>;
+  using T = X<A{}>;
+  using T = X<(A())>;
+  // Once we support bit-casts involving bit-fields, this should be valid too.
+  using T = X<__builtin_bit_cast(A, 0)>; // expected-error {{constant}} expected-note {{not yet supported}}
+}
diff --git a/clang/test/lit.cfg.py b/clang/test/lit.cfg.py
index 004ee39a9795f..91161b8244f22 100644
--- a/clang/test/lit.cfg.py
+++ b/clang/test/lit.cfg.py
@@ -1,7 +1,6 @@
 # -*- Python -*-
 
 import os
-import sys
 import platform
 import re
 import subprocess
@@ -168,9 +167,11 @@ def is_filesystem_case_insensitive():
 if platform.system() not in ['Windows']:
     config.available_features.add('can-remove-opened-file')
 
-# Check 64-bit host
-if sys.maxsize > 2**32:
-  config.available_features.add("clang-64-bits")
+# Features
+known_arches = ["x86_64", "mips64", "ppc64", "aarch64"]
+if (any(config.target_triple.startswith(x) for x in known_arches)):
+  config.available_features.add("clang-target-64-bits")
+
 
 
 def calculate_arch_features(arch_string):
diff --git a/clang/tools/libclang/CIndex.cpp b/clang/tools/libclang/CIndex.cpp
index 2662092960445..2b20b7c29224a 100644
--- a/clang/tools/libclang/CIndex.cpp
+++ b/clang/tools/libclang/CIndex.cpp
@@ -6381,6 +6381,7 @@ CXCursor clang_getCursorDefinition(CXCursor C) {
   case Decl::Binding:
   case Decl::MSProperty:
   case Decl::MSGuid:
+  case Decl::TemplateParamObject:
   case Decl::IndirectField:
   case Decl::ObjCIvar:
   case Decl::ObjCAtDefsField:
diff --git a/clang/tools/libclang/CMakeLists.txt b/clang/tools/libclang/CMakeLists.txt
index 15f7ff94dfead..51ff2e7e15659 100644
--- a/clang/tools/libclang/CMakeLists.txt
+++ b/clang/tools/libclang/CMakeLists.txt
@@ -174,7 +174,6 @@ install(DIRECTORY ../../include/clang-c
   DESTINATION "${LIBCLANG_HEADERS_INSTALL_DESTINATION}"
   FILES_MATCHING
   PATTERN "*.h"
-  PATTERN ".svn" EXCLUDE
   )
 
 # LLVM_DISTRIBUTION_COMPONENTS requires that each component have both a
diff --git a/compiler-rt/lib/asan/asan_report.cpp b/compiler-rt/lib/asan/asan_report.cpp
index 4b4db1db6dc9c..03f1ed2b01866 100644
--- a/compiler-rt/lib/asan/asan_report.cpp
+++ b/compiler-rt/lib/asan/asan_report.cpp
@@ -151,7 +151,8 @@ class ScopedInErrorReport {
     if (common_flags()->print_cmdline)
       PrintCmdline();
 
-    if (common_flags()->print_module_map == 2) PrintModuleMap();
+    if (common_flags()->print_module_map == 2)
+      DumpProcessMap();
 
     // Copy the message buffer so that we could start logging without holding a
     // lock that gets aquired during printing.
diff --git a/compiler-rt/lib/asan/asan_rtl.cpp b/compiler-rt/lib/asan/asan_rtl.cpp
index 115733cdaa48e..7b5a929963c6a 100644
--- a/compiler-rt/lib/asan/asan_rtl.cpp
+++ b/compiler-rt/lib/asan/asan_rtl.cpp
@@ -45,7 +45,8 @@ static void AsanDie() {
     // Don't die twice - run a busy loop.
     while (1) { }
   }
-  if (common_flags()->print_module_map >= 1) PrintModuleMap();
+  if (common_flags()->print_module_map >= 1)
+    DumpProcessMap();
   if (flags()->sleep_before_dying) {
     Report("Sleeping for %d second(s)\n", flags()->sleep_before_dying);
     SleepForSeconds(flags()->sleep_before_dying);
diff --git a/compiler-rt/lib/builtins/riscv/int_mul_impl.inc b/compiler-rt/lib/builtins/riscv/int_mul_impl.inc
index 50951d5f4195f..53699b356f6a8 100644
--- a/compiler-rt/lib/builtins/riscv/int_mul_impl.inc
+++ b/compiler-rt/lib/builtins/riscv/int_mul_impl.inc
@@ -10,7 +10,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#if !defined(__riscv_mul)
+#ifndef __mulxi3
+#error "__mulxi3 must be defined to use this generic implementation"
+#endif
+
 	.text
 	.align 2
 
@@ -28,4 +31,3 @@ __mulxi3:
 	slli   a2, a2, 1
 	bnez   a1, .L1
 	ret
-#endif
diff --git a/compiler-rt/lib/lsan/lsan.cpp b/compiler-rt/lib/lsan/lsan.cpp
index c8cc045783d45..2c0a3bf0787c2 100644
--- a/compiler-rt/lib/lsan/lsan.cpp
+++ b/compiler-rt/lib/lsan/lsan.cpp
@@ -77,7 +77,7 @@ static void InitializeFlags() {
   parser.ParseString(lsan_default_options);
   parser.ParseStringFromEnv("LSAN_OPTIONS");
 
-  SetVerbosity(common_flags()->verbosity);
+  InitializeCommonFlags();
 
   if (Verbosity()) ReportUnrecognizedFlags();
 
diff --git a/compiler-rt/lib/lsan/lsan_interceptors.cpp b/compiler-rt/lib/lsan/lsan_interceptors.cpp
index 39d075ddec09c..bf8d316770ee1 100644
--- a/compiler-rt/lib/lsan/lsan_interceptors.cpp
+++ b/compiler-rt/lib/lsan/lsan_interceptors.cpp
@@ -115,7 +115,11 @@ INTERCEPTOR(void*, memalign, uptr alignment, uptr size) {
   return lsan_memalign(alignment, size, stack);
 }
 #define LSAN_MAYBE_INTERCEPT_MEMALIGN INTERCEPT_FUNCTION(memalign)
+#else
+#define LSAN_MAYBE_INTERCEPT_MEMALIGN
+#endif  // SANITIZER_INTERCEPT_MEMALIGN
 
+#if SANITIZER_INTERCEPT___LIBC_MEMALIGN
 INTERCEPTOR(void *, __libc_memalign, uptr alignment, uptr size) {
   ENSURE_LSAN_INITED;
   GET_STACK_TRACE_MALLOC;
@@ -125,9 +129,8 @@ INTERCEPTOR(void *, __libc_memalign, uptr alignment, uptr size) {
 }
 #define LSAN_MAYBE_INTERCEPT___LIBC_MEMALIGN INTERCEPT_FUNCTION(__libc_memalign)
 #else
-#define LSAN_MAYBE_INTERCEPT_MEMALIGN
 #define LSAN_MAYBE_INTERCEPT___LIBC_MEMALIGN
-#endif // SANITIZER_INTERCEPT_MEMALIGN
+#endif  // SANITIZER_INTERCEPT___LIBC_MEMALIGN
 
 #if SANITIZER_INTERCEPT_ALIGNED_ALLOC
 INTERCEPTOR(void*, aligned_alloc, uptr alignment, uptr size) {
diff --git a/compiler-rt/lib/memprof/memprof_interface_internal.h b/compiler-rt/lib/memprof/memprof_interface_internal.h
index ea410f5cab063..a51d83a2521c9 100644
--- a/compiler-rt/lib/memprof/memprof_interface_internal.h
+++ b/compiler-rt/lib/memprof/memprof_interface_internal.h
@@ -46,6 +46,9 @@ const char *__memprof_default_options();
 SANITIZER_INTERFACE_ATTRIBUTE
 extern uptr __memprof_shadow_memory_dynamic_address;
 
+SANITIZER_INTERFACE_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE extern char
+    __memprof_profile_filename[1];
+
 SANITIZER_INTERFACE_ATTRIBUTE void __memprof_load(uptr p);
 SANITIZER_INTERFACE_ATTRIBUTE void __memprof_store(uptr p);
 
diff --git a/compiler-rt/lib/memprof/memprof_rtl.cpp b/compiler-rt/lib/memprof/memprof_rtl.cpp
index 61270406e48c6..93264dd562d08 100644
--- a/compiler-rt/lib/memprof/memprof_rtl.cpp
+++ b/compiler-rt/lib/memprof/memprof_rtl.cpp
@@ -27,6 +27,9 @@
 
 uptr __memprof_shadow_memory_dynamic_address; // Global interface symbol.
 
+// Allow the user to specify a profile output file via the binary.
+SANITIZER_WEAK_ATTRIBUTE char __memprof_profile_filename[1];
+
 namespace __memprof {
 
 static void MemprofDie() {
@@ -172,7 +175,12 @@ static void MemprofInitInternal() {
   AddDieCallback(MemprofDie);
   SetCheckFailedCallback(MemprofCheckFailed);
 
-  __sanitizer_set_report_path(common_flags()->log_path);
+  // Use profile name specified via the binary itself if it exists, and hasn't
+  // been overrriden by a flag at runtime.
+  if (__memprof_profile_filename[0] != 0 && !common_flags()->log_path)
+    __sanitizer_set_report_path(__memprof_profile_filename);
+  else
+    __sanitizer_set_report_path(common_flags()->log_path);
 
   __sanitizer::InitializePlatformEarly();
 
diff --git a/compiler-rt/lib/memprof/weak_symbols.txt b/compiler-rt/lib/memprof/weak_symbols.txt
index bb2dea85574fa..271813612ab60 100644
--- a/compiler-rt/lib/memprof/weak_symbols.txt
+++ b/compiler-rt/lib/memprof/weak_symbols.txt
@@ -1 +1 @@
-___memprof_default_options
+___memprof_default_options __memprof_profile_filename
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_common.h b/compiler-rt/lib/sanitizer_common/sanitizer_common.h
index 040db6fc2a16b..bce24d68045b1 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_common.h
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_common.h
@@ -254,7 +254,6 @@ void UpdateProcessName();
 void CacheBinaryName();
 void DisableCoreDumperIfNecessary();
 void DumpProcessMap();
-void PrintModuleMap();
 const char *GetEnv(const char *name);
 bool SetEnv(const char *name, const char *value);
 
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_file.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_file.cpp
index 79930d7942501..6a7996797cbca 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_file.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_file.cpp
@@ -69,24 +69,23 @@ void ReportFile::ReopenIfNecessary() {
 }
 
 void ReportFile::SetReportPath(const char *path) {
-  if (!path)
-    return;
-  uptr len = internal_strlen(path);
-  if (len > sizeof(path_prefix) - 100) {
-    Report("ERROR: Path is too long: %c%c%c%c%c%c%c%c...\n",
-           path[0], path[1], path[2], path[3],
-           path[4], path[5], path[6], path[7]);
-    Die();
+  if (path) {
+    uptr len = internal_strlen(path);
+    if (len > sizeof(path_prefix) - 100) {
+      Report("ERROR: Path is too long: %c%c%c%c%c%c%c%c...\n", path[0], path[1],
+             path[2], path[3], path[4], path[5], path[6], path[7]);
+      Die();
+    }
   }
 
   SpinMutexLock l(mu);
   if (fd != kStdoutFd && fd != kStderrFd && fd != kInvalidFd)
     CloseFile(fd);
   fd = kInvalidFd;
-  if (internal_strcmp(path, "stdout") == 0) {
-    fd = kStdoutFd;
-  } else if (internal_strcmp(path, "stderr") == 0) {
+  if (!path || internal_strcmp(path, "stderr") == 0) {
     fd = kStderrFd;
+  } else if (internal_strcmp(path, "stdout") == 0) {
+    fd = kStdoutFd;
   } else {
     internal_snprintf(path_prefix, kMaxPathLength, "%s", path);
   }
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_flags.inc b/compiler-rt/lib/sanitizer_common/sanitizer_flags.inc
index d8e809b060944..cfb5822645f13 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_flags.inc
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_flags.inc
@@ -52,9 +52,9 @@ COMMON_FLAG(bool, handle_ioctl, false, "Intercept and handle ioctl requests.")
 COMMON_FLAG(int, malloc_context_size, 1,
             "Max number of stack frames kept for each allocation/deallocation.")
 COMMON_FLAG(
-    const char *, log_path, "stderr",
+    const char *, log_path, nullptr,
     "Write logs to \"log_path.pid\". The special values are \"stdout\" and "
-    "\"stderr\". The default is \"stderr\".")
+    "\"stderr\". If unspecified, defaults to \"stderr\".")
 COMMON_FLAG(
     bool, log_exe_name, false,
     "Mention name of executable when reporting error and "
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_fuchsia.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_fuchsia.cpp
index 6d1ad79467706..a6034ebda936e 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_fuchsia.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_fuchsia.cpp
@@ -105,8 +105,6 @@ void SetAlternateSignalStack() {}
 void UnsetAlternateSignalStack() {}
 void InitTlsSize() {}
 
-void PrintModuleMap() {}
-
 bool SignalContext::IsStackOverflow() const { return false; }
 void SignalContext::DumpAllRegisters(void *context) { UNIMPLEMENTED(); }
 const char *SignalContext::Describe() const { UNIMPLEMENTED(); }
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_linux.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_linux.cpp
index 024bef3c05829..0a1bb1771ecab 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_linux.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_linux.cpp
@@ -2218,8 +2218,6 @@ void CheckMPROTECT() {
 #endif
 }
 
-void PrintModuleMap() { }
-
 void CheckNoDeepBind(const char *filename, int flag) {
 #ifdef RTLD_DEEPBIND
   if (flag & RTLD_DEEPBIND) {
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_mac.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_mac.cpp
index b1271120c0041..ab7005c086930 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_mac.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_mac.cpp
@@ -1300,7 +1300,7 @@ void FormatUUID(char *out, uptr size, const u8 *uuid) {
                     uuid[12], uuid[13], uuid[14], uuid[15]);
 }
 
-void PrintModuleMap() {
+void DumpProcessMap() {
   Printf("Process module map:\n");
   MemoryMappingLayout memory_mapping(false);
   InternalMmapVector<LoadedModule> modules;
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h b/compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h
index 80a6b5990f24d..3e512d0f40ee0 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h
@@ -499,12 +499,15 @@
    SI_NOT_RTEMS && !SI_SOLARIS)  // NOLINT
 #define SANITIZER_INTERCEPT_MEMALIGN \
   (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_RTEMS)
+#define SANITIZER_INTERCEPT___LIBC_MEMALIGN                               \
+  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && !SI_OPENBSD && SI_NOT_RTEMS && \
+   !SI_ANDROID)
 #define SANITIZER_INTERCEPT_PVALLOC                                         \
   (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_FUCHSIA && \
    SI_NOT_RTEMS && !SI_SOLARIS)  // NOLINT
-#define SANITIZER_INTERCEPT_CFREE                                           \
-  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_FUCHSIA && \
-   SI_NOT_RTEMS && !SI_SOLARIS)  // NOLINT
+#define SANITIZER_INTERCEPT_CFREE                                            \
+  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_FUCHSIA && SI_NOT_RTEMS && \
+   !SI_SOLARIS && !SANITIZER_ANDROID)  // NOLINT
 #define SANITIZER_INTERCEPT_REALLOCARRAY SI_POSIX
 #define SANITIZER_INTERCEPT_ALIGNED_ALLOC (!SI_MAC && SI_NOT_RTEMS)
 #define SANITIZER_INTERCEPT_MALLOC_USABLE_SIZE \
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_posix.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_posix.cpp
index b8b75c20d9f9a..2e080098283fd 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_posix.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_posix.cpp
@@ -239,6 +239,7 @@ bool MemoryRangeIsAvailable(uptr range_start, uptr range_end) {
   return true;
 }
 
+#if !SANITIZER_MAC
 void DumpProcessMap() {
   MemoryMappingLayout proc_maps(/*cache_enabled*/true);
   const sptr kBufSize = 4095;
@@ -252,6 +253,7 @@ void DumpProcessMap() {
   Report("End of process memory map.\n");
   UnmapOrDie(filename, kBufSize);
 }
+#endif
 
 const char *GetPwd() {
   return GetEnv("PWD");
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_rtems.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_rtems.cpp
index 29bcfcfa6f158..d58bd08fb1a89 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_rtems.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_rtems.cpp
@@ -108,8 +108,6 @@ void SetAlternateSignalStack() {}
 void UnsetAlternateSignalStack() {}
 void InitTlsSize() {}
 
-void PrintModuleMap() {}
-
 void SignalContext::DumpAllRegisters(void *context) {}
 const char *DescribeSignalOrException(int signo) { UNIMPLEMENTED(); }
 
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_win.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_win.cpp
index 53a537d398475..6c02b331daffd 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_win.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_win.cpp
@@ -491,8 +491,6 @@ void DumpProcessMap() {
 }
 #endif
 
-void PrintModuleMap() { }
-
 void DisableCoreDumperIfNecessary() {
   // Do nothing.
 }
diff --git a/compiler-rt/lib/tsan/rtl/tsan_report.cpp b/compiler-rt/lib/tsan/rtl/tsan_report.cpp
index 4892c446c104b..968c7b97553c6 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_report.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_report.cpp
@@ -386,7 +386,8 @@ void PrintReport(const ReportDesc *rep) {
       ReportErrorSummary(rep_typ_str, frame->info);
   }
 
-  if (common_flags()->print_module_map == 2) PrintModuleMap();
+  if (common_flags()->print_module_map == 2)
+    DumpProcessMap();
 
   Printf("==================\n");
 }
diff --git a/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp b/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp
index 7b37ed50681e6..3d721eb95a2cf 100644
--- a/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp
+++ b/compiler-rt/lib/tsan/rtl/tsan_rtl.cpp
@@ -446,7 +446,8 @@ void MaybeSpawnBackgroundThread() {
 int Finalize(ThreadState *thr) {
   bool failed = false;
 
-  if (common_flags()->print_module_map == 1) PrintModuleMap();
+  if (common_flags()->print_module_map == 1)
+    DumpProcessMap();
 
   if (flags()->atexit_sleep_ms > 0 && ThreadCount(thr) > 1)
     SleepForMillis(flags()->atexit_sleep_ms);
diff --git a/compiler-rt/test/asan/TestCases/coverage-and-lsan.cpp b/compiler-rt/test/asan/TestCases/coverage-and-lsan.cpp
index 60851dabb6b7f..d95d95fc5d5fe 100644
--- a/compiler-rt/test/asan/TestCases/coverage-and-lsan.cpp
+++ b/compiler-rt/test/asan/TestCases/coverage-and-lsan.cpp
@@ -9,7 +9,8 @@
 // RUN: %sancov print %t-dir/*.sancov 2>&1
 //
 // REQUIRES: leak-detection
-
+// FIXME: sancov paths not work with adb
+// UNSUPPORTED: android
 int *g = new int;
 int main(int argc, char **argv) {
   g = 0;
diff --git a/compiler-rt/test/asan/lit.cfg.py b/compiler-rt/test/asan/lit.cfg.py
index 3a338d79e2225..4f56c886e92e3 100644
--- a/compiler-rt/test/asan/lit.cfg.py
+++ b/compiler-rt/test/asan/lit.cfg.py
@@ -209,10 +209,11 @@ def build_invocation(compile_flags):
   config.available_features.add('fast-unwinder-works')
 
 # Turn on leak detection on 64-bit Linux.
-leak_detection_linux = (config.host_os == 'Linux') and (not config.android) and (config.target_arch == 'x86_64' or config.target_arch == 'i386')
+leak_detection_android = config.android and 'android-thread-properties-api' in config.available_features and (config.target_arch in ['x86_64', 'i386', 'i686', 'aarch64'])
+leak_detection_linux = (config.host_os == 'Linux') and (not config.android) and (config.target_arch in ['x86_64', 'i386'])
 leak_detection_mac = (config.host_os == 'Darwin') and (config.target_arch == 'x86_64')
 leak_detection_netbsd = (config.host_os == 'NetBSD') and (config.target_arch in ['x86_64', 'i386'])
-if leak_detection_linux or leak_detection_mac or leak_detection_netbsd:
+if leak_detection_android or leak_detection_linux or leak_detection_mac or leak_detection_netbsd:
   config.available_features.add('leak-detection')
 
 # Set LD_LIBRARY_PATH to pick dynamic runtime up properly.
diff --git a/compiler-rt/test/builtins/Unit/muldi3_test.c b/compiler-rt/test/builtins/Unit/muldi3_test.c
index 74530cbfb09a6..c03570f029809 100644
--- a/compiler-rt/test/builtins/Unit/muldi3_test.c
+++ b/compiler-rt/test/builtins/Unit/muldi3_test.c
@@ -11,7 +11,7 @@ int test__muldi3(di_int a, di_int b, di_int expected)
     di_int x = __muldi3(a, b);
     if (x != expected)
         printf("error in __muldi3: %lld * %lld = %lld, expected %lld\n",
-               a, b, __muldi3(a, b), expected);
+               a, b, x, expected);
     return x != expected;
 }
 
diff --git a/compiler-rt/test/builtins/Unit/riscv/mulsi3_test.c b/compiler-rt/test/builtins/Unit/mulsi3_test.c
similarity index 90%
rename from compiler-rt/test/builtins/Unit/riscv/mulsi3_test.c
rename to compiler-rt/test/builtins/Unit/mulsi3_test.c
index d4621acbbbb5a..0b0c14204efdf 100644
--- a/compiler-rt/test/builtins/Unit/riscv/mulsi3_test.c
+++ b/compiler-rt/test/builtins/Unit/mulsi3_test.c
@@ -1,13 +1,10 @@
-// REQUIRES: riscv32-target-arch
 // RUN: %clang_builtins %s %librt -o %t && %run %t
+// REQUIRES: librt_has_mulsi3
 
 #include "int_lib.h"
 #include <stdio.h>
 #include <limits.h>
 
-#if !defined(__riscv_mul) && __riscv_xlen == 32
-// Based on mulsi3_test.c
-
 COMPILER_RT_ABI si_int __mulsi3(si_int a, si_int b);
 
 int test__mulsi3(si_int a, si_int b, si_int expected)
@@ -15,14 +12,12 @@ int test__mulsi3(si_int a, si_int b, si_int expected)
     si_int x = __mulsi3(a, b);
     if (x != expected)
         printf("error in __mulsi3: %d * %d = %d, expected %d\n",
-               a, b, __mulsi3(a, b), expected);
+               a, b, x, expected);
     return x != expected;
 }
-#endif
 
 int main()
 {
-#if !defined(__riscv_mul) && __riscv_xlen == 32
     if (test__mulsi3(0, 0, 0))
         return 1;
     if (test__mulsi3(0, 1, 0))
@@ -99,9 +94,6 @@ int main()
         return 1;
     if (test__mulsi3(-8192, -4194303, 34359730176))
         return 1;
-#else
-    printf("skipped\n");
-#endif
 
     return 0;
 }
diff --git a/compiler-rt/test/lit.common.cfg.py b/compiler-rt/test/lit.common.cfg.py
index d955f8829e5d3..742da04fe059c 100644
--- a/compiler-rt/test/lit.common.cfg.py
+++ b/compiler-rt/test/lit.common.cfg.py
@@ -357,6 +357,14 @@ def get_macos_aligned_version(macos_vers):
     config.environment['ANDROID_SERIAL'] = config.android_serial
 
   adb = os.environ.get('ADB', 'adb')
+
+  # These are needed for tests to upload/download temp files, such as
+  # suppression-files, to device.
+  config.substitutions.append( ('%device_rundir', "/data/local/tmp/Output") )
+  config.substitutions.append( ('%push_to_device', "%s -s '%s' push " % (adb, env['ANDROID_SERIAL']) ) )
+  config.substitutions.append( ('%pull_from_device', "%s -s '%s' pull " % (adb, env['ANDROID_SERIAL']) ) )
+  config.substitutions.append( ('%adb_shell ', "%s -s '%s' shell " % (adb, env['ANDROID_SERIAL']) ) )
+
   try:
     android_api_level_str = subprocess.check_output([adb, "shell", "getprop", "ro.build.version.sdk"], env=env).rstrip()
   except (subprocess.CalledProcessError, OSError):
@@ -375,6 +383,11 @@ def get_macos_aligned_version(macos_vers):
   subprocess.check_call([adb, "shell", "mkdir", "-p", android_tmpdir], env=env)
   for file in config.android_files_to_push:
     subprocess.check_call([adb, "push", file, android_tmpdir], env=env)
+else:
+  config.substitutions.append( ('%device_rundir', "") )
+  config.substitutions.append( ('%push_to_device', "echo ") )
+  config.substitutions.append( ('%pull_from_device', "echo ") )
+  config.substitutions.append( ('%adb_shell', "echo ") )
 
 if config.host_os == 'Linux':
   # detect whether we are using glibc, and which version
diff --git a/compiler-rt/test/lsan/TestCases/Linux/cleanup_in_tsd_destructor.c b/compiler-rt/test/lsan/TestCases/Linux/cleanup_in_tsd_destructor.c
index cf080e4dd7562..18476d5a9adec 100644
--- a/compiler-rt/test/lsan/TestCases/Linux/cleanup_in_tsd_destructor.c
+++ b/compiler-rt/test/lsan/TestCases/Linux/cleanup_in_tsd_destructor.c
@@ -3,7 +3,7 @@
 // user-installed TSD destructors have finished running (since they may contain
 // additional cleanup tasks). LSan doesn't actually meet that goal 100%, but it
 // makes its best effort.
-// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_globals=0"
+// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0"
 // RUN: %clang_lsan %s -o %t
 // RUN: %env_lsan_opts=$LSAN_BASE:use_tls=1 %run %t
 // RUN: %env_lsan_opts=$LSAN_BASE:use_tls=0 not %run %t 2>&1 | FileCheck %s
diff --git a/compiler-rt/test/lsan/TestCases/Linux/disabler_in_tsd_destructor.c b/compiler-rt/test/lsan/TestCases/Linux/disabler_in_tsd_destructor.c
index 52819bb9fccf2..be36b52fc34f2 100644
--- a/compiler-rt/test/lsan/TestCases/Linux/disabler_in_tsd_destructor.c
+++ b/compiler-rt/test/lsan/TestCases/Linux/disabler_in_tsd_destructor.c
@@ -1,5 +1,5 @@
 // Regression test. Disabler should not depend on TSD validity.
-// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_globals=0:use_tls=1:use_ld_allocations=0"
+// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_tls=1:use_ld_allocations=0"
 // RUN: %clang_lsan %s -o %t
 // RUN: %env_lsan_opts=$LSAN_BASE %run %t
 
diff --git a/compiler-rt/test/lsan/TestCases/Linux/guard-page.c b/compiler-rt/test/lsan/TestCases/Linux/guard-page.c
index 7b353ce30f896..ba03c1ff3535b 100644
--- a/compiler-rt/test/lsan/TestCases/Linux/guard-page.c
+++ b/compiler-rt/test/lsan/TestCases/Linux/guard-page.c
@@ -1,6 +1,9 @@
 // Check that if LSan finds that SP doesn't point into thread stack (e.g.
 // if swapcontext is used), LSan will not hit the guard page.
 // RUN: %clang_lsan %s -o %t && %run %t
+// Missing 'getcontext' and 'makecontext' on Android.
+// UNSUPPORTED: android
+
 #include <errno.h>
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/compiler-rt/test/lsan/TestCases/Linux/log-path_test.cpp b/compiler-rt/test/lsan/TestCases/Linux/log-path_test.cpp
index a31b4f64acc5d..aba3aa12411ce 100644
--- a/compiler-rt/test/lsan/TestCases/Linux/log-path_test.cpp
+++ b/compiler-rt/test/lsan/TestCases/Linux/log-path_test.cpp
@@ -7,7 +7,11 @@
 
 // Good log_path.
 // RUN: rm -f %t.log.*
-// RUN: %env_lsan_opts="use_stacks=0:log_path='"%t.log"'" not %run %t > %t.out 2>&1
+// RUN: %adb_shell 'rm -f %t.log.*'
+// RUN: %env_lsan_opts="use_stacks=0:log_path='"%device_rundir/%t.log"'" not %run %t > %t.out 2>&1
+// adb-pull doesn't support wild cards so we need to rename the log file.
+// RUN: %adb_shell 'mv %device_rundir/%t.log.* %device_rundir/%t.log'
+// RUN: %pull_from_device %device_rundir/%t.log %t.log.ANDROID
 // RUN: FileCheck %s --check-prefix=CHECK-ERROR < %t.log.*
 
 #include <stdio.h>
diff --git a/compiler-rt/test/lsan/TestCases/Linux/use_tls_dynamic.cpp b/compiler-rt/test/lsan/TestCases/Linux/use_tls_dynamic.cpp
index b7ca2d754b1c1..14368ef407beb 100644
--- a/compiler-rt/test/lsan/TestCases/Linux/use_tls_dynamic.cpp
+++ b/compiler-rt/test/lsan/TestCases/Linux/use_tls_dynamic.cpp
@@ -1,6 +1,6 @@
 // Test that dynamically allocated TLS space is included in the root set.
 
-// This is known to be broken with glibc-2.27+
+// This is known to be broken with glibc-2.27+ but it should pass with Bionic
 // https://bugs.llvm.org/show_bug.cgi?id=37804
 // XFAIL: glibc-2.27
 
@@ -10,7 +10,7 @@
 // RUN: %env_lsan_opts=$LSAN_BASE:"use_tls=0" not %run %t 2>&1 | FileCheck %s
 // RUN: %env_lsan_opts=$LSAN_BASE:"use_tls=1" %run %t 2>&1
 // RUN: %env_lsan_opts="" %run %t 2>&1
-// UNSUPPORTED: i386-linux,arm,powerpc
+// UNSUPPORTED: arm,powerpc
 
 #ifndef BUILD_DSO
 #include <assert.h>
@@ -27,8 +27,15 @@ int main(int argc, char *argv[]) {
   assert(handle != 0);
   typedef void **(* store_t)(void *p);
   store_t StoreToTLS = (store_t)dlsym(handle, "StoreToTLS");
-  assert(dlerror() == 0);
 
+  // Sometimes dlerror() occurs when we broke the interceptors.
+  // Add the message here to make the error more obvious.
+  const char *dlerror_msg = dlerror();
+  assert(dlerror_msg == nullptr);
+  if (dlerror_msg != nullptr) {
+    fprintf(stderr, "DLERROR: %s\n", dlerror_msg);
+    fflush(stderr);
+  }
   void *p = malloc(1337);
   // If we don't  know about dynamic TLS, we will return a false leak above.
   void **p_in_tls = StoreToTLS(p);
diff --git a/compiler-rt/test/lsan/TestCases/Linux/use_tls_pthread_specific_static.cpp b/compiler-rt/test/lsan/TestCases/Linux/use_tls_pthread_specific_static.cpp
index cafe40f063765..c4398c56ea0b3 100644
--- a/compiler-rt/test/lsan/TestCases/Linux/use_tls_pthread_specific_static.cpp
+++ b/compiler-rt/test/lsan/TestCases/Linux/use_tls_pthread_specific_static.cpp
@@ -19,7 +19,10 @@ int main() {
   int res;
   res = pthread_key_create(&key, NULL);
   assert(res == 0);
+#if !defined(__ANDROID__) && !defined(__BIONIC__)
+  // Bionic doesn't have specific limit.
   assert(key < PTHREAD_KEY_2NDLEVEL_SIZE);
+#endif
   void *p = malloc(1337);
   res = pthread_setspecific(key, p);
   assert(res == 0);
diff --git a/compiler-rt/test/lsan/TestCases/disabler.c b/compiler-rt/test/lsan/TestCases/disabler.c
index f8b7f0da1ba6c..73f1b3e36e937 100644
--- a/compiler-rt/test/lsan/TestCases/disabler.c
+++ b/compiler-rt/test/lsan/TestCases/disabler.c
@@ -1,5 +1,5 @@
 // Test for __lsan_disable() / __lsan_enable().
-// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_globals=0:use_tls=0"
+// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_tls=0"
 // RUN: %clang_lsan %s -o %t
 // RUN: %env_lsan_opts=$LSAN_BASE not %run %t 2>&1 | FileCheck %s
 
diff --git a/compiler-rt/test/lsan/TestCases/disabler.cpp b/compiler-rt/test/lsan/TestCases/disabler.cpp
index c5ffdb0bf6f10..21e508e5314b0 100644
--- a/compiler-rt/test/lsan/TestCases/disabler.cpp
+++ b/compiler-rt/test/lsan/TestCases/disabler.cpp
@@ -1,5 +1,5 @@
 // Test for ScopedDisabler.
-// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_globals=0:use_tls=0"
+// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_tls=0"
 // RUN: %clangxx_lsan %s -o %t
 // RUN: %env_lsan_opts=$LSAN_BASE not %run %t 2>&1 | FileCheck %s
 
diff --git a/compiler-rt/test/lsan/TestCases/ignore_object.c b/compiler-rt/test/lsan/TestCases/ignore_object.c
index 53dea75947bb4..92f03ee33a05d 100644
--- a/compiler-rt/test/lsan/TestCases/ignore_object.c
+++ b/compiler-rt/test/lsan/TestCases/ignore_object.c
@@ -1,5 +1,5 @@
 // Test for __lsan_ignore_object().
-// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_globals=0:use_tls=0"
+// RUN: LSAN_BASE="report_objects=1:use_registers=0:use_stacks=0:use_tls=0"
 // RUN: %clang_lsan %s -o %t
 // RUN: %env_lsan_opts=$LSAN_BASE not %run %t 2>&1 | FileCheck %s
 
diff --git a/compiler-rt/test/lsan/TestCases/large_allocation_leak.cpp b/compiler-rt/test/lsan/TestCases/large_allocation_leak.cpp
index 66f364fffcd52..07095388982c3 100644
--- a/compiler-rt/test/lsan/TestCases/large_allocation_leak.cpp
+++ b/compiler-rt/test/lsan/TestCases/large_allocation_leak.cpp
@@ -5,7 +5,7 @@
 
 // For 32 bit LSan it's pretty likely that large chunks are "reachable" from some
 // internal data structures (e.g. Glibc global data).
-// UNSUPPORTED: x86, arm
+// UNSUPPORTED: x86,i686,arm
 
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/compiler-rt/test/lsan/TestCases/many_threads_detach.cpp b/compiler-rt/test/lsan/TestCases/many_threads_detach.cpp
index 1f85bcf1825aa..857766ddee24e 100644
--- a/compiler-rt/test/lsan/TestCases/many_threads_detach.cpp
+++ b/compiler-rt/test/lsan/TestCases/many_threads_detach.cpp
@@ -1,5 +1,7 @@
 // Test that threads are reused.
-// RUN: %clangxx_lsan %s -o %t -lpthread && %run %t
+// On Android, pthread_* are in libc.so. So the `-lpthread` is not supported.
+// Use `-pthread` so that its driver will DTRT (ie., ignore it).
+// RUN: %clangxx_lsan %s -o %t -pthread && %run %t
 
 #include <pthread.h>
 #include <stdlib.h>
diff --git a/compiler-rt/test/lsan/TestCases/stale_stack_leak.cpp b/compiler-rt/test/lsan/TestCases/stale_stack_leak.cpp
index 505e62762629f..72aee5f56fd96 100644
--- a/compiler-rt/test/lsan/TestCases/stale_stack_leak.cpp
+++ b/compiler-rt/test/lsan/TestCases/stale_stack_leak.cpp
@@ -6,7 +6,7 @@
 //
 // x86 passes parameters through stack that may lead to false negatives
 // The same applies to s390x register save areas.
-// UNSUPPORTED: x86,powerpc64,arm,s390x
+// UNSUPPORTED: x86,i686,powerpc64,arm,s390x
 
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/compiler-rt/test/lsan/TestCases/strace_test.cpp b/compiler-rt/test/lsan/TestCases/strace_test.cpp
index b25e057538487..18c809ca32960 100644
--- a/compiler-rt/test/lsan/TestCases/strace_test.cpp
+++ b/compiler-rt/test/lsan/TestCases/strace_test.cpp
@@ -2,6 +2,9 @@
 // REQUIRES: strace
 // RUN: %clangxx_lsan %s -o %t
 // RUN: not strace -o /dev/null %run %t 2>&1 | FileCheck %s
+// FIXME: This technically works in practice but cannot be tested because the
+// fatal-error caused adb to failed. Could not be captured to stderr to lit-check.
+// XFAIL: android
 
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/compiler-rt/test/lsan/TestCases/suppressions_file.cpp b/compiler-rt/test/lsan/TestCases/suppressions_file.cpp
index 33cf0202d7319..29422bd3c2c0a 100644
--- a/compiler-rt/test/lsan/TestCases/suppressions_file.cpp
+++ b/compiler-rt/test/lsan/TestCases/suppressions_file.cpp
@@ -3,13 +3,16 @@
 
 // RUN: rm -f %t.supp
 // RUN: touch %t.supp
-// RUN: %env_lsan_opts="$LSAN_BASE:suppressions='%t.supp'" not %run %t 2>&1 | FileCheck %s --check-prefix=NOSUPP
+// RUN: %push_to_device %t.supp %device_rundir/%t.supp
+// RUN: %env_lsan_opts="$LSAN_BASE:suppressions='%device_rundir/%t.supp'" not %run %t 2>&1 | FileCheck %s --check-prefix=NOSUPP
 
 // RUN: echo "leak:*LSanTestLeakingFunc*" > %t.supp
-// RUN: %env_lsan_opts="$LSAN_BASE:suppressions='%t.supp'" not %run %t 2>&1 | FileCheck %s
-
+// RUN: %push_to_device  %t.supp %device_rundir/%t.supp
+// RUN: %env_lsan_opts="$LSAN_BASE:suppressions='%device_rundir/%t.supp'" not %run %t 2>&1 | FileCheck %s
+//
 // RUN: echo "leak:%t" > %t.supp
-// RUN: %env_lsan_opts="$LSAN_BASE:suppressions='%t.supp':symbolize=false" %run %t
+// RUN: %push_to_device  %t.supp %device_rundir/%t.supp
+// RUN: %env_lsan_opts="$LSAN_BASE:suppressions='%device_rundir/%t.supp':symbolize=false" %run %t
 
 #include <stdio.h>
 #include <stdlib.h>
diff --git a/compiler-rt/test/lsan/TestCases/swapcontext.cpp b/compiler-rt/test/lsan/TestCases/swapcontext.cpp
index afce8d9fe3b18..d0999598ad4c2 100644
--- a/compiler-rt/test/lsan/TestCases/swapcontext.cpp
+++ b/compiler-rt/test/lsan/TestCases/swapcontext.cpp
@@ -4,7 +4,8 @@
 // RUN: %clangxx_lsan %s -o %t
 // RUN: %env_lsan_opts= %run %t 2>&1
 // RUN: %env_lsan_opts= not %run %t foo 2>&1 | FileCheck %s
-// UNSUPPORTED: arm,powerpc64
+// Missing 'getcontext' and 'makecontext' on Android.
+// UNSUPPORTED: arm,powerpc64,android
 
 #include "sanitizer_common/sanitizer_ucontext.h"
 #include <stdio.h>
diff --git a/compiler-rt/test/lsan/TestCases/use_registers.cpp b/compiler-rt/test/lsan/TestCases/use_registers.cpp
index 60071767456d4..dcf3bb9009334 100644
--- a/compiler-rt/test/lsan/TestCases/use_registers.cpp
+++ b/compiler-rt/test/lsan/TestCases/use_registers.cpp
@@ -20,8 +20,8 @@ extern "C" void *registers_thread_func(void *arg) {
 
   // To store the pointer, choose a register which is unlikely to be reused by
   // a function call.
-#if defined(__i386__)
-  asm("mov %0, %%esi"
+#if defined(__i386__) || defined(__i686__)
+  asm("mov %0, %%edi"
       :
       : "r"(p));
 #elif defined(__x86_64__)
@@ -36,6 +36,12 @@ extern "C" void *registers_thread_func(void *arg) {
   asm("mov r5, %0"
       :
       : "r"(p));
+#elif defined(__aarch64__)
+  // x9-10are used. x11-12 are probably used.
+  // So we pick x13 to be safe.
+  asm("mov x13, %0"
+      :
+      : "r"(p));
 #elif defined(__powerpc__)
   asm("mr 30, %0"
       :
diff --git a/compiler-rt/test/lsan/lit.common.cfg.py b/compiler-rt/test/lsan/lit.common.cfg.py
index 1d393880af6a9..5adeec3e83a25 100644
--- a/compiler-rt/test/lsan/lit.common.cfg.py
+++ b/compiler-rt/test/lsan/lit.common.cfg.py
@@ -21,6 +21,7 @@ def get_required_attr(config, attr_name):
 
 # Choose between standalone and LSan+ASan modes.
 lsan_lit_test_mode = get_required_attr(config, 'lsan_lit_test_mode')
+
 if lsan_lit_test_mode == "Standalone":
   config.name = "LeakSanitizer-Standalone"
   lsan_cflags = ["-fsanitize=leak"]
@@ -35,7 +36,8 @@ def get_required_attr(config, attr_name):
 config.name += config.name_suffix
 
 # Platform-specific default LSAN_OPTIONS for lit tests.
-default_lsan_opts = 'detect_leaks=1'
+default_common_opts_str = ':'.join(list(config.default_sanitizer_opts))
+default_lsan_opts = default_common_opts_str + ':detect_leaks=1'
 if config.host_os == 'Darwin':
   # On Darwin, we default to `abort_on_error=1`, which would make tests run
   # much slower. Let's override this and run lit tests with 'abort_on_error=0'.
@@ -53,6 +55,8 @@ def get_required_attr(config, attr_name):
   config.available_features.add('strace')
 
 clang_cflags = ["-O0", config.target_cflags] + config.debug_info_flags
+if config.android:
+  clang_cflags = clang_cflags + ["-fno-emulated-tls"]
 clang_cxxflags = config.cxx_mode_flags + clang_cflags
 lsan_incdir = config.test_source_root + "/../"
 clang_lsan_cflags = clang_cflags + lsan_cflags + ["-I%s" % lsan_incdir]
@@ -69,11 +73,13 @@ def build_invocation(compile_flags):
 config.substitutions.append( ("%clang_lsan ", build_invocation(clang_lsan_cflags)) )
 config.substitutions.append( ("%clangxx_lsan ", build_invocation(clang_lsan_cxxflags)) )
 
-# LeakSanitizer tests are currently supported on x86-64 Linux, PowerPC64 Linux, arm Linux, mips64 Linux, s390x Linux and x86_64 Darwin.
-supported_linux = config.host_os == 'Linux' and config.host_arch in ['x86_64', 'ppc64', 'ppc64le', 'mips64', 'arm', 'armhf', 'armv7l', 's390x']
+# LeakSanitizer tests are currently supported on
+# Android{aarch64, x86, x86_64}, x86-64 Linux, PowerPC64 Linux, arm Linux, mips64 Linux, s390x Linux and x86_64 Darwin.
+supported_android = config.android and config.target_arch in ['x86_64', 'i386', 'aarch64'] and 'android-thread-properties-api' in config.available_features
+supported_linux = (not config.android) and config.host_os == 'Linux' and config.host_arch in ['x86_64', 'ppc64', 'ppc64le', 'mips64', 'arm', 'armhf', 'armv7l', 's390x']
 supported_darwin = config.host_os == 'Darwin' and config.target_arch in ['x86_64']
 supported_netbsd = config.host_os == 'NetBSD' and config.target_arch in ['x86_64', 'i386']
-if not (supported_linux or supported_darwin or supported_netbsd):
+if not (supported_android or supported_linux or supported_darwin or supported_netbsd):
   config.unsupported = True
 
 # Don't support Thumb due to broken fast unwinder
diff --git a/compiler-rt/test/memprof/TestCases/log_path_test.cpp b/compiler-rt/test/memprof/TestCases/log_path_test.cpp
index 0b1d4982f11cd..fbc05185af232 100644
--- a/compiler-rt/test/memprof/TestCases/log_path_test.cpp
+++ b/compiler-rt/test/memprof/TestCases/log_path_test.cpp
@@ -21,6 +21,24 @@
 // RUN:   not %run %t 2> %t.out
 // RUN: FileCheck %s --check-prefix=CHECK-LONG < %t.out
 
+// Specifying the log name via the __memprof_profile_filename variable.
+// RUN: %clangxx_memprof  %s -o %t -DPROFILE_NAME_VAR="%t.log2"
+// RUN: rm -f %t.log2.*
+// RUN: %run %t 2> %t.out
+// RUN: FileCheck %s --check-prefix=CHECK-GOOD < %t.log2.*
+
+// Check that the log_path option overrides the log name set via the
+// __memprof_profile_filename variable.
+// RUN: rm -f %t.log.*
+// RUN: %env_memprof_opts=log_path=%t.log %run %t 2> %t.out
+// RUN: FileCheck %s --check-prefix=CHECK-GOOD < %t.log.*
+
+#ifdef PROFILE_NAME_VAR
+#define xstr(s) str(s)
+#define str(s) #s
+char __memprof_profile_filename[] = xstr(PROFILE_NAME_VAR);
+#endif
+
 #include <stdlib.h>
 #include <string.h>
 int main(int argc, char **argv) {
diff --git a/flang/lib/Evaluate/intrinsics.cpp b/flang/lib/Evaluate/intrinsics.cpp
index 83fdc76c9bcd6..eaff9620a22dd 100644
--- a/flang/lib/Evaluate/intrinsics.cpp
+++ b/flang/lib/Evaluate/intrinsics.cpp
@@ -1959,6 +1959,19 @@ static bool CheckAssociated(SpecificCall &call,
                     *pointerSymbol);
               } else {
                 // object pointer and target
+                if (const Symbol * targetSymbol{GetLastSymbol(*targetExpr)}) {
+                  if (!(targetSymbol->attrs().test(semantics::Attr::POINTER) ||
+                          targetSymbol->attrs().test(
+                              semantics::Attr::TARGET))) {
+                    AttachDeclaration(
+                        messages.Say("TARGET= argument '%s' must have either "
+                                     "the POINTER or the TARGET "
+                                     "attribute"_err_en_US,
+                            targetName),
+                        *targetSymbol);
+                  }
+                }
+
                 if (const auto pointerType{pointerArg->GetType()}) {
                   if (const auto targetType{targetArg->GetType()}) {
                     ok = pointerType->IsTkCompatibleWith(*targetType);
diff --git a/flang/test/Semantics/associated.f90 b/flang/test/Semantics/associated.f90
index b78ccb017b162..641b7d97d78af 100644
--- a/flang/test/Semantics/associated.f90
+++ b/flang/test/Semantics/associated.f90
@@ -74,7 +74,6 @@ subroutine test()
     lVar = associated(intVar, intVar)
     !ERROR: POINTER= argument of ASSOCIATED() must be a POINTER
     lVar = associated(intAllocVar)
-    lVar = associated(intPointerVar1, intVar) !OK
     !ERROR: Arguments of ASSOCIATED() must be a POINTER and an optional valid target
     lVar = associated(intPointerVar1, targetRealVar)
     lVar = associated(intPointerVar1, targetIntVar1) !OK
@@ -82,6 +81,10 @@ subroutine test()
     lVar = associated(intPointerVar1, targetIntVar2)
     lVar = associated(intPointerVar1) !OK
     lVar = associated(intPointerVar1, intPointerVar2) !OK
+    !ERROR: In assignment to object pointer 'intpointervar1', the target 'intvar' is not an object with POINTER or TARGET attributes
+    intPointerVar1 => intVar
+    !ERROR: TARGET= argument 'intvar' must have either the POINTER or the TARGET attribute
+    lVar = associated(intPointerVar1, intVar)
 
     ! Procedure pointer tests
     intprocPointer1 => intProc !OK
diff --git a/libcxx/CMakeLists.txt b/libcxx/CMakeLists.txt
index 015a359bfb487..eb19646646f0f 100644
--- a/libcxx/CMakeLists.txt
+++ b/libcxx/CMakeLists.txt
@@ -397,9 +397,11 @@ endif ()
 # Configure System
 #===============================================================================
 
+# TODO: Other projects rely on the location where we generate the libc++ headers
+#       internally. We should get rid of these dependencies.
 if(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR AND NOT APPLE)
   set(LIBCXX_LIBRARY_DIR ${LLVM_LIBRARY_OUTPUT_INTDIR}/${LLVM_DEFAULT_TARGET_TRIPLE}/c++)
-  set(LIBCXX_HEADER_DIR ${LLVM_BINARY_DIR})
+  set(LIBCXX_GENERATED_INCLUDE_DIR "${LLVM_BINARY_DIR}/include/c++/v1")
   set(LIBCXX_INSTALL_LIBRARY_DIR lib${LLVM_LIBDIR_SUFFIX}/${LLVM_DEFAULT_TARGET_TRIPLE}/c++)
   if(LIBCXX_LIBDIR_SUBDIR)
     string(APPEND LIBCXX_LIBRARY_DIR /${LIBCXX_LIBDIR_SUBDIR})
@@ -407,11 +409,11 @@ if(LLVM_ENABLE_PER_TARGET_RUNTIME_DIR AND NOT APPLE)
   endif()
 elseif(LLVM_LIBRARY_OUTPUT_INTDIR)
   set(LIBCXX_LIBRARY_DIR ${LLVM_LIBRARY_OUTPUT_INTDIR})
-  set(LIBCXX_HEADER_DIR  ${LLVM_BINARY_DIR})
+  set(LIBCXX_GENERATED_INCLUDE_DIR  "${LLVM_BINARY_DIR}/include/c++/v1")
   set(LIBCXX_INSTALL_LIBRARY_DIR lib${LIBCXX_LIBDIR_SUFFIX})
 else()
   set(LIBCXX_LIBRARY_DIR ${CMAKE_BINARY_DIR}/lib${LIBCXX_LIBDIR_SUFFIX})
-  set(LIBCXX_HEADER_DIR  ${CMAKE_BINARY_DIR})
+  set(LIBCXX_GENERATED_INCLUDE_DIR  "${CMAKE_BINARY_DIR}/include/c++/v1")
   set(LIBCXX_INSTALL_LIBRARY_DIR lib${LIBCXX_LIBDIR_SUFFIX})
 endif()
 
diff --git a/libcxx/benchmarks/CMakeLists.txt b/libcxx/benchmarks/CMakeLists.txt
index 42d25c20c8115..12acffcbc98b7 100644
--- a/libcxx/benchmarks/CMakeLists.txt
+++ b/libcxx/benchmarks/CMakeLists.txt
@@ -8,7 +8,7 @@ include(CheckCXXCompilerFlag)
 set(BENCHMARK_LIBCXX_COMPILE_FLAGS
     -Wno-unused-command-line-argument
     -nostdinc++
-    -isystem ${LIBCXX_SOURCE_DIR}/include
+    -isystem "${LIBCXX_GENERATED_INCLUDE_DIR}"
     -L${LIBCXX_LIBRARY_DIR}
     -Wl,-rpath,${LIBCXX_LIBRARY_DIR}
     ${SANITIZER_FLAGS}
@@ -18,7 +18,6 @@ if (DEFINED LIBCXX_CXX_ABI_LIBRARY_PATH)
           -L${LIBCXX_CXX_ABI_LIBRARY_PATH}
           -Wl,-rpath,${LIBCXX_CXX_ABI_LIBRARY_PATH})
 endif()
-list(APPEND BENCHMARK_LIBCXX_COMPILE_FLAGS -include "${LIBCXX_BINARY_DIR}/__config_site")
 split_list(BENCHMARK_LIBCXX_COMPILE_FLAGS)
 
 ExternalProject_Add(google-benchmark-libcxx
diff --git a/libcxx/cmake/Modules/HandleLibCXXABI.cmake b/libcxx/cmake/Modules/HandleLibCXXABI.cmake
index c5aa26739e367..6f50853189a2f 100644
--- a/libcxx/cmake/Modules/HandleLibCXXABI.cmake
+++ b/libcxx/cmake/Modules/HandleLibCXXABI.cmake
@@ -52,8 +52,10 @@ macro(setup_abi_lib abidefines abishared abistatic abifiles abidirs)
             COMMENT "Copying C++ ABI header ${fpath}...")
         list(APPEND abilib_headers "${dst}")
 
-        if (LIBCXX_HEADER_DIR)
-          set(dst "${LIBCXX_HEADER_DIR}/include/c++/v1/${dstdir}/${fpath}")
+        if (LIBCXX_GENERATED_INCLUDE_DIR) # always true
+          # TODO: libc++ shouldn't be responsible for copying the libc++abi
+          #       headers into the right location.
+          set(dst "${LIBCXX_GENERATED_INCLUDE_DIR}/include/c++/v1/${dstdir}/${fpath}")
           add_custom_command(OUTPUT ${dst}
               DEPENDS ${src}
               COMMAND ${CMAKE_COMMAND} -E copy_if_different ${src} ${dst}
diff --git a/libcxx/docs/TestingLibcxx.rst b/libcxx/docs/TestingLibcxx.rst
index d42becfd5c3d8..ec017e23b147a 100644
--- a/libcxx/docs/TestingLibcxx.rst
+++ b/libcxx/docs/TestingLibcxx.rst
@@ -26,8 +26,8 @@ Usage
 
 After building libc++, you can run parts of the libc++ test suite by simply
 running ``llvm-lit`` on a specified test or directory. If you're unsure
-whether the required libraries have been built, you can use the
-`check-cxx-deps` target. For example:
+whether the required targets have been built, you can use the `check-cxx-deps`
+target to build them. For example:
 
 .. code-block:: bash
 
@@ -37,6 +37,12 @@ whether the required libraries have been built, you can use the
   $ <build>/bin/llvm-lit -sv libcxx/test/std/depr/depr.c.headers/stdlib_h.pass.cpp # Run a single test
   $ <build>/bin/llvm-lit -sv libcxx/test/std/atomics libcxx/test/std/threads # Test std::thread and std::atomic
 
+In the default configuration, the tests are built against headers that form a
+fake installation root of libc++. This installation root has to be updated when
+changes are made to the headers, so you should re-run the `check-cxx-deps` target
+before running the tests manually with `lit` when you make any sort of change,
+including to the headers.
+
 Sometimes you'll want to change the way LIT is running the tests. Custom options
 can be specified using the `--param=<name>=<val>` flag. The most common option
 you'll want to change is the standard dialect (ie -std=c++XX). By default the
diff --git a/libcxx/include/CMakeLists.txt b/libcxx/include/CMakeLists.txt
index 7c97db41bb73a..a63b565e4400f 100644
--- a/libcxx/include/CMakeLists.txt
+++ b/libcxx/include/CMakeLists.txt
@@ -2,6 +2,7 @@ set(files
   __bit_reference
   __bsd_locale_defaults.h
   __bsd_locale_fallbacks.h
+  __config
   __errc
   __debug
   __functional_03
@@ -184,62 +185,28 @@ if(LIBCXX_INSTALL_SUPPORT_HEADERS)
     )
 endif()
 
-configure_file("__config_site.in"
-               "${LIBCXX_BINARY_DIR}/__config_site"
-               @ONLY)
+if (LIBCXX_GENERATED_INCLUDE_DIR) # Always true
+  configure_file("__config_site.in" "${LIBCXX_GENERATED_INCLUDE_DIR}/__config_site" @ONLY)
 
-# Generate a custom __config header. The new header is created
-# by prepending __config_site to the current __config header.
-add_custom_command(OUTPUT ${LIBCXX_BINARY_DIR}/__generated_config
-  COMMAND ${Python3_EXECUTABLE} ${LIBCXX_SOURCE_DIR}/utils/cat_files.py
-    ${LIBCXX_BINARY_DIR}/__config_site
-    ${LIBCXX_SOURCE_DIR}/include/__config
-    -o ${LIBCXX_BINARY_DIR}/__generated_config
-  DEPENDS ${LIBCXX_SOURCE_DIR}/include/__config
-          ${LIBCXX_BINARY_DIR}/__config_site
-)
-# Add a target that executes the generation commands.
-add_custom_target(cxx-generated-config ALL
-  DEPENDS ${LIBCXX_BINARY_DIR}/__generated_config)
-
-if(LIBCXX_HEADER_DIR)
-  set(output_dir ${LIBCXX_HEADER_DIR}/include/c++/v1)
-
-  set(out_files)
+  set(_all_includes "${LIBCXX_GENERATED_INCLUDE_DIR}/__config_site")
   foreach(f ${files})
-    set(src ${CMAKE_CURRENT_SOURCE_DIR}/${f})
-    set(dst ${output_dir}/${f})
+    set(src "${CMAKE_CURRENT_SOURCE_DIR}/${f}")
+    set(dst "${LIBCXX_GENERATED_INCLUDE_DIR}/${f}")
     add_custom_command(OUTPUT ${dst}
       DEPENDS ${src}
       COMMAND ${CMAKE_COMMAND} -E copy_if_different ${src} ${dst}
       COMMENT "Copying CXX header ${f}")
-    list(APPEND out_files ${dst})
+    list(APPEND _all_includes "${dst}")
   endforeach()
-
-  # Copy the generated header as __config into build directory.
-  set(src ${LIBCXX_BINARY_DIR}/__generated_config)
-  set(dst ${output_dir}/__config)
-  add_custom_command(OUTPUT ${dst}
-      DEPENDS ${src} cxx-generated-config
-      COMMAND ${CMAKE_COMMAND} -E copy_if_different ${src} ${dst}
-      COMMENT "Copying CXX __config")
-  list(APPEND out_files ${dst})
-  add_custom_target(generate-cxx-headers DEPENDS ${out_files})
+  add_custom_target(generate-cxx-headers DEPENDS ${_all_includes})
 
   add_library(cxx-headers INTERFACE)
   add_dependencies(cxx-headers generate-cxx-headers ${LIBCXX_CXX_ABI_HEADER_TARGET})
   # TODO: Use target_include_directories once we figure out why that breaks the runtimes build
   if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "MSVC" OR "${CMAKE_CXX_SIMULATE_ID}" STREQUAL "MSVC")
-    target_compile_options(cxx-headers INTERFACE /I "${output_dir}")
-  else()
-    target_compile_options(cxx-headers INTERFACE -I "${output_dir}")
-  endif()
-
-  # Make sure the generated __config_site header is included when we build the library.
-  if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "MSVC" OR "${CMAKE_CXX_SIMULATE_ID}" STREQUAL "MSVC")
-    target_compile_options(cxx-headers INTERFACE /FI "${LIBCXX_BINARY_DIR}/__config_site")
+    target_compile_options(cxx-headers INTERFACE /I "${LIBCXX_GENERATED_INCLUDE_DIR}")
   else()
-    target_compile_options(cxx-headers INTERFACE -include "${LIBCXX_BINARY_DIR}/__config_site")
+    target_compile_options(cxx-headers INTERFACE -I "${LIBCXX_GENERATED_INCLUDE_DIR}")
   endif()
 else()
   add_library(cxx-headers INTERFACE)
@@ -255,16 +222,15 @@ if (LIBCXX_INSTALL_HEADERS)
     )
   endforeach()
 
-  # Install the generated header as __config.
-  install(FILES ${LIBCXX_BINARY_DIR}/__generated_config
+  # Install the generated __config_site.
+  install(FILES ${LIBCXX_GENERATED_INCLUDE_DIR}/__config_site
     DESTINATION ${LIBCXX_INSTALL_HEADER_PREFIX}include/c++/v1
     PERMISSIONS OWNER_READ OWNER_WRITE GROUP_READ WORLD_READ
-    RENAME __config
     COMPONENT cxx-headers)
 
   if (NOT CMAKE_CONFIGURATION_TYPES)
     add_custom_target(install-cxx-headers
-                      DEPENDS cxx-headers cxx-generated-config
+                      DEPENDS cxx-headers
                       COMMAND "${CMAKE_COMMAND}"
                               -DCMAKE_INSTALL_COMPONENT=cxx-headers
                               -P "${CMAKE_BINARY_DIR}/cmake_install.cmake")
diff --git a/libcxx/include/__config b/libcxx/include/__config
index 8d2ab2e51b115..d99e972004b86 100644
--- a/libcxx/include/__config
+++ b/libcxx/include/__config
@@ -10,6 +10,8 @@
 #ifndef _LIBCPP_CONFIG
 #define _LIBCPP_CONFIG
 
+#include <__config_site>
+
 #if defined(_MSC_VER) && !defined(__clang__)
 #  if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
 #    define _LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER
diff --git a/libcxx/test/configs/legacy.cfg.in b/libcxx/test/configs/legacy.cfg.in
index f0a4e8a73e09a..148ac3fc5a899 100644
--- a/libcxx/test/configs/legacy.cfg.in
+++ b/libcxx/test/configs/legacy.cfg.in
@@ -3,6 +3,7 @@
 import os
 import site
 
+config.cxx_headers              = "@LIBCXX_GENERATED_INCLUDE_DIR@"
 config.cxx_under_test           = "@CMAKE_CXX_COMPILER@"
 config.project_obj_root         = "@CMAKE_BINARY_DIR@"
 config.libcxx_src_root          = "@LIBCXX_SOURCE_DIR@"
diff --git a/libcxx/test/std/input.output/filesystems/fs.enum/enum.copy_options.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.enum/enum.copy_options.pass.cpp
index c5d73eb96c445..52903efb0d5f6 100644
--- a/libcxx/test/std/input.output/filesystems/fs.enum/enum.copy_options.pass.cpp
+++ b/libcxx/test/std/input.output/filesystems/fs.enum/enum.copy_options.pass.cpp
@@ -35,26 +35,27 @@ int main(int, char**) {
   typedef check_bitmask_type<E, E::skip_existing, E::update_existing> BitmaskTester;
   assert(BitmaskTester::check());
 
-  static_assert(
+  // The standard doesn't specify the numeric values of the enum.
+  LIBCPP_STATIC_ASSERT(
           E::none == ME(0),
         "Expected enumeration values do not match");
   // Option group for copy_file
-  static_assert(
+  LIBCPP_STATIC_ASSERT(
           E::skip_existing      == ME(1) &&
           E::overwrite_existing == ME(2) &&
           E::update_existing    == ME(4),
         "Expected enumeration values do not match");
   // Option group for copy on directories
-  static_assert(
+  LIBCPP_STATIC_ASSERT(
           E::recursive == ME(8),
         "Expected enumeration values do not match");
   // Option group for copy on symlinks
-  static_assert(
+  LIBCPP_STATIC_ASSERT(
           E::copy_symlinks == ME(16) &&
           E::skip_symlinks == ME(32),
         "Expected enumeration values do not match");
   // Option group for changing form of copy
-  static_assert(
+  LIBCPP_STATIC_ASSERT(
           E::directories_only    == ME(64) &&
           E::create_symlinks     == ME(128) &&
           E::create_hard_links   == ME(256),
diff --git a/libcxx/test/std/input.output/filesystems/fs.enum/enum.file_type.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.enum/enum.file_type.pass.cpp
index dd2a868f2d0aa..81e7227deaeb5 100644
--- a/libcxx/test/std/input.output/filesystems/fs.enum/enum.file_type.pass.cpp
+++ b/libcxx/test/std/input.output/filesystems/fs.enum/enum.file_type.pass.cpp
@@ -31,7 +31,8 @@ int main(int, char**) {
 
   LIBCPP_ONLY(static_assert(std::is_same<UT, signed char>::value, "")); // Implementation detail
 
-  static_assert(
+  // The standard doesn't specify the numeric values of the enum.
+  LIBCPP_STATIC_ASSERT(
           E::none == ME(0) &&
           E::not_found == ME(-1) &&
           E::regular == ME(1) &&
diff --git a/libcxx/test/std/input.output/filesystems/fs.enum/enum.path.format.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.enum/enum.path.format.pass.cpp
index fc9fa1936119d..f328167833371 100644
--- a/libcxx/test/std/input.output/filesystems/fs.enum/enum.path.format.pass.cpp
+++ b/libcxx/test/std/input.output/filesystems/fs.enum/enum.path.format.pass.cpp
@@ -25,7 +25,7 @@ int main(int, char**) {
 
   // Check that E is a scoped enum by checking for conversions.
   typedef std::underlying_type<E>::type UT;
-  static_assert(!std::is_convertible<E, UT>::value, "");
+  LIBCPP_STATIC_ASSERT(!std::is_convertible<E, UT>::value, "");
 
   LIBCPP_ONLY(static_assert(std::is_same<UT, unsigned char>::value, "")); // Implementation detail
 
diff --git a/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type.pass.cpp
index 09b8aed5a5c8e..afdf0ec15b943 100644
--- a/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type.pass.cpp
+++ b/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type.pass.cpp
@@ -32,16 +32,8 @@ void test_trivial_clock() {
   ((void)odr_use);
 }
 
-void test_time_point_resolution_and_range() {
-  using namespace fs;
-  using Dur = file_time_type::duration;
-  using Period = Dur::period;
-  ASSERT_SAME_TYPE(Period, std::nano);
-}
-
 int main(int, char**) {
   test_trivial_clock();
-  test_time_point_resolution_and_range();
 
   return 0;
 }
diff --git a/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type_resolution.compile.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type_resolution.compile.pass.cpp
new file mode 100644
index 0000000000000..965e7a4620b87
--- /dev/null
+++ b/libcxx/test/std/input.output/filesystems/fs.filesystem.synopsis/file_time_type_resolution.compile.pass.cpp
@@ -0,0 +1,27 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+// UNSUPPORTED: c++03, windows
+
+// MS STL and libstdc++ use the native windows file timestamp resolution,
+// with 100 ns resolution.
+
+// <filesystem>
+
+// typedef TrivialClock file_time_type;
+
+#include "filesystem_include.h"
+#include <chrono>
+#include <type_traits>
+
+#include "test_macros.h"
+
+using namespace fs;
+using Dur = file_time_type::duration;
+using Period = Dur::period;
+ASSERT_SAME_TYPE(Period, std::nano);
diff --git a/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.space/space.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.space/space.pass.cpp
index f607b44e544de..c0317966d4fe8 100644
--- a/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.space/space.pass.cpp
+++ b/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.space/space.pass.cpp
@@ -14,7 +14,6 @@
 // space_info space(const path& p, error_code& ec) noexcept;
 
 #include "filesystem_include.h"
-#include <sys/statvfs.h>
 
 #include "test_macros.h"
 #include "rapid-cxx-test.h"
@@ -85,23 +84,12 @@ TEST_CASE(basic_space_test)
     // All the test cases should reside on the same filesystem and therefore
     // should have the same expected result. Compute this expected result
     // one and check that it looks semi-sane.
-    struct statvfs expect;
-    TEST_REQUIRE(::statvfs(static_env.Dir.string().c_str(), &expect) != -1);
-    TEST_CHECK(expect.f_bavail > 0);
-    TEST_CHECK(expect.f_bfree > 0);
-    TEST_CHECK(expect.f_bsize > 0);
-    TEST_CHECK(expect.f_blocks > 0);
-    TEST_REQUIRE(expect.f_frsize > 0);
-    auto do_mult = [&](std::uintmax_t val) {
-        std::uintmax_t fsize = expect.f_frsize;
-        std::uintmax_t new_val = val * fsize;
-        TEST_CHECK(new_val / fsize == val); // Test for overflow
-        return new_val;
-    };
     const std::uintmax_t bad_value = static_cast<std::uintmax_t>(-1);
-    const std::uintmax_t expect_capacity = do_mult(expect.f_blocks);
-    const std::uintmax_t expect_free = do_mult(expect.f_bfree);
-    const std::uintmax_t expect_avail = do_mult(expect.f_bavail);
+    std::uintmax_t expect_capacity;
+    std::uintmax_t expect_free;
+    std::uintmax_t expect_avail;
+    TEST_REQUIRE(utils::space(static_env.Dir.string(), expect_capacity,
+                              expect_free, expect_avail));
 
     // Other processes running on the operating system may have changed
     // the amount of space available. Check that these are within tolerances.
diff --git a/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.symlink_status/symlink_status.pass.cpp b/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.symlink_status/symlink_status.pass.cpp
index 8d320d750bc96..d2116779ba1fe 100644
--- a/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.symlink_status/symlink_status.pass.cpp
+++ b/libcxx/test/std/input.output/filesystems/fs.op.funcs/fs.op.symlink_status/symlink_status.pass.cpp
@@ -83,7 +83,7 @@ TEST_CASE(test_symlink_status_cannot_resolve)
 #ifndef TEST_HAS_NO_EXCEPTIONS
         { // test throwing case
             try {
-                symlink_status(p);
+                (void)symlink_status(p);
             } catch (filesystem_error const& err) {
                 TEST_CHECK(err.path1() == p);
                 TEST_CHECK(err.path2() == "");
diff --git a/libcxx/test/support/filesystem_test_helper.h b/libcxx/test/support/filesystem_test_helper.h
index 704e2b26a0b48..ac2014fa34d1c 100644
--- a/libcxx/test/support/filesystem_test_helper.h
+++ b/libcxx/test/support/filesystem_test_helper.h
@@ -6,6 +6,7 @@
 #include <sys/stat.h> // for stat, mkdir, mkfifo
 #ifndef _WIN32
 #include <unistd.h> // for ftruncate, link, symlink, getcwd, chdir
+#include <sys/statvfs.h>
 #else
 #include <io.h>
 #include <direct.h>
@@ -52,6 +53,21 @@ namespace utils {
     inline int unsetenv(const char *var) {
         return ::_putenv((std::string(var) + "=").c_str());
     }
+    inline bool space(std::string path, std::uintmax_t &capacity,
+                      std::uintmax_t &free, std::uintmax_t &avail) {
+        ULARGE_INTEGER FreeBytesAvailableToCaller, TotalNumberOfBytes,
+                       TotalNumberOfFreeBytes;
+        if (!GetDiskFreeSpaceExA(path.c_str(), &FreeBytesAvailableToCaller,
+                                 &TotalNumberOfBytes, &TotalNumberOfFreeBytes))
+          return false;
+        capacity = TotalNumberOfBytes.QuadPart;
+        free = TotalNumberOfFreeBytes.QuadPart;
+        avail = FreeBytesAvailableToCaller.QuadPart;
+        assert(capacity > 0);
+        assert(free > 0);
+        assert(avail > 0);
+        return true;
+    }
 #else
     using ::mkdir;
     using ::ftruncate;
@@ -63,6 +79,27 @@ namespace utils {
     inline int unsetenv(const char *var) {
         return ::unsetenv(var);
     }
+    inline bool space(std::string path, std::uintmax_t &capacity,
+                      std::uintmax_t &free, std::uintmax_t &avail) {
+        struct statvfs expect;
+        if (::statvfs(path.c_str(), &expect) == -1)
+          return false;
+        assert(expect.f_bavail > 0);
+        assert(expect.f_bfree > 0);
+        assert(expect.f_bsize > 0);
+        assert(expect.f_blocks > 0);
+        assert(expect.f_frsize > 0);
+        auto do_mult = [&](std::uintmax_t val) {
+            std::uintmax_t fsize = expect.f_frsize;
+            std::uintmax_t new_val = val * fsize;
+            assert(new_val / fsize == val); // Test for overflow
+            return new_val;
+        };
+        capacity = do_mult(expect.f_blocks);
+        free = do_mult(expect.f_bfree);
+        avail = do_mult(expect.f_bavail);
+        return true;
+    }
 #endif
 
     inline std::string getcwd() {
diff --git a/libcxx/utils/ci/buildkite-pipeline-premerge.sh b/libcxx/utils/ci/buildkite-pipeline-premerge.sh
new file mode 100755
index 0000000000000..ba6b5ae6b1aca
--- /dev/null
+++ b/libcxx/utils/ci/buildkite-pipeline-premerge.sh
@@ -0,0 +1,42 @@
+#!/usr/bin/env bash
+#===----------------------------------------------------------------------===##
+#
+# Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+# See https://llvm.org/LICENSE.txt for license information.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+#
+#===----------------------------------------------------------------------===##
+
+#
+# This file generates a Buildkite pipeline that triggers the libc++ CI
+# job(s) if needed.
+# See https://buildkite.com/docs/agent/v3/cli-pipeline#pipeline-format.
+#
+# Invoked by CI on pre-merge check for a commit.
+#
+
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+
+if ! git diff --name-only HEAD~1 | grep -q -E "libcxx/|libcxxabi/"; then
+  # libcxx is not affected
+  exit 0
+fi
+
+reviewID="$(git log --format=%B -n 1 | sed -nE 's/^Review-ID:[[:space:]]*(.+)$/\1/p')"
+if [[ "${reviewID}" != "" ]]; then
+  buildMessage="https://llvm.org/${reviewID}"
+else
+  buildMessage="Push to branch ${BUILDKITE_BRANCH}"
+fi
+
+
+cat <<EOF
+steps:
+  - trigger: "libcxx-ci"
+    build:
+      message: "${buildMessage}"
+      commit: "${BUILDKITE_COMMIT}"
+      branch: "${BUILDKITE_BRANCH}"
+EOF
+
+
diff --git a/libcxx/utils/ci/buildkite-pipeline-snapshot.sh b/libcxx/utils/ci/buildkite-pipeline-snapshot.sh
new file mode 100755
index 0000000000000..b48aad791c78a
--- /dev/null
+++ b/libcxx/utils/ci/buildkite-pipeline-snapshot.sh
@@ -0,0 +1,25 @@
+#!/usr/bin/env bash
+#===----------------------------------------------------------------------===##
+#
+# Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+# See https://llvm.org/LICENSE.txt for license information.
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+#
+#===----------------------------------------------------------------------===##
+
+#
+# This file generates a Buildkite pipeline that triggers the libc++ CI jobs.
+# See https://buildkite.com/docs/agent/v3/cli-pipeline#pipeline-format.
+#
+# Invoked by CI on full builds.
+#
+
+DIR="$( cd "$( dirname "${BASH_SOURCE[0]}" )" >/dev/null 2>&1 && pwd )"
+cat <<EOF
+steps:
+  - trigger: "libcxx-ci"
+    build:
+      message: "${BUILDKITE_MESSAGE}"
+      commit: "${BUILDKITE_COMMIT}"
+      branch: "${BUILDKITE_BRANCH}"
+EOF
diff --git a/libcxx/utils/ci/buildkite-pipeline.yml b/libcxx/utils/ci/buildkite-pipeline.yml
index 6d416a36cba25..76cf1c4a76669 100644
--- a/libcxx/utils/ci/buildkite-pipeline.yml
+++ b/libcxx/utils/ci/buildkite-pipeline.yml
@@ -16,86 +16,119 @@
 
 steps:
   - label: "C++03"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-cxx03 | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-cxx03"
+    artifact_paths:
+      - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "C++11"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-cxx11 | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-cxx11"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "C++14"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-cxx14 | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-cxx14"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "C++17"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-cxx17 | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-cxx17"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "C++20"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-cxx2a | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-cxx2a"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "-fno-exceptions"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-noexceptions | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-noexceptions"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "GCC/C++20"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-gcc | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-gcc"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "ASAN"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-asan | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-asan"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "TSAN"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-tsan | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-tsan"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "UBSAN"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-ubsan | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-ubsan"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "With LLVM's libunwind"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-with_llvm_unwinder | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-with_llvm_unwinder"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "Single-threaded"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-singlethreaded | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-singlethreaded"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "No debug mode"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-nodebug | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-nodebug"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-builders"
 
   - label: "No random device"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-no-random_device | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-no-random_device"
     agents:
       queue: "libcxx-builders"
 
   - label: "MacOS C++20"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh generic-cxx2a | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh generic-cxx2a"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-macos-builders"
 
   # Build with the configuration we use to generate libc++.dylib on Apple platforms
   - label: "Apple system"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh x86_64-apple-system | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh x86_64-apple-system"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-macos-builders"
+
   - label: "Apple system -fno-exceptions"
-    command: "set -o pipefail && libcxx/utils/ci/run-buildbot.sh x86_64-apple-system-noexceptions | libcxx/utils/ci/phabricator-report"
+    command: "libcxx/utils/ci/run-buildbot.sh x86_64-apple-system-noexceptions"
+    artifact_paths:
+    - "**/test-results.xml"
     agents:
       queue: "libcxx-macos-builders"
diff --git a/libcxx/utils/ci/run-buildbot.sh b/libcxx/utils/ci/run-buildbot.sh
index 87a1f0f1d6228..df166b9ed40dd 100755
--- a/libcxx/utils/ci/run-buildbot.sh
+++ b/libcxx/utils/ci/run-buildbot.sh
@@ -12,52 +12,54 @@ set -ex
 BUILDER="${1}"
 MONOREPO_ROOT="$(git rev-parse --show-toplevel)"
 BUILD_DIR="${MONOREPO_ROOT}/build/${BUILDER}"
+INSTALL_DIR="${MONOREPO_ROOT}/build/${BUILDER}/install"
 
 args=()
 args+=("-DLLVM_ENABLE_PROJECTS=libcxx;libunwind;libcxxabi")
+args+=("-DCMAKE_INSTALL_PREFIX=${INSTALL_DIR}")
 args+=("-DLIBCXX_CXX_ABI=libcxxabi")
 
 case "${BUILDER}" in
 generic-cxx03)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-cxx03.cmake")
 ;;
 generic-cxx11)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-cxx11.cmake")
 ;;
 generic-cxx14)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-cxx14.cmake")
 ;;
 generic-cxx17)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-cxx17.cmake")
 ;;
 generic-cxx2a)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-cxx2a.cmake")
 ;;
 generic-noexceptions)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-noexceptions.cmake")
 ;;
 generic-32bit)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-32bits.cmake")
 ;;
 generic-gcc)
@@ -65,66 +67,66 @@ generic-gcc)
     export CXX=g++
     # FIXME: Re-enable experimental testing on GCC. GCC cares about the order
     #        in which we link -lc++experimental, which causes issues.
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --param enable_experimental=False")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --param enable_experimental=False --xunit-xml-output test-results.xml")
 ;;
 generic-asan)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-asan.cmake")
 ;;
 generic-msan)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-msan.cmake")
 ;;
 generic-tsan)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-tsan.cmake")
 ;;
 generic-ubsan)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-ubsan.cmake")
 ;;
 generic-with_llvm_unwinder)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-DLIBCXXABI_USE_LLVM_UNWINDER=ON")
 ;;
 generic-singlethreaded)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-singlethreaded.cmake")
 ;;
 generic-nodebug)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-nodebug.cmake")
 ;;
 generic-no-random_device)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Generic-no-random_device.cmake")
 ;;
 x86_64-apple-system)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Apple.cmake")
 ;;
 x86_64-apple-system-noexceptions)
     export CC=clang
     export CXX=clang++
-    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported")
+    args+=("-DLLVM_LIT_ARGS=-sv --show-unsupported --xunit-xml-output test-results.xml")
     args+=("-C${MONOREPO_ROOT}/libcxx/cmake/caches/Apple.cmake")
     args+=("-DLIBCXX_ENABLE_EXCEPTIONS=OFF")
     args+=("-DLIBCXXABI_ENABLE_EXCEPTIONS=OFF")
@@ -148,5 +150,8 @@ ninja -C "${BUILD_DIR}" check-cxx
 echo "+++ Running the libc++abi tests"
 ninja -C "${BUILD_DIR}" check-cxxabi
 
+echo "+++ Installing libc++ and libc++abi to a fake location"
+ninja -C "${BUILD_DIR}" install-cxx install-cxxabi
+
 # echo "+++ Running the libc++ benchmarks"
 # ninja -C "${BUILD_DIR}" check-cxx-benchmarks
diff --git a/libcxx/utils/libcxx/test/config.py b/libcxx/utils/libcxx/test/config.py
index 0d21aa17afd27..c5050d1b6b9a7 100644
--- a/libcxx/utils/libcxx/test/config.py
+++ b/libcxx/utils/libcxx/test/config.py
@@ -331,7 +331,6 @@ def configure_default_compile_flags(self):
 
     def configure_compile_flags_header_includes(self):
         support_path = os.path.join(self.libcxx_src_root, 'test', 'support')
-        self.configure_config_site_header()
         if self.cxx_stdlib_under_test != 'libstdc++' and \
            not self.target_info.is_windows():
             self.cxx.compile_flags += [
@@ -348,16 +347,12 @@ def configure_compile_flags_header_includes(self):
                                          'set_windows_crt_report_mode.h')
             ]
         cxx_headers = self.get_lit_conf('cxx_headers')
-        if cxx_headers == '' or (cxx_headers is None
-                                 and self.cxx_stdlib_under_test != 'libc++'):
+        if cxx_headers is None and self.cxx_stdlib_under_test != 'libc++':
             self.lit_config.note('using the system cxx headers')
             return
         self.cxx.compile_flags += ['-nostdinc++']
-        if cxx_headers is None:
-            cxx_headers = os.path.join(self.libcxx_src_root, 'include')
         if not os.path.isdir(cxx_headers):
-            self.lit_config.fatal("cxx_headers='%s' is not a directory."
-                                  % cxx_headers)
+            self.lit_config.fatal("cxx_headers='{}' is not a directory.".format(cxx_headers))
         self.cxx.compile_flags += ['-I' + cxx_headers]
         if self.libcxx_obj_root is not None:
             cxxabi_headers = os.path.join(self.libcxx_obj_root, 'include',
@@ -365,16 +360,6 @@ def configure_compile_flags_header_includes(self):
             if os.path.isdir(cxxabi_headers):
                 self.cxx.compile_flags += ['-I' + cxxabi_headers]
 
-    def configure_config_site_header(self):
-        # Check for a possible __config_site in the build directory. We
-        # use this if it exists.
-        if self.libcxx_obj_root is None:
-            return
-        config_site_header = os.path.join(self.libcxx_obj_root, '__config_site')
-        if not os.path.isfile(config_site_header):
-            return
-        self.cxx.compile_flags += ['-include', config_site_header]
-
     def configure_link_flags(self):
         # Configure library path
         self.configure_link_flags_cxx_library_path()
diff --git a/libcxxabi/CMakeLists.txt b/libcxxabi/CMakeLists.txt
index 3f37dbf4cd36f..f0593bc2a076b 100644
--- a/libcxxabi/CMakeLists.txt
+++ b/libcxxabi/CMakeLists.txt
@@ -138,10 +138,6 @@ if (NOT LIBCXXABI_ENABLE_SHARED AND NOT LIBCXXABI_ENABLE_STATIC)
   message(FATAL_ERROR "libc++abi must be built as either a shared or static library.")
 endif()
 
-set(LIBCXXABI_LIBCXX_INCLUDES "${LIBCXXABI_LIBCXX_PATH}/include" CACHE PATH
-    "Specify path to libc++ includes.")
-message(STATUS "Libc++abi will be using libc++ includes from ${LIBCXXABI_LIBCXX_INCLUDES}")
-
 option(LIBCXXABI_HERMETIC_STATIC_LIBRARY
   "Do not export any symbols from the static library." OFF)
 
diff --git a/libcxxabi/include/__cxxabi_config.h b/libcxxabi/include/__cxxabi_config.h
index 98f55b2a256d9..cffedb88df543 100644
--- a/libcxxabi/include/__cxxabi_config.h
+++ b/libcxxabi/include/__cxxabi_config.h
@@ -18,6 +18,19 @@
 #define __has_attribute(_attribute_) 0
 #endif
 
+#if defined(__clang__)
+#  define _LIBCXXABI_COMPILER_CLANG
+#  ifndef __apple_build_version__
+#    define _LIBCXXABI_CLANG_VER (__clang_major__ * 100 + __clang_minor__)
+#  endif
+#elif defined(__GNUC__)
+#  define _LIBCXXABI_COMPILER_GCC
+#elif defined(_MSC_VER)
+#  define _LIBCXXABI_COMPILER_MSVC
+#elif defined(__IBMCPP__)
+#  define _LIBCXXABI_COMPILER_IBM
+#endif
+
 #if defined(_WIN32)
  #if defined(_LIBCXXABI_DISABLE_VISIBILITY_ANNOTATIONS)
   #define _LIBCXXABI_HIDDEN
@@ -53,7 +66,7 @@
  #endif
 #endif
 
-#if defined(_WIN32)
+#if defined(_LIBCXXABI_COMPILER_MSVC)
 #define _LIBCXXABI_WEAK
 #else
 #define _LIBCXXABI_WEAK __attribute__((__weak__))
diff --git a/libcxxabi/src/CMakeLists.txt b/libcxxabi/src/CMakeLists.txt
index e9e454082a054..8d4d84c225e23 100644
--- a/libcxxabi/src/CMakeLists.txt
+++ b/libcxxabi/src/CMakeLists.txt
@@ -55,8 +55,6 @@ if (MSVC_IDE OR XCODE)
   endif()
 endif()
 
-include_directories("${LIBCXXABI_LIBCXX_INCLUDES}")
-
 if (LIBCXXABI_HAS_CXA_THREAD_ATEXIT_IMPL)
   add_definitions(-DHAVE___CXA_THREAD_ATEXIT_IMPL)
 endif()
@@ -168,7 +166,7 @@ if (LIBCXXABI_ENABLE_SHARED)
   if(COMMAND llvm_setup_rpath)
     llvm_setup_rpath(cxxabi_shared)
   endif()
-  target_link_libraries(cxxabi_shared PRIVATE ${LIBCXXABI_SHARED_LIBRARIES} ${LIBCXXABI_LIBRARIES})
+  target_link_libraries(cxxabi_shared PRIVATE cxx-headers ${LIBCXXABI_SHARED_LIBRARIES} ${LIBCXXABI_LIBRARIES})
   if (TARGET pstl::ParallelSTL)
     target_link_libraries(cxxabi_shared PUBLIC pstl::ParallelSTL)
   endif()
@@ -233,7 +231,7 @@ endif()
 # Build the static library.
 if (LIBCXXABI_ENABLE_STATIC)
   add_library(cxxabi_static STATIC ${LIBCXXABI_SOURCES} ${LIBCXXABI_HEADERS})
-  target_link_libraries(cxxabi_static PRIVATE ${LIBCXXABI_STATIC_LIBRARIES} ${LIBCXXABI_LIBRARIES})
+  target_link_libraries(cxxabi_static PRIVATE cxx-headers ${LIBCXXABI_STATIC_LIBRARIES} ${LIBCXXABI_LIBRARIES})
   if (TARGET pstl::ParallelSTL)
     target_link_libraries(cxxabi_static PUBLIC pstl::ParallelSTL)
   endif()
diff --git a/libcxxabi/test/libcxxabi/test/config.py b/libcxxabi/test/libcxxabi/test/config.py
index 45fb0f5d7afcb..e7c019688f1a3 100644
--- a/libcxxabi/test/libcxxabi/test/config.py
+++ b/libcxxabi/test/libcxxabi/test/config.py
@@ -56,10 +56,9 @@ def configure_compile_flags(self):
         super(Configuration, self).configure_compile_flags()
 
     def configure_compile_flags_header_includes(self):
-        self.configure_config_site_header()
         cxx_headers = self.get_lit_conf(
             'cxx_headers',
-            os.path.join(self.libcxx_src_root, '/include'))
+            os.path.join(self.project_obj_root, 'include', 'c++', 'v1'))
         if cxx_headers == '':
             self.lit_config.note('using the systems c++ headers')
         else:
diff --git a/libcxxabi/test/lit.site.cfg.in b/libcxxabi/test/lit.site.cfg.in
index 87f955e321610..a6d43c6e2da19 100644
--- a/libcxxabi/test/lit.site.cfg.in
+++ b/libcxxabi/test/lit.site.cfg.in
@@ -9,7 +9,6 @@ config.libcxxabi_src_root       = "@LIBCXXABI_SOURCE_DIR@"
 config.libcxxabi_obj_root       = "@LIBCXXABI_BINARY_DIR@"
 config.abi_library_path         = "@LIBCXXABI_LIBRARY_DIR@"
 config.libcxx_src_root          = "@LIBCXXABI_LIBCXX_PATH@"
-config.cxx_headers              = "@LIBCXXABI_LIBCXX_INCLUDES@"
 config.libunwind_headers        = "@LIBCXXABI_LIBUNWIND_INCLUDES_INTERNAL@"
 config.cxx_library_root         = "@LIBCXXABI_LIBCXX_LIBRARY_PATH@"
 config.llvm_unwinder            = @LIBCXXABI_USE_LLVM_UNWINDER@
diff --git a/libcxxabi/test/test_exception_storage.pass.cpp b/libcxxabi/test/test_exception_storage.pass.cpp
index 5c29c4524a41d..bf423667b7734 100644
--- a/libcxxabi/test/test_exception_storage.pass.cpp
+++ b/libcxxabi/test/test_exception_storage.pass.cpp
@@ -29,7 +29,9 @@ void *thread_code (void *parm) {
         std::printf("Got different globals!\n");
 
     *result = (size_t) glob1;
+#ifndef _LIBCXXABI_HAS_NO_THREADS
     sleep ( 1 );
+#endif
     return parm;
 }
 
diff --git a/libunwind/test/libunwind/test/config.py b/libunwind/test/libunwind/test/config.py
index 977f9a0fb3f93..183c48f6a8016 100644
--- a/libunwind/test/libunwind/test/config.py
+++ b/libunwind/test/libunwind/test/config.py
@@ -49,8 +49,6 @@ def configure_compile_flags(self):
         super(Configuration, self).configure_compile_flags()
 
     def configure_compile_flags_header_includes(self):
-        self.configure_config_site_header()
-
         libunwind_headers = self.get_lit_conf(
             'libunwind_headers',
             os.path.join(self.libunwind_src_root, 'include'))
diff --git a/lld/CMakeLists.txt b/lld/CMakeLists.txt
index 8b8c7178c616c..82b4b9b9b1981 100644
--- a/lld/CMakeLists.txt
+++ b/lld/CMakeLists.txt
@@ -195,7 +195,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     DESTINATION include
     FILES_MATCHING
     PATTERN "*.h"
-    PATTERN ".svn" EXCLUDE
     )
 endif()
 
diff --git a/lld/ELF/MarkLive.cpp b/lld/ELF/MarkLive.cpp
index af6c08c215816..fe5d32c80bfea 100644
--- a/lld/ELF/MarkLive.cpp
+++ b/lld/ELF/MarkLive.cpp
@@ -339,16 +339,16 @@ template <class ELFT> void elf::markLive() {
 
   // Otherwise, do mark-sweep GC.
   //
-  // The -gc-sections option works only for SHF_ALLOC sections
-  // (sections that are memory-mapped at runtime). So we can
-  // unconditionally make non-SHF_ALLOC sections alive except
-  // SHF_LINK_ORDER and SHT_REL/SHT_RELA sections.
+  // The -gc-sections option works only for SHF_ALLOC sections (sections that
+  // are memory-mapped at runtime). So we can unconditionally make non-SHF_ALLOC
+  // sections alive except SHF_LINK_ORDER, SHT_REL/SHT_RELA sections, and
+  // sections in a group.
   //
   // Usually, non-SHF_ALLOC sections are not removed even if they are
-  // unreachable through relocations because reachability is not
-  // a good signal whether they are garbage or not (e.g. there is
-  // usually no section referring to a .comment section, but we
-  // want to keep it.).
+  // unreachable through relocations because reachability is not a good signal
+  // whether they are garbage or not (e.g. there is usually no section referring
+  // to a .comment section, but we want to keep it.) When a non-SHF_ALLOC
+  // section is retained, we also retain sections dependent on it.
   //
   // Note on SHF_LINK_ORDER: Such sections contain metadata and they
   // have a reverse dependency on the InputSection they are linked with.
@@ -370,8 +370,11 @@ template <class ELFT> void elf::markLive() {
     bool isLinkOrder = (sec->flags & SHF_LINK_ORDER);
     bool isRel = (sec->type == SHT_REL || sec->type == SHT_RELA);
 
-    if (!isAlloc && !isLinkOrder && !isRel && !sec->nextInSectionGroup)
+    if (!isAlloc && !isLinkOrder && !isRel && !sec->nextInSectionGroup) {
       sec->markLive();
+      for (InputSection *isec : sec->dependentSections)
+        isec->markLive();
+    }
   }
 
   // Follow the graph to mark all live sections.
diff --git a/lld/test/ELF/gc-sections-linkorder2.s b/lld/test/ELF/gc-sections-linkorder2.s
index b22bc589f56ea..2024972b3937e 100644
--- a/lld/test/ELF/gc-sections-linkorder2.s
+++ b/lld/test/ELF/gc-sections-linkorder2.s
@@ -14,3 +14,9 @@ _start:
 .quad 0
 .section .zed,"ao",@progbits,.foo
 .quad 0
+
+.section .nonalloc
+.quad 0
+
+.section .nonalloc_linkorder,"o",@progbits,.nonalloc
+.quad 0
diff --git a/lldb/cmake/modules/FindPythonAndSwig.cmake b/lldb/cmake/modules/FindPythonAndSwig.cmake
index de274ede5dbf6..dcbc386b70ddc 100644
--- a/lldb/cmake/modules/FindPythonAndSwig.cmake
+++ b/lldb/cmake/modules/FindPythonAndSwig.cmake
@@ -48,7 +48,7 @@ else()
   get_property(MULTI_CONFIG GLOBAL PROPERTY GENERATOR_IS_MULTI_CONFIG)
   if ("${Python3_VERSION}" VERSION_GREATER_EQUAL "3.7" AND
       "${SWIG_VERSION}" VERSION_LESS "4.0" AND WIN32 AND (
-      ${MULTI_CONFIG} OR (${uppercase_CMAKE_BUILD_TYPE} STREQUAL "DEBUG")))
+      ${MULTI_CONFIG} OR (uppercase_CMAKE_BUILD_TYPE STREQUAL "DEBUG")))
     # Technically this can happen with non-Windows builds too, but we are not
     # able to detect whether Python was built with assertions, and only Windows
     # has the requirement that Debug LLDB must use Debug Python.
diff --git a/lldb/cmake/modules/LLDBConfig.cmake b/lldb/cmake/modules/LLDBConfig.cmake
index 5fbc89892c73c..2fdf1502d0559 100644
--- a/lldb/cmake/modules/LLDBConfig.cmake
+++ b/lldb/cmake/modules/LLDBConfig.cmake
@@ -228,7 +228,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     DESTINATION include
     FILES_MATCHING
     PATTERN "*.h"
-    PATTERN ".svn" EXCLUDE
     PATTERN ".cmake" EXCLUDE
     )
 
@@ -237,7 +236,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     DESTINATION include
     FILES_MATCHING
     PATTERN "*.h"
-    PATTERN ".svn" EXCLUDE
     PATTERN ".cmake" EXCLUDE
     )
 
diff --git a/lldb/include/lldb/Core/Disassembler.h b/lldb/include/lldb/Core/Disassembler.h
index d3b903b83c191..c38f1f7975d56 100644
--- a/lldb/include/lldb/Core/Disassembler.h
+++ b/lldb/include/lldb/Core/Disassembler.h
@@ -48,6 +48,7 @@ class DataExtractor;
 class Debugger;
 class Disassembler;
 class Module;
+class StackFrame;
 class Stream;
 class SymbolContext;
 class SymbolContextList;
@@ -404,11 +405,8 @@ class Disassembler : public std::enable_shared_from_this<Disassembler>,
                           uint32_t num_mixed_context_lines, uint32_t options,
                           Stream &strm);
 
-  static bool
-  Disassemble(Debugger &debugger, const ArchSpec &arch, const char *plugin_name,
-              const char *flavor, const ExecutionContext &exe_ctx,
-              uint32_t num_instructions, bool mixed_source_and_assembly,
-              uint32_t num_mixed_context_lines, uint32_t options, Stream &strm);
+  static bool Disassemble(Debugger &debugger, const ArchSpec &arch,
+                          StackFrame &frame, Stream &strm);
 
   // Constructors and Destructors
   Disassembler(const ArchSpec &arch, const char *flavor);
diff --git a/lldb/include/lldb/Utility/Args.h b/lldb/include/lldb/Utility/Args.h
index 2cce7d0c697c7..b93677b539350 100644
--- a/lldb/include/lldb/Utility/Args.h
+++ b/lldb/include/lldb/Utility/Args.h
@@ -66,6 +66,7 @@ class Args {
 
   Args(const Args &rhs);
   explicit Args(const StringList &list);
+  explicit Args(llvm::ArrayRef<llvm::StringRef> args);
 
   Args &operator=(const Args &rhs);
 
@@ -262,9 +263,8 @@ class Args {
 
   static uint32_t StringToGenericRegister(llvm::StringRef s);
 
-  static const char *GetShellSafeArgument(const FileSpec &shell,
-                                          const char *unsafe_arg,
-                                          std::string &safe_arg);
+  static std::string GetShellSafeArgument(const FileSpec &shell,
+                                          llvm::StringRef unsafe_arg);
 
   // EncodeEscapeSequences will change the textual representation of common
   // escape sequences like "\n" (two characters) into a single '\n'. It does
diff --git a/lldb/packages/Python/lldbsuite/test/builders/builder.py b/lldb/packages/Python/lldbsuite/test/builders/builder.py
index fbfa86700e227..6c9584224f4a9 100644
--- a/lldb/packages/Python/lldbsuite/test/builders/builder.py
+++ b/lldb/packages/Python/lldbsuite/test/builders/builder.py
@@ -93,11 +93,7 @@ def getArchSpec(self, architecture):
         Helper function to return the key-value string to specify the architecture
         used for the make system.
         """
-        arch = architecture if architecture else None
-        if not arch and configuration.arch:
-            arch = configuration.arch
-
-        return ("ARCH=" + arch) if arch else ""
+        return ("ARCH=" + architecture) if architecture else ""
 
     def getCCSpec(self, compiler):
         """
diff --git a/lldb/packages/Python/lldbsuite/test/lldbtest.py b/lldb/packages/Python/lldbsuite/test/lldbtest.py
index 69da3914f1f22..f469ce86f26e3 100644
--- a/lldb/packages/Python/lldbsuite/test/lldbtest.py
+++ b/lldb/packages/Python/lldbsuite/test/lldbtest.py
@@ -2606,6 +2606,9 @@ def build(
         """Platform specific way to build the default binaries."""
         module = builder_module()
 
+        if not architecture and configuration.arch:
+            architecture = configuration.arch
+
         dictionary = lldbplatformutil.finalize_build_dictionary(dictionary)
         if self.getDebugInfo() is None:
             return self.buildDefault(architecture, compiler, dictionary)
diff --git a/lldb/source/Core/Disassembler.cpp b/lldb/source/Core/Disassembler.cpp
index cc98b7309d6de..111cec6e29683 100644
--- a/lldb/source/Core/Disassembler.cpp
+++ b/lldb/source/Core/Disassembler.cpp
@@ -540,34 +540,29 @@ void Disassembler::PrintInstructions(Debugger &debugger, const ArchSpec &arch,
 }
 
 bool Disassembler::Disassemble(Debugger &debugger, const ArchSpec &arch,
-                               const char *plugin_name, const char *flavor,
-                               const ExecutionContext &exe_ctx,
-                               uint32_t num_instructions,
-                               bool mixed_source_and_assembly,
-                               uint32_t num_mixed_context_lines,
-                               uint32_t options, Stream &strm) {
+                               StackFrame &frame, Stream &strm) {
   AddressRange range;
-  StackFrame *frame = exe_ctx.GetFramePtr();
-  if (frame) {
-    SymbolContext sc(
-        frame->GetSymbolContext(eSymbolContextFunction | eSymbolContextSymbol));
-    if (sc.function) {
-      range = sc.function->GetAddressRange();
-    } else if (sc.symbol && sc.symbol->ValueIsAddress()) {
-      range.GetBaseAddress() = sc.symbol->GetAddressRef();
-      range.SetByteSize(sc.symbol->GetByteSize());
-    } else {
-      range.GetBaseAddress() = frame->GetFrameCodeAddress();
-    }
+  SymbolContext sc(
+      frame.GetSymbolContext(eSymbolContextFunction | eSymbolContextSymbol));
+  if (sc.function) {
+    range = sc.function->GetAddressRange();
+  } else if (sc.symbol && sc.symbol->ValueIsAddress()) {
+    range.GetBaseAddress() = sc.symbol->GetAddressRef();
+    range.SetByteSize(sc.symbol->GetByteSize());
+  } else {
+    range.GetBaseAddress() = frame.GetFrameCodeAddress();
+  }
 
     if (range.GetBaseAddress().IsValid() && range.GetByteSize() == 0)
       range.SetByteSize(DEFAULT_DISASM_BYTE_SIZE);
-  }
 
-  return Disassemble(
-      debugger, arch, plugin_name, flavor, exe_ctx, range.GetBaseAddress(),
-      {Limit::Instructions, num_instructions}, mixed_source_and_assembly,
-      num_mixed_context_lines, options, strm);
+    Disassembler::Limit limit = {Disassembler::Limit::Bytes,
+                                 range.GetByteSize()};
+    if (limit.value == 0)
+      limit.value = DEFAULT_DISASM_BYTE_SIZE;
+
+    return Disassemble(debugger, arch, nullptr, nullptr, frame,
+                       range.GetBaseAddress(), limit, false, 0, 0, strm);
 }
 
 Instruction::Instruction(const Address &address, AddressClass addr_class)
diff --git a/lldb/source/Host/common/ProcessLaunchInfo.cpp b/lldb/source/Host/common/ProcessLaunchInfo.cpp
index 851069d0f20fb..01cf3d76314e4 100644
--- a/lldb/source/Host/common/ProcessLaunchInfo.cpp
+++ b/lldb/source/Host/common/ProcessLaunchInfo.cpp
@@ -253,7 +253,6 @@ bool ProcessLaunchInfo::ConvertArgumentsForLaunchingInShell(
       if (argv == nullptr || argv[0] == nullptr)
         return false;
       Args shell_arguments;
-      std::string safe_arg;
       shell_arguments.AppendArgument(shell_executable);
       const llvm::Triple &triple = GetArchitecture().GetTriple();
       if (triple.getOS() == llvm::Triple::Win32 &&
@@ -330,9 +329,10 @@ bool ProcessLaunchInfo::ConvertArgumentsForLaunchingInShell(
           return false;
       } else {
         for (size_t i = 0; argv[i] != nullptr; ++i) {
-          const char *arg =
-              Args::GetShellSafeArgument(m_shell, argv[i], safe_arg);
-          shell_command.Printf(" %s", arg);
+          std::string safe_arg = Args::GetShellSafeArgument(m_shell, argv[i]);
+          // Add a space to separate this arg from the previous one.
+          shell_command.PutCString(" ");
+          shell_command.PutCString(safe_arg);
         }
       }
       shell_arguments.AppendArgument(shell_command.GetString());
diff --git a/lldb/source/Plugins/SymbolFile/DWARF/DWARFUnit.h b/lldb/source/Plugins/SymbolFile/DWARF/DWARFUnit.h
index abe16182ef620..d6009518da127 100644
--- a/lldb/source/Plugins/SymbolFile/DWARF/DWARFUnit.h
+++ b/lldb/source/Plugins/SymbolFile/DWARF/DWARFUnit.h
@@ -166,7 +166,7 @@ class DWARFUnit : public lldb_private::UserID {
 
   void SetBaseAddress(dw_addr_t base_addr);
 
-  DWARFBaseDIE GetUnitDIEOnly() { return DWARFDIE(this, GetUnitDIEPtrOnly()); }
+  DWARFBaseDIE GetUnitDIEOnly() { return {this, GetUnitDIEPtrOnly()}; }
 
   DWARFDIE DIE() { return DWARFDIE(this, DIEPtr()); }
 
diff --git a/lldb/source/Target/StackFrame.cpp b/lldb/source/Target/StackFrame.cpp
index 22bca52d7f98a..131581567d736 100644
--- a/lldb/source/Target/StackFrame.cpp
+++ b/lldb/source/Target/StackFrame.cpp
@@ -229,21 +229,16 @@ bool StackFrame::ChangePC(addr_t pc) {
 
 const char *StackFrame::Disassemble() {
   std::lock_guard<std::recursive_mutex> guard(m_mutex);
-  if (m_disassembly.Empty()) {
-    ExecutionContext exe_ctx(shared_from_this());
-    Target *target = exe_ctx.GetTargetPtr();
-    if (target) {
-      const char *plugin_name = nullptr;
-      const char *flavor = nullptr;
-      Disassembler::Disassemble(target->GetDebugger(),
-                                target->GetArchitecture(), plugin_name, flavor,
-                                exe_ctx, 0, false, 0, 0, m_disassembly);
-    }
-    if (m_disassembly.Empty())
-      return nullptr;
+  if (!m_disassembly.Empty())
+    return m_disassembly.GetData();
+
+  ExecutionContext exe_ctx(shared_from_this());
+  if (Target *target = exe_ctx.GetTargetPtr()) {
+    Disassembler::Disassemble(target->GetDebugger(), target->GetArchitecture(),
+                              *this, m_disassembly);
   }
 
-  return m_disassembly.GetData();
+  return m_disassembly.Empty() ? nullptr : m_disassembly.GetData();
 }
 
 Block *StackFrame::GetFrameBlock() {
diff --git a/lldb/source/Utility/Args.cpp b/lldb/source/Utility/Args.cpp
index 4f3285404b6d2..152e73db08eaa 100644
--- a/lldb/source/Utility/Args.cpp
+++ b/lldb/source/Utility/Args.cpp
@@ -175,6 +175,11 @@ Args::Args(const StringList &list) : Args() {
     AppendArgument(arg);
 }
 
+Args::Args(llvm::ArrayRef<llvm::StringRef> args) : Args() {
+  for (llvm::StringRef arg : args)
+    AppendArgument(arg);
+}
+
 Args &Args::operator=(const Args &rhs) {
   Clear();
 
@@ -374,9 +379,8 @@ void Args::Clear() {
   m_argv.push_back(nullptr);
 }
 
-const char *Args::GetShellSafeArgument(const FileSpec &shell,
-                                       const char *unsafe_arg,
-                                       std::string &safe_arg) {
+std::string Args::GetShellSafeArgument(const FileSpec &shell,
+                                       llvm::StringRef unsafe_arg) {
   struct ShellDescriptor {
     ConstString m_basename;
     const char *m_escapables;
@@ -398,7 +402,7 @@ const char *Args::GetShellSafeArgument(const FileSpec &shell,
     }
   }
 
-  safe_arg.assign(unsafe_arg);
+  std::string safe_arg(unsafe_arg);
   size_t prev_pos = 0;
   while (prev_pos < safe_arg.size()) {
     // Escape spaces and quotes
@@ -409,7 +413,7 @@ const char *Args::GetShellSafeArgument(const FileSpec &shell,
     } else
       break;
   }
-  return safe_arg.c_str();
+  return safe_arg;
 }
 
 lldb::Encoding Args::StringToEncoding(llvm::StringRef s,
diff --git a/lldb/test/API/commands/disassemble/basic/TestFrameDisassemble.py b/lldb/test/API/commands/disassemble/basic/TestFrameDisassemble.py
index 156458234a16f..5c573f064e1e3 100644
--- a/lldb/test/API/commands/disassemble/basic/TestFrameDisassemble.py
+++ b/lldb/test/API/commands/disassemble/basic/TestFrameDisassemble.py
@@ -57,4 +57,6 @@ def frame_disassemble_test(self):
 
         frame = threads[0].GetFrameAtIndex(0)
         disassembly = frame.Disassemble()
-        self.assertNotEqual(len(disassembly), 0, "Disassembly was empty.")
+        self.assertNotEqual(disassembly, "")
+        self.assertNotIn("error", disassembly)
+        self.assertIn(": nop", disassembly)
diff --git a/lldb/test/API/commands/disassemble/basic/main.cpp b/lldb/test/API/commands/disassemble/basic/main.cpp
index cfa2498fc0208..4a2d9c106e219 100644
--- a/lldb/test/API/commands/disassemble/basic/main.cpp
+++ b/lldb/test/API/commands/disassemble/basic/main.cpp
@@ -2,6 +2,7 @@ int
 sum (int a, int b)
 {
     int result = a + b; // Set a breakpoint here
+    asm("nop");
     return result;
 }
 
diff --git a/lldb/test/API/python_api/target/TestTargetAPI.py b/lldb/test/API/python_api/target/TestTargetAPI.py
index a4d3aea94bec8..0f5fe80177e6c 100644
--- a/lldb/test/API/python_api/target/TestTargetAPI.py
+++ b/lldb/test/API/python_api/target/TestTargetAPI.py
@@ -331,7 +331,7 @@ def get_description(self):
         if not desc:
             self.fail("SBTarget.GetDescription() failed")
         self.expect(desc, exe=False,
-                    substrs=['a.out', 'Target', 'Module', 'Breakpoint'])
+                    substrs=['Target', 'Module', 'a.out', 'Breakpoint'])
 
     @not_remote_testsuite_ready
     @add_test_categories(['pyapi'])
diff --git a/lldb/test/Shell/lldb-server/TestErrorMessages.test b/lldb/test/Shell/lldb-server/TestErrorMessages.test
new file mode 100644
index 0000000000000..ef64ec6e5aba3
--- /dev/null
+++ b/lldb/test/Shell/lldb-server/TestErrorMessages.test
@@ -0,0 +1,14 @@
+RUN: lldb-server gdbserver --fd 2>&1 | FileCheck --check-prefixes=FD1,ALL %s
+FD1: error: --fd: missing argument
+
+RUN: lldb-server gdbserver --fd three 2>&1 | FileCheck --check-prefixes=FD2,ALL %s
+FD2: error: invalid '--fd' argument
+
+RUN: lldb-server gdbserver --bogus 2>&1 | FileCheck --check-prefixes=BOGUS,ALL %s
+BOGUS: error: unknown argument '--bogus'
+
+RUN: lldb-server gdbserver 2>&1 | FileCheck --check-prefixes=CONN,ALL %s
+CONN: error: no connection arguments
+
+ALL: Use '{{.*}} g[dbserver] --help' for a complete list of options.
+
diff --git a/lldb/tools/lldb-server/CMakeLists.txt b/lldb/tools/lldb-server/CMakeLists.txt
index 6e7b30df5c581..930c327cf072c 100644
--- a/lldb/tools/lldb-server/CMakeLists.txt
+++ b/lldb/tools/lldb-server/CMakeLists.txt
@@ -1,3 +1,8 @@
+set(LLVM_TARGET_DEFINITIONS LLGSOptions.td)
+tablegen(LLVM LLGSOptions.inc -gen-opt-parser-defs)
+add_public_tablegen_target(LLGSOptionsTableGen)
+set_target_properties(LLGSOptionsTableGen PROPERTIES FOLDER "lldb misc")
+
 set(LLDB_PLUGINS)
 
 if(CMAKE_SYSTEM_NAME MATCHES "Linux|Android")
@@ -53,8 +58,13 @@ add_lldb_tool(lldb-server
       ${LLDB_SYSTEM_LIBS}
 
     LINK_COMPONENTS
+      Option
       Support
 )
 
+add_dependencies(lldb-server
+  LLGSOptionsTableGen
+  ${tablegen_deps}
+)
 target_include_directories(lldb-server PRIVATE "${LLDB_SOURCE_DIR}/source")
 target_link_libraries(lldb-server PRIVATE ${LLDB_SYSTEM_LIBS})
diff --git a/lldb/tools/lldb-server/LLGSOptions.td b/lldb/tools/lldb-server/LLGSOptions.td
new file mode 100644
index 0000000000000..429a4671764f3
--- /dev/null
+++ b/lldb/tools/lldb-server/LLGSOptions.td
@@ -0,0 +1,62 @@
+include "llvm/Option/OptParser.td"
+
+class F<string name>: Flag<["--", "-"], name>;
+class R<list<string> prefixes, string name>
+  : Option<prefixes, name, KIND_REMAINING_ARGS>;
+
+multiclass SJ<string name, string help> {
+  def NAME: Separate<["--", "-"], name>,
+    HelpText<help>;
+  def NAME # _eq: Joined<["--", "-"], name # "=">,
+    Alias<!cast<Separate>(NAME)>;
+}
+
+def grp_connect : OptionGroup<"connection">, HelpText<"CONNECTION">;
+
+defm fd: SJ<"fd", "Communicate over the given file descriptor.">,
+  MetaVarName<"<fd>">,
+  Group<grp_connect>;
+
+defm named_pipe: SJ<"named-pipe", "Write port lldb-server will listen on to the given named pipe.">,
+  MetaVarName<"<name>">,
+  Group<grp_connect>;
+
+defm pipe: SJ<"pipe", "Write port lldb-server will listen on to the given file descriptor.">,
+  MetaVarName<"<fd>">,
+  Group<grp_connect>;
+
+def reverse_connect: F<"reverse-connect">,
+  HelpText<"Connect to the client instead of passively waiting for a connection. In this case [host]:port denotes the remote address to connect to.">,
+  Group<grp_connect>;
+
+def grp_general : OptionGroup<"general options">, HelpText<"GENERAL OPTIONS">;
+
+defm log_channels: SJ<"log-channels", "Channels to log. A colon-separated list of entries. Each entry starts with a channel followed by a space-separated list of categories.">,
+  MetaVarName<"<channel1 categories...:channel2 categories...>">,
+  Group<grp_general>;
+
+defm log_file: SJ<"log-file", "Destination file to log to. If empty, log to stderr.">,
+  MetaVarName<"<file>">,
+  Group<grp_general>;
+
+def setsid: F<"setsid">, HelpText<"Run lldb-server in a new session.">,
+  Group<grp_general>;
+def: Flag<["-"], "S">, Alias<setsid>,
+  Group<grp_general>;
+
+def help: F<"help">, HelpText<"Prints out the usage information for lldb-server.">,
+  Group<grp_general>;
+def: Flag<["-"], "h">, Alias<help>,
+  Group<grp_general>;
+
+def grp_target : OptionGroup<"target selection">, HelpText<"TARGET SELECTION">;
+
+defm attach: SJ<"attach", "Attach to the process given by a (numeric) process id or a name.">,
+  MetaVarName<"<pid-or-name>">,
+  Group<grp_target>;
+
+def REM : R<["--"], "">, HelpText<"Launch program for debugging.">,
+  MetaVarName<"program args">,
+  Group<grp_target>;
+
+def: F<"native-regs">; // Noop. Present for backwards compatibility only.
diff --git a/lldb/tools/lldb-server/lldb-gdbserver.cpp b/lldb/tools/lldb-server/lldb-gdbserver.cpp
index 633e37c3a0435..9b8c67af14db2 100644
--- a/lldb/tools/lldb-server/lldb-gdbserver.cpp
+++ b/lldb/tools/lldb-server/lldb-gdbserver.cpp
@@ -17,7 +17,6 @@
 #include <unistd.h>
 #endif
 
-
 #include "Acceptor.h"
 #include "LLDBServerUtilities.h"
 #include "Plugins/Process/gdb-remote/GDBRemoteCommunicationServerLLGS.h"
@@ -25,8 +24,6 @@
 #include "lldb/Host/Config.h"
 #include "lldb/Host/ConnectionFileDescriptor.h"
 #include "lldb/Host/FileSystem.h"
-#include "lldb/Host/HostGetOpt.h"
-#include "lldb/Host/OptionParser.h"
 #include "lldb/Host/Pipe.h"
 #include "lldb/Host/Socket.h"
 #include "lldb/Host/StringConvert.h"
@@ -34,7 +31,11 @@
 #include "lldb/Target/Process.h"
 #include "lldb/Utility/Status.h"
 #include "llvm/ADT/StringRef.h"
+#include "llvm/Option/ArgList.h"
+#include "llvm/Option/OptTable.h"
+#include "llvm/Option/Option.h"
 #include "llvm/Support/Errno.h"
+#include "llvm/Support/WithColor.h"
 
 #if defined(__linux__)
 #include "Plugins/Process/Linux/NativeProcessLinux.h"
@@ -88,31 +89,6 @@ class NativeProcessFactory : public NativeProcessProtocol::Factory {
 #endif
 }
 
-// option descriptors for getopt_long_only()
-
-static int g_debug = 0;
-static int g_verbose = 0;
-
-static struct option g_long_options[] = {
-    {"debug", no_argument, &g_debug, 1},
-    {"verbose", no_argument, &g_verbose, 1},
-    {"log-file", required_argument, nullptr, 'l'},
-    {"log-channels", required_argument, nullptr, 'c'},
-    {"attach", required_argument, nullptr, 'a'},
-    {"named-pipe", required_argument, nullptr, 'N'},
-    {"pipe", required_argument, nullptr, 'U'},
-    {"native-regs", no_argument, nullptr,
-     'r'}, // Specify to use the native registers instead of the gdb defaults
-           // for the architecture.  NOTE: this is a do-nothing arg as it's
-           // behavior is default now.  FIXME remove call from lldb-platform.
-    {"reverse-connect", no_argument, nullptr,
-     'R'}, // Specifies that llgs attaches to the client address:port rather
-           // than llgs listening for a connection from address on port.
-    {"setsid", no_argument, nullptr,
-     'S'}, // Call setsid() to make llgs run in its own session.
-    {"fd", required_argument, nullptr, 'F'},
-    {nullptr, 0, nullptr, 0}};
-
 #ifndef _WIN32
 // Watch for signals
 static int g_sighup_received_count = 0;
@@ -129,20 +105,6 @@ static void sighup_handler(MainLoopBase &mainloop) {
 }
 #endif // #ifndef _WIN32
 
-static void display_usage(const char *progname, const char *subcommand) {
-  fprintf(stderr, "Usage:\n  %s %s "
-                  "[--log-file log-file-name] "
-                  "[--log-channels log-channel-list] "
-                  "[--setsid] "
-                  "[--fd file-descriptor]"
-                  "[--named-pipe named-pipe-path] "
-                  "[--native-regs] "
-                  "[--attach pid] "
-                  "[[HOST]:PORT] "
-                  "[-- PROGRAM ARG1 ARG2 ...]\n",
-          progname, subcommand);
-}
-
 void handle_attach_to_pid(GDBRemoteCommunicationServerLLGS &gdb_server,
                           lldb::pid_t pid) {
   Status error = gdb_server.AttachToProcess(pid);
@@ -176,12 +138,12 @@ void handle_attach(GDBRemoteCommunicationServerLLGS &gdb_server,
     handle_attach_to_process_name(gdb_server, attach_target);
 }
 
-void handle_launch(GDBRemoteCommunicationServerLLGS &gdb_server, int argc,
-                   const char *const argv[]) {
+void handle_launch(GDBRemoteCommunicationServerLLGS &gdb_server,
+                   llvm::ArrayRef<llvm::StringRef> Arguments) {
   ProcessLaunchInfo info;
   info.GetFlags().Set(eLaunchFlagStopAtEntry | eLaunchFlagDebug |
                       eLaunchFlagDisableASLR);
-  info.SetArguments(const_cast<const char **>(argv), true);
+  info.SetArguments(Args(Arguments), true);
 
   llvm::SmallString<64> cwd;
   if (std::error_code ec = llvm::sys::fs::current_path(cwd)) {
@@ -198,7 +160,7 @@ void handle_launch(GDBRemoteCommunicationServerLLGS &gdb_server, int argc,
   Status error = gdb_server.LaunchProcess();
   if (error.Fail()) {
     llvm::errs() << llvm::formatv("error: failed to launch '{0}': {1}\n",
-                                  argv[0], error);
+                                  Arguments[0], error);
     exit(1);
   }
 }
@@ -229,7 +191,7 @@ Status writeSocketIdToPipe(lldb::pipe_t unnamed_pipe,
 
 void ConnectToRemote(MainLoop &mainloop,
                      GDBRemoteCommunicationServerLLGS &gdb_server,
-                     bool reverse_connect, const char *const host_and_port,
+                     bool reverse_connect, llvm::StringRef host_and_port,
                      const char *const progname, const char *const subcommand,
                      const char *const named_pipe_path, pipe_t unnamed_pipe,
                      int connection_fd) {
@@ -258,7 +220,7 @@ void ConnectToRemote(MainLoop &mainloop,
               connection_url, error.AsCString());
       exit(-1);
     }
-  } else if (host_and_port && host_and_port[0]) {
+  } else if (!host_and_port.empty()) {
     // Parse out host and port.
     std::string final_host_and_port;
     std::string connection_host;
@@ -269,7 +231,7 @@ void ConnectToRemote(MainLoop &mainloop,
     // expect the remainder to be the port.
     if (host_and_port[0] == ':')
       final_host_and_port.append("localhost");
-    final_host_and_port.append(host_and_port);
+    final_host_and_port.append(host_and_port.str());
 
     // Note: use rfind, because the host/port may look like "[::1]:12345".
     const std::string::size_type colon_pos = final_host_and_port.rfind(':');
@@ -361,7 +323,57 @@ void ConnectToRemote(MainLoop &mainloop,
   printf("Connection established.\n");
 }
 
-// main
+namespace {
+enum ID {
+  OPT_INVALID = 0, // This is not an option ID.
+#define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM,  \
+               HELPTEXT, METAVAR, VALUES)                                      \
+  OPT_##ID,
+#include "LLGSOptions.inc"
+#undef OPTION
+};
+
+#define PREFIX(NAME, VALUE) const char *const NAME[] = VALUE;
+#include "LLGSOptions.inc"
+#undef PREFIX
+
+const opt::OptTable::Info InfoTable[] = {
+#define OPTION(PREFIX, NAME, ID, KIND, GROUP, ALIAS, ALIASARGS, FLAGS, PARAM,  \
+               HELPTEXT, METAVAR, VALUES)                                      \
+  {                                                                            \
+      PREFIX,      NAME,      HELPTEXT,                                        \
+      METAVAR,     OPT_##ID,  opt::Option::KIND##Class,                        \
+      PARAM,       FLAGS,     OPT_##GROUP,                                     \
+      OPT_##ALIAS, ALIASARGS, VALUES},
+#include "LLGSOptions.inc"
+#undef OPTION
+};
+
+class LLGSOptTable : public opt::OptTable {
+public:
+  LLGSOptTable() : OptTable(InfoTable) {}
+
+  void PrintHelp(llvm::StringRef Name) {
+    std::string Usage =
+        (Name + " [options] [[host]:port] [[--] program args...]").str();
+    OptTable::PrintHelp(llvm::outs(), Usage.c_str(), "lldb-server");
+    llvm::outs() << R"(
+DESCRIPTION
+  lldb-server connects to the LLDB client, which drives the debugging session.
+  If no connection options are given, the [host]:port argument must be present
+  and will denote the address that lldb-server will listen on. [host] defaults
+  to "localhost" if empty. Port can be zero, in which case the port number will
+  be chosen dynamically and written to destinations given by --named-pipe and
+  --pipe arguments.
+
+  If no target is selected at startup, lldb-server can be directed by the LLDB
+  client to launch or attach to a process.
+
+)";
+  }
+};
+} // namespace
+
 int main_gdbserver(int argc, char *argv[]) {
   Status error;
   MainLoop mainloop;
@@ -374,10 +386,6 @@ int main_gdbserver(int argc, char *argv[]) {
 
   const char *progname = argv[0];
   const char *subcommand = argv[1];
-  argc--;
-  argv++;
-  int long_option_index = 0;
-  int ch;
   std::string attach_target;
   std::string named_pipe_path;
   std::string log_file;
@@ -390,94 +398,71 @@ int main_gdbserver(int argc, char *argv[]) {
   // ProcessLaunchInfo launch_info;
   ProcessAttachInfo attach_info;
 
-  bool show_usage = false;
-  int option_error = 0;
-#if __GLIBC__
-  optind = 0;
-#else
-  optreset = 1;
-  optind = 1;
-#endif
-
-  std::string short_options(OptionParser::GetShortOptionString(g_long_options));
-
-  while ((ch = getopt_long_only(argc, argv, short_options.c_str(),
-                                g_long_options, &long_option_index)) != -1) {
-    switch (ch) {
-    case 0: // Any optional that auto set themselves will return 0
-      break;
-
-    case 'l': // Set Log File
-      if (optarg && optarg[0])
-        log_file.assign(optarg);
-      break;
-
-    case 'c': // Log Channels
-      if (optarg && optarg[0])
-        log_channels = StringRef(optarg);
-      break;
-
-    case 'N': // named pipe
-      if (optarg && optarg[0])
-        named_pipe_path = optarg;
-      break;
-
-    case 'U': // unnamed pipe
-      if (optarg && optarg[0])
-        unnamed_pipe = (pipe_t)StringConvert::ToUInt64(optarg, -1);
-      break;
-
-    case 'r':
-      // Do nothing, native regs is the default these days
-      break;
-
-    case 'R':
-      reverse_connect = true;
-      break;
+  LLGSOptTable Opts;
+  llvm::BumpPtrAllocator Alloc;
+  llvm::StringSaver Saver(Alloc);
+  bool HasError = false;
+  opt::InputArgList Args = Opts.parseArgs(argc - 1, argv + 1, OPT_UNKNOWN,
+                                          Saver, [&](llvm::StringRef Msg) {
+                                            WithColor::error() << Msg << "\n";
+                                            HasError = true;
+                                          });
+  std::string Name =
+      (llvm::sys::path::filename(argv[0]) + " g[dbserver]").str();
+  std::string HelpText =
+      "Use '" + Name + " --help' for a complete list of options.\n";
+  if (HasError) {
+    llvm::errs() << HelpText;
+    return 1;
+  }
 
-    case 'F':
-      connection_fd = StringConvert::ToUInt32(optarg, -1);
-      break;
+  if (Args.hasArg(OPT_help)) {
+    Opts.PrintHelp(Name);
+    return 0;
+  }
 
 #ifndef _WIN32
-    case 'S':
-      // Put llgs into a new session. Terminals group processes
-      // into sessions and when a special terminal key sequences
-      // (like control+c) are typed they can cause signals to go out to
-      // all processes in a session. Using this --setsid (-S) option
-      // will cause debugserver to run in its own sessions and be free
-      // from such issues.
-      //
-      // This is useful when llgs is spawned from a command
-      // line application that uses llgs to do the debugging,
-      // yet that application doesn't want llgs receiving the
-      // signals sent to the session (i.e. dying when anyone hits ^C).
-      {
-        const ::pid_t new_sid = setsid();
-        if (new_sid == -1) {
-          llvm::errs() << llvm::formatv(
-              "failed to set new session id for {0} ({1})\n", LLGS_PROGRAM_NAME,
-              llvm::sys::StrError());
-        }
+  if (Args.hasArg(OPT_setsid)) {
+    // Put llgs into a new session. Terminals group processes
+    // into sessions and when a special terminal key sequences
+    // (like control+c) are typed they can cause signals to go out to
+    // all processes in a session. Using this --setsid (-S) option
+    // will cause debugserver to run in its own sessions and be free
+    // from such issues.
+    //
+    // This is useful when llgs is spawned from a command
+    // line application that uses llgs to do the debugging,
+    // yet that application doesn't want llgs receiving the
+    // signals sent to the session (i.e. dying when anyone hits ^C).
+    {
+      const ::pid_t new_sid = setsid();
+      if (new_sid == -1) {
+        WithColor::warning()
+            << llvm::formatv("failed to set new session id for {0} ({1})\n",
+                             LLGS_PROGRAM_NAME, llvm::sys::StrError());
       }
-      break;
+    }
+  }
 #endif
 
-    case 'a': // attach {pid|process_name}
-      if (optarg && optarg[0])
-        attach_target = optarg;
-      break;
-
-    case 'h': /* fall-through is intentional */
-    case '?':
-      show_usage = true;
-      break;
+  log_file = Args.getLastArgValue(OPT_log_file).str();
+  log_channels = Args.getLastArgValue(OPT_log_channels);
+  named_pipe_path = Args.getLastArgValue(OPT_named_pipe).str();
+  reverse_connect = Args.hasArg(OPT_reverse_connect);
+  attach_target = Args.getLastArgValue(OPT_attach).str();
+  if (Args.hasArg(OPT_pipe)) {
+    uint64_t Arg;
+    if (!llvm::to_integer(Args.getLastArgValue(OPT_pipe), Arg)) {
+      WithColor::error() << "invalid '--pipe' argument\n" << HelpText;
+      return 1;
     }
+    unnamed_pipe = (pipe_t)Arg;
   }
-
-  if (show_usage || option_error) {
-    display_usage(progname, subcommand);
-    exit(option_error);
+  if (Args.hasArg(OPT_fd)) {
+    if (!llvm::to_integer(Args.getLastArgValue(OPT_fd), connection_fd)) {
+      WithColor::error() << "invalid '--fd' argument\n" << HelpText;
+      return 1;
+    }
   }
 
   if (!LLDBServerUtilities::SetupLogging(
@@ -486,30 +471,26 @@ int main_gdbserver(int argc, char *argv[]) {
               LLDB_LOG_OPTION_PREPEND_FILE_FUNCTION))
     return -1;
 
-  Log *log(lldb_private::GetLogIfAnyCategoriesSet(GDBR_LOG_PROCESS));
-  if (log) {
-    LLDB_LOGF(log, "lldb-server launch");
-    for (int i = 0; i < argc; i++) {
-      LLDB_LOGF(log, "argv[%i] = '%s'", i, argv[i]);
-    }
+  std::vector<llvm::StringRef> Inputs;
+  for (opt::Arg *Arg : Args.filtered(OPT_INPUT))
+    Inputs.push_back(Arg->getValue());
+  if (opt::Arg *Arg = Args.getLastArg(OPT_REM)) {
+    for (const char *Val : Arg->getValues())
+      Inputs.push_back(Val);
   }
-
-  // Skip any options we consumed with getopt_long_only.
-  argc -= optind;
-  argv += optind;
-
-  if (argc == 0 && connection_fd == -1) {
-    fputs("No arguments\n", stderr);
-    display_usage(progname, subcommand);
-    exit(255);
+  if (Inputs.empty() && connection_fd == -1) {
+    WithColor::error() << "no connection arguments\n" << HelpText;
+    return 1;
   }
 
   NativeProcessFactory factory;
   GDBRemoteCommunicationServerLLGS gdb_server(mainloop, factory);
 
-  const char *const host_and_port = argv[0];
-  argc -= 1;
-  argv += 1;
+  llvm::StringRef host_and_port;
+  if (!Inputs.empty()) {
+    host_and_port = Inputs.front();
+    Inputs.erase(Inputs.begin());
+  }
 
   // Any arguments left over are for the program that we need to launch. If
   // there
@@ -520,8 +501,8 @@ int main_gdbserver(int argc, char *argv[]) {
   // explicitly asked to attach with the --attach={pid|program_name} form.
   if (!attach_target.empty())
     handle_attach(gdb_server, attach_target);
-  else if (argc > 0)
-    handle_launch(gdb_server, argc, argv);
+  else if (!Inputs.empty())
+    handle_launch(gdb_server, Inputs);
 
   // Print version info.
   printf("%s-%s\n", LLGS_PROGRAM_NAME, LLGS_VERSION_STR);
@@ -532,7 +513,6 @@ int main_gdbserver(int argc, char *argv[]) {
 
   if (!gdb_server.IsConnected()) {
     fprintf(stderr, "no connection information provided, unable to run\n");
-    display_usage(progname, subcommand);
     return 1;
   }
 
diff --git a/lldb/unittests/Utility/ArgsTest.cpp b/lldb/unittests/Utility/ArgsTest.cpp
index ed1235f99c718..cd62c08df31e4 100644
--- a/lldb/unittests/Utility/ArgsTest.cpp
+++ b/lldb/unittests/Utility/ArgsTest.cpp
@@ -9,6 +9,7 @@
 #include "gtest/gtest.h"
 
 #include "lldb/Utility/Args.h"
+#include "lldb/Utility/FileSpec.h"
 #include "lldb/Utility/StringList.h"
 
 #include <limits>
@@ -313,3 +314,25 @@ TEST(ArgsTest, Yaml) {
   EXPECT_EQ(entries[2].GetQuoteChar(), '"');
   EXPECT_EQ(entries[3].GetQuoteChar(), '\0');
 }
+
+TEST(ArgsTest, GetShellSafeArgument) {
+  // Try escaping with bash at start/middle/end of the argument.
+  FileSpec bash("/bin/bash", FileSpec::Style::posix);
+  EXPECT_EQ(Args::GetShellSafeArgument(bash, "\"b"), "\\\"b");
+  EXPECT_EQ(Args::GetShellSafeArgument(bash, "a\""), "a\\\"");
+  EXPECT_EQ(Args::GetShellSafeArgument(bash, "a\"b"), "a\\\"b");
+
+  // String that doesn't need to be escaped
+  EXPECT_EQ(Args::GetShellSafeArgument(bash, "a"), "a");
+
+  // Try escaping with tcsh and the tcsh-specific "$" escape.
+  FileSpec tcsh("/bin/tcsh", FileSpec::Style::posix);
+  EXPECT_EQ(Args::GetShellSafeArgument(tcsh, "a$b"), "a\\$b");
+  // Bash however doesn't need escaping for "$".
+  EXPECT_EQ(Args::GetShellSafeArgument(bash, "a$b"), "a$b");
+
+  // Try escaping with an unknown shell.
+  FileSpec unknown_shell("/bin/unknown_shell", FileSpec::Style::posix);
+  EXPECT_EQ(Args::GetShellSafeArgument(unknown_shell, "a'b"), "a\\'b");
+  EXPECT_EQ(Args::GetShellSafeArgument(unknown_shell, "a\"b"), "a\\\"b");
+}
diff --git a/llvm/CMakeLists.txt b/llvm/CMakeLists.txt
index 2f5206c15d061..6b9fa7b212e28 100644
--- a/llvm/CMakeLists.txt
+++ b/llvm/CMakeLists.txt
@@ -1104,7 +1104,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     PATTERN "*.td"
     PATTERN "*.inc"
     PATTERN "LICENSE.TXT"
-    PATTERN ".svn" EXCLUDE
     )
 
   install(DIRECTORY ${LLVM_INCLUDE_DIR}/llvm ${LLVM_INCLUDE_DIR}/llvm-c
@@ -1118,7 +1117,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     # Exclude include/llvm/CMakeFiles/intrinsics_gen.dir, matched by "*.def"
     PATTERN "CMakeFiles" EXCLUDE
     PATTERN "config.h" EXCLUDE
-    PATTERN ".svn" EXCLUDE
     )
 
   if (LLVM_INSTALL_MODULEMAPS)
diff --git a/llvm/cmake/modules/CMakeLists.txt b/llvm/cmake/modules/CMakeLists.txt
index 4b8879f65fe47..505dc9a29d700 100644
--- a/llvm/cmake/modules/CMakeLists.txt
+++ b/llvm/cmake/modules/CMakeLists.txt
@@ -93,7 +93,6 @@ set(llvm_config_include_buildtree_only_exports)
 file(COPY .
   DESTINATION ${llvm_cmake_builddir}
   FILES_MATCHING PATTERN *.cmake
-  PATTERN .svn EXCLUDE
   PATTERN CMakeFiles EXCLUDE
   )
 
@@ -152,7 +151,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     DESTINATION ${LLVM_INSTALL_PACKAGE_DIR}
     COMPONENT cmake-exports
     FILES_MATCHING PATTERN *.cmake
-    PATTERN .svn EXCLUDE
     PATTERN LLVMConfig.cmake EXCLUDE
     PATTERN LLVMConfigExtensions.cmake EXCLUDE
     PATTERN LLVMConfigVersion.cmake EXCLUDE
diff --git a/llvm/docs/ReleaseNotes.rst b/llvm/docs/ReleaseNotes.rst
index d85f4bab210f3..5089cc6ca93e8 100644
--- a/llvm/docs/ReleaseNotes.rst
+++ b/llvm/docs/ReleaseNotes.rst
@@ -118,7 +118,8 @@ During this release ...
   the "target-cpu" attribute or TargetMachine CPU which will be used to select
   Instruction Set. If the attribute is not present, the tune CPU will follow
   the target CPU.
-* Support for ISA HRESET has been added.
+* Support for ``HRESET`` instructions has been added.
+* Support for ``UINTR`` instructions has been added.
 
 Changes to the AMDGPU Target
 -----------------------------
diff --git a/llvm/docs/TableGen/BackGuide.rst b/llvm/docs/TableGen/BackGuide.rst
index 430e640778cf2..3a84babc4c399 100644
--- a/llvm/docs/TableGen/BackGuide.rst
+++ b/llvm/docs/TableGen/BackGuide.rst
@@ -22,8 +22,8 @@ for C++, but may be any type of file that the backend developer needs.
 This document is a guide to writing a backend for TableGen. It is not a
 complete reference manual, but rather a guide to using the facilities
 provided by TableGen for the backends. For a complete reference to the
-various data structures and functions involved, see the Doxygen
-documentation.
+various data structures and functions involved, see the primary TableGen
+header file (``record.h``) and/or the Doxygen documentation.
 
 This document assumes that you have read the :doc:`TableGen Programmer's
 Reference <./ProgRef>`, which provides a detailed reference for coding
diff --git a/llvm/include/llvm/ADT/ImmutableMap.h b/llvm/include/llvm/ADT/ImmutableMap.h
index 30689d2274a87..81b21a7319a71 100644
--- a/llvm/include/llvm/ADT/ImmutableMap.h
+++ b/llvm/include/llvm/ADT/ImmutableMap.h
@@ -355,7 +355,7 @@ class ImmutableMapRef {
   unsigned getHeight() const { return Root ? Root->getHeight() : 0; }
 
   static inline void Profile(FoldingSetNodeID &ID, const ImmutableMapRef &M) {
-    ID.AddPointer(M.Root);
+    ID.AddPointer(M.Root.get());
   }
 
   inline void Profile(FoldingSetNodeID &ID) const { return Profile(ID, *this); }
diff --git a/llvm/include/llvm/ADT/Triple.h b/llvm/include/llvm/ADT/Triple.h
index 80e7ae0b48f3b..5bac9162d444a 100644
--- a/llvm/include/llvm/ADT/Triple.h
+++ b/llvm/include/llvm/ADT/Triple.h
@@ -500,6 +500,12 @@ class Triple {
     return getEnvironment() == Triple::MacABI;
   }
 
+  /// Returns true for targets that run on a macOS machine.
+  bool isTargetMachineMac() const {
+    return isMacOSX() || (isOSDarwin() && (isSimulatorEnvironment() ||
+                                           isMacCatalystEnvironment()));
+  }
+
   bool isOSNetBSD() const {
     return getOS() == Triple::NetBSD;
   }
diff --git a/llvm/include/llvm/Analysis/RegionPass.h b/llvm/include/llvm/Analysis/RegionPass.h
index 995c5dca3de3b..5c7fa5f56693a 100644
--- a/llvm/include/llvm/Analysis/RegionPass.h
+++ b/llvm/include/llvm/Analysis/RegionPass.h
@@ -85,8 +85,6 @@ class RegionPass : public Pass {
 /// The pass manager to schedule RegionPasses.
 class RGPassManager : public FunctionPass, public PMDataManager {
   std::deque<Region*> RQ;
-  bool skipThisRegion;
-  bool redoThisRegion;
   RegionInfo *RI;
   Region *CurrentRegion;
 
diff --git a/llvm/include/llvm/Analysis/ScalarEvolution.h b/llvm/include/llvm/Analysis/ScalarEvolution.h
index 691c424a280c5..c8a5d02f71797 100644
--- a/llvm/include/llvm/Analysis/ScalarEvolution.h
+++ b/llvm/include/llvm/Analysis/ScalarEvolution.h
@@ -1317,39 +1317,40 @@ class ScalarEvolution {
     /// never have more than one computable exit.
     SmallVector<ExitNotTakenInfo, 1> ExitNotTaken;
 
-    /// The pointer part of \c MaxAndComplete is an expression indicating the
-    /// least maximum backedge-taken count of the loop that is known, or a
-    /// SCEVCouldNotCompute. This expression is only valid if the predicates
-    /// associated with all loop exits are true.
-    ///
-    /// The integer part of \c MaxAndComplete is a boolean indicating if \c
-    /// ExitNotTaken has an element for every exiting block in the loop.
-    PointerIntPair<const SCEV *, 1> MaxAndComplete;
+    /// Expression indicating the least maximum backedge-taken count of the loop
+    /// that is known, or a SCEVCouldNotCompute. This expression is only valid
+    /// if the redicates associated with all loop exits are true.
+    const SCEV *ConstantMax;
+
+    /// Indicating if \c ExitNotTaken has an element for every exiting block in
+    /// the loop.
+    bool IsComplete;
 
     /// True iff the backedge is taken either exactly Max or zero times.
     bool MaxOrZero = false;
 
-    /// \name Helper projection functions on \c MaxAndComplete.
+    /// \name Helper projection functions on \c ConstantMaxAndComplete.
     /// @{
-    bool isComplete() const { return MaxAndComplete.getInt(); }
-    const SCEV *getMax() const { return MaxAndComplete.getPointer(); }
+    bool isComplete() const { return IsComplete; }
+    const SCEV *getConstantMax() const { return ConstantMax; }
     /// @}
 
   public:
-    BackedgeTakenInfo() : MaxAndComplete(nullptr, 0) {}
+    BackedgeTakenInfo() : ConstantMax(nullptr), IsComplete(false) {}
     BackedgeTakenInfo(BackedgeTakenInfo &&) = default;
     BackedgeTakenInfo &operator=(BackedgeTakenInfo &&) = default;
 
     using EdgeExitInfo = std::pair<BasicBlock *, ExitLimit>;
 
     /// Initialize BackedgeTakenInfo from a list of exact exit counts.
-    BackedgeTakenInfo(ArrayRef<EdgeExitInfo> ExitCounts, bool Complete,
-                      const SCEV *MaxCount, bool MaxOrZero);
+    BackedgeTakenInfo(ArrayRef<EdgeExitInfo> ExitCounts, bool IsComplete,
+                      const SCEV *ConstantMax, bool MaxOrZero);
 
     /// Test whether this BackedgeTakenInfo contains any computed information,
     /// or whether it's all SCEVCouldNotCompute values.
     bool hasAnyInfo() const {
-      return !ExitNotTaken.empty() || !isa<SCEVCouldNotCompute>(getMax());
+      return !ExitNotTaken.empty() ||
+             !isa<SCEVCouldNotCompute>(getConstantMax());
     }
 
     /// Test whether this BackedgeTakenInfo contains complete information.
@@ -1383,16 +1384,16 @@ class ScalarEvolution {
     const SCEV *getExact(const BasicBlock *ExitingBlock,
                          ScalarEvolution *SE) const;
 
-    /// Get the max backedge taken count for the loop.
-    const SCEV *getMax(ScalarEvolution *SE) const;
+    /// Get the constant max backedge taken count for the loop.
+    const SCEV *getConstantMax(ScalarEvolution *SE) const;
 
-    /// Get the max backedge taken count for the particular loop exit.
-    const SCEV *getMax(const BasicBlock *ExitingBlock,
-                       ScalarEvolution *SE) const;
+    /// Get the constant max backedge taken count for the particular loop exit.
+    const SCEV *getConstantMax(const BasicBlock *ExitingBlock,
+                               ScalarEvolution *SE) const;
 
     /// Return true if the number of times this backedge is taken is either the
-    /// value returned by getMax or zero.
-    bool isMaxOrZero(ScalarEvolution *SE) const;
+    /// value returned by getConstantMax or zero.
+    bool isConstantMaxOrZero(ScalarEvolution *SE) const;
 
     /// Return true if any backedge taken count expressions refer to the given
     /// subexpression.
@@ -1497,13 +1498,6 @@ class ScalarEvolution {
   ConstantRange getRangeForAffineAR(const SCEV *Start, const SCEV *Stop,
                                     const SCEV *MaxBECount, unsigned BitWidth);
 
-  /// Determines the range for the affine non-self-wrapping SCEVAddRecExpr {\p
-  /// Start,+,\p Stop}<nw>.
-  ConstantRange getRangeForAffineNoSelfWrappingAR(const SCEVAddRecExpr *AddRec,
-                                                  const SCEV *MaxBECount,
-                                                  unsigned BitWidth,
-                                                  RangeSignHint SignHint);
-
   /// Try to compute a range for the affine SCEVAddRecExpr {\p Start,+,\p
   /// Stop} by "factoring out" a ternary expression from the add recurrence.
   /// Helper called by \c getRange.
diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index abcd612c7b553..817cabd6344e8 100644
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -1166,15 +1166,9 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
       break;
 
     case Intrinsic::memcpy:
-      // FIXME: all cost kinds should default to the same thing?
-      if (CostKind != TTI::TCK_RecipThroughput)
-        return thisT()->getMemcpyCost(ICA.getInst());
-      return BaseT::getIntrinsicInstrCost(ICA, CostKind);
+      return thisT()->getMemcpyCost(ICA.getInst());
 
     case Intrinsic::masked_scatter: {
-      // FIXME: all cost kinds should default to the same thing?
-      if (CostKind != TTI::TCK_RecipThroughput)
-        return BaseT::getIntrinsicInstrCost(ICA, CostKind);
       assert(VF == 1 && "Can't vectorize types here.");
       const Value *Mask = Args[3];
       bool VarMask = !isa<Constant>(Mask);
@@ -1184,9 +1178,6 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
                                              VarMask, Alignment, CostKind, I);
     }
     case Intrinsic::masked_gather: {
-      // FIXME: all cost kinds should default to the same thing?
-      if (CostKind != TTI::TCK_RecipThroughput)
-        return BaseT::getIntrinsicInstrCost(ICA, CostKind);
       assert(VF == 1 && "Can't vectorize types here.");
       const Value *Mask = Args[2];
       bool VarMask = !isa<Constant>(Mask);
diff --git a/llvm/include/llvm/CodeGen/CommandFlags.h b/llvm/include/llvm/CodeGen/CommandFlags.h
index b58fe2d43cdba..dd9a74cad8d70 100644
--- a/llvm/include/llvm/CodeGen/CommandFlags.h
+++ b/llvm/include/llvm/CodeGen/CommandFlags.h
@@ -99,6 +99,10 @@ bool getIgnoreXCOFFVisibility();
 
 std::string getBBSections();
 
+std::string getStackProtectorGuard();
+unsigned getStackProtectorGuardOffset();
+std::string getStackProtectorGuardReg();
+
 unsigned getTLSSize();
 
 bool getEmulatedTLS();
@@ -135,6 +139,9 @@ struct RegisterCodeGenFlags {
 
 llvm::BasicBlockSection getBBSectionsMode(llvm::TargetOptions &Options);
 
+llvm::StackProtectorGuards
+getStackProtectorGuardMode(llvm::TargetOptions &Options);
+
 /// Common utility function tightly tied to the options listed here. Initializes
 /// a TargetOptions object with CodeGen flags and returns it.
 /// \p TheTriple is used to determine the default value for options if
diff --git a/llvm/include/llvm/CodeGen/MachineFunction.h b/llvm/include/llvm/CodeGen/MachineFunction.h
index 01f59045ecb4a..e9979c788ce0b 100644
--- a/llvm/include/llvm/CodeGen/MachineFunction.h
+++ b/llvm/include/llvm/CodeGen/MachineFunction.h
@@ -458,7 +458,11 @@ class MachineFunction {
   /// which has the same signature (i.e., def operands in the same place) but
   /// a modified instruction type, flags, or otherwise. An example: X86 moves
   /// are sometimes transformed into equivalent LEAs.
-  void substituteDebugValuesForInst(const MachineInstr &Old, MachineInstr &New);
+  /// If the two instructions are not the same opcode, limit which operands to
+  /// examine for substitutions to the first N operands by setting
+  /// \p MaxOperand.
+  void substituteDebugValuesForInst(const MachineInstr &Old, MachineInstr &New,
+                                    unsigned MaxOperand = UINT_MAX);
 
   MachineFunction(Function &F, const LLVMTargetMachine &Target,
                   const TargetSubtargetInfo &STI, unsigned FunctionNum,
diff --git a/llvm/include/llvm/CodeGen/ReachingDefAnalysis.h b/llvm/include/llvm/CodeGen/ReachingDefAnalysis.h
index 1d2251dbededb..bcb48de2fe5a6 100644
--- a/llvm/include/llvm/CodeGen/ReachingDefAnalysis.h
+++ b/llvm/include/llvm/CodeGen/ReachingDefAnalysis.h
@@ -139,23 +139,25 @@ class ReachingDefAnalysis : public MachineFunctionPass {
 
   /// Provides the instruction id of the closest reaching def instruction of
   /// PhysReg that reaches MI, relative to the begining of MI's basic block.
-  int getReachingDef(MachineInstr *MI, int PhysReg) const;
+  int getReachingDef(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Return whether A and B use the same def of PhysReg.
-  bool hasSameReachingDef(MachineInstr *A, MachineInstr *B, int PhysReg) const;
+  bool hasSameReachingDef(MachineInstr *A, MachineInstr *B,
+                          MCRegister PhysReg) const;
 
   /// Return whether the reaching def for MI also is live out of its parent
   /// block.
-  bool isReachingDefLiveOut(MachineInstr *MI, int PhysReg) const;
+  bool isReachingDefLiveOut(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Return the local MI that produces the live out value for PhysReg, or
   /// nullptr for a non-live out or non-local def.
   MachineInstr *getLocalLiveOutMIDef(MachineBasicBlock *MBB,
-                                     int PhysReg) const;
+                                     MCRegister PhysReg) const;
 
   /// If a single MachineInstr creates the reaching definition, then return it.
   /// Otherwise return null.
-  MachineInstr *getUniqueReachingMIDef(MachineInstr *MI, int PhysReg) const;
+  MachineInstr *getUniqueReachingMIDef(MachineInstr *MI,
+                                       MCRegister PhysReg) const;
 
   /// If a single MachineInstr creates the reaching definition, for MIs operand
   /// at Idx, then return it. Otherwise return null.
@@ -167,44 +169,44 @@ class ReachingDefAnalysis : public MachineFunctionPass {
 
   /// Provide whether the register has been defined in the same basic block as,
   /// and before, MI.
-  bool hasLocalDefBefore(MachineInstr *MI, int PhysReg) const;
+  bool hasLocalDefBefore(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Return whether the given register is used after MI, whether it's a local
   /// use or a live out.
-  bool isRegUsedAfter(MachineInstr *MI, int PhysReg) const;
+  bool isRegUsedAfter(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Return whether the given register is defined after MI.
-  bool isRegDefinedAfter(MachineInstr *MI, int PhysReg) const;
+  bool isRegDefinedAfter(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Provides the clearance - the number of instructions since the closest
   /// reaching def instuction of PhysReg that reaches MI.
-  int getClearance(MachineInstr *MI, MCPhysReg PhysReg) const;
+  int getClearance(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Provides the uses, in the same block as MI, of register that MI defines.
   /// This does not consider live-outs.
-  void getReachingLocalUses(MachineInstr *MI, int PhysReg,
+  void getReachingLocalUses(MachineInstr *MI, MCRegister PhysReg,
                             InstSet &Uses) const;
 
   /// Search MBB for a definition of PhysReg and insert it into Defs. If no
   /// definition is found, recursively search the predecessor blocks for them.
-  void getLiveOuts(MachineBasicBlock *MBB, int PhysReg, InstSet &Defs,
+  void getLiveOuts(MachineBasicBlock *MBB, MCRegister PhysReg, InstSet &Defs,
                    BlockSet &VisitedBBs) const;
-  void getLiveOuts(MachineBasicBlock *MBB, int PhysReg, InstSet &Defs) const;
+  void getLiveOuts(MachineBasicBlock *MBB, MCRegister PhysReg,
+                   InstSet &Defs) const;
 
   /// For the given block, collect the instructions that use the live-in
   /// value of the provided register. Return whether the value is still
   /// live on exit.
-  bool getLiveInUses(MachineBasicBlock *MBB, int PhysReg,
+  bool getLiveInUses(MachineBasicBlock *MBB, MCRegister PhysReg,
                      InstSet &Uses) const;
 
   /// Collect the users of the value stored in PhysReg, which is defined
   /// by MI.
-  void getGlobalUses(MachineInstr *MI, int PhysReg,
-                     InstSet &Uses) const;
+  void getGlobalUses(MachineInstr *MI, MCRegister PhysReg, InstSet &Uses) const;
 
   /// Collect all possible definitions of the value stored in PhysReg, which is
   /// used by MI.
-  void getGlobalReachingDefs(MachineInstr *MI, int PhysReg,
+  void getGlobalReachingDefs(MachineInstr *MI, MCRegister PhysReg,
                              InstSet &Defs) const;
 
   /// Return whether From can be moved forwards to just before To.
@@ -229,12 +231,13 @@ class ReachingDefAnalysis : public MachineFunctionPass {
 
   /// Return whether a MachineInstr could be inserted at MI and safely define
   /// the given register without affecting the program.
-  bool isSafeToDefRegAt(MachineInstr *MI, int PhysReg) const;
+  bool isSafeToDefRegAt(MachineInstr *MI, MCRegister PhysReg) const;
 
   /// Return whether a MachineInstr could be inserted at MI and safely define
   /// the given register without affecting the program, ignoring any effects
   /// on the provided instructions.
-  bool isSafeToDefRegAt(MachineInstr *MI, int PhysReg, InstSet &Ignore) const;
+  bool isSafeToDefRegAt(MachineInstr *MI, MCRegister PhysReg,
+                        InstSet &Ignore) const;
 
 private:
   /// Set up LiveRegs by merging predecessor live-out values.
@@ -269,7 +272,8 @@ class ReachingDefAnalysis : public MachineFunctionPass {
 
   /// Provides the instruction of the closest reaching def instruction of
   /// PhysReg that reaches MI, relative to the begining of MI's basic block.
-  MachineInstr *getReachingLocalMIDef(MachineInstr *MI, int PhysReg) const;
+  MachineInstr *getReachingLocalMIDef(MachineInstr *MI,
+                                      MCRegister PhysReg) const;
 };
 
 } // namespace llvm
diff --git a/llvm/include/llvm/CodeGen/TargetInstrInfo.h b/llvm/include/llvm/CodeGen/TargetInstrInfo.h
index 96cca0257782b..5d5a6efab2207 100644
--- a/llvm/include/llvm/CodeGen/TargetInstrInfo.h
+++ b/llvm/include/llvm/CodeGen/TargetInstrInfo.h
@@ -1399,8 +1399,13 @@ class TargetInstrInfo : public MCInstrInfo {
   /// If the specified instruction defines any predicate
   /// or condition code register(s) used for predication, returns true as well
   /// as the definition predicate(s) by reference.
-  virtual bool DefinesPredicate(MachineInstr &MI,
-                                std::vector<MachineOperand> &Pred) const {
+  /// SkipDead should be set to false at any point that dead
+  /// predicate instructions should be considered as being defined.
+  /// A dead predicate instruction is one that is guaranteed to be removed
+  /// after a call to PredicateInstruction.
+  virtual bool ClobbersPredicate(MachineInstr &MI,
+                                 std::vector<MachineOperand> &Pred,
+                                 bool SkipDead) const {
     return false;
   }
 
diff --git a/llvm/include/llvm/CodeGen/TargetLowering.h b/llvm/include/llvm/CodeGen/TargetLowering.h
index 562416679a8ac..07ff21a10e3fa 100644
--- a/llvm/include/llvm/CodeGen/TargetLowering.h
+++ b/llvm/include/llvm/CodeGen/TargetLowering.h
@@ -1688,7 +1688,7 @@ class TargetLoweringBase {
 
   /// If a physical register, this specifies the register that
   /// llvm.savestack/llvm.restorestack should save and restore.
-  unsigned getStackPointerRegisterToSaveRestore() const {
+  Register getStackPointerRegisterToSaveRestore() const {
     return StackPointerRegisterToSaveRestore;
   }
 
diff --git a/llvm/include/llvm/CodeGen/ValueTypes.h b/llvm/include/llvm/CodeGen/ValueTypes.h
index 958711f1529cc..a03ec010f0477 100644
--- a/llvm/include/llvm/CodeGen/ValueTypes.h
+++ b/llvm/include/llvm/CodeGen/ValueTypes.h
@@ -92,26 +92,17 @@ namespace llvm {
     /// with the element type converted to an integer type with the same
     /// bitwidth.
     EVT changeVectorElementTypeToInteger() const {
-      if (!isSimple())
-        return changeExtendedVectorElementTypeToInteger();
-      MVT EltTy = getSimpleVT().getVectorElementType();
-      unsigned BitWidth = EltTy.getSizeInBits();
-      MVT IntTy = MVT::getIntegerVT(BitWidth);
-      MVT VecTy = MVT::getVectorVT(IntTy, getVectorElementCount());
-      assert(VecTy.SimpleTy != MVT::INVALID_SIMPLE_VALUE_TYPE &&
-             "Simple vector VT not representable by simple integer vector VT!");
-      return VecTy;
+      if (isSimple())
+        return getSimpleVT().changeVectorElementTypeToInteger();
+      return changeExtendedVectorElementTypeToInteger();
     }
 
     /// Return a VT for a vector type whose attributes match ourselves
     /// with the exception of the element type that is chosen by the caller.
     EVT changeVectorElementType(EVT EltVT) const {
-      if (!isSimple())
-        return changeExtendedVectorElementType(EltVT);
-      MVT VecTy = MVT::getVectorVT(EltVT.V, getVectorElementCount());
-      assert(VecTy.SimpleTy != MVT::INVALID_SIMPLE_VALUE_TYPE &&
-             "Simple vector VT not representable by simple integer vector VT!");
-      return VecTy;
+      if (isSimple() && EltVT.isSimple())
+        return getSimpleVT().changeVectorElementType(EltVT.getSimpleVT());
+      return changeExtendedVectorElementType(EltVT);
     }
 
     /// Return the type converted to an equivalently sized integer or vector
@@ -122,8 +113,7 @@ namespace llvm {
         return changeVectorElementTypeToInteger();
 
       if (isSimple())
-        return MVT::getIntegerVT(getSizeInBits());
-
+        return getSimpleVT().changeTypeToInteger();
       return changeExtendedTypeToInteger();
     }
 
diff --git a/llvm/include/llvm/DebugInfo/Symbolize/Symbolize.h b/llvm/include/llvm/DebugInfo/Symbolize/Symbolize.h
index 43c89aafea9bd..085e4bb4ccb8f 100644
--- a/llvm/include/llvm/DebugInfo/Symbolize/Symbolize.h
+++ b/llvm/include/llvm/DebugInfo/Symbolize/Symbolize.h
@@ -49,7 +49,6 @@ class LLVMSymbolizer {
     std::string FallbackDebugPath;
     std::string DWPName;
     std::vector<std::string> DebugFileDirectory;
-    std::function<void(Error)> RecoverableErrorHandler = WithColor::defaultErrorHandler;
   };
 
   LLVMSymbolizer() = default;
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h b/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h
index 41827181e30cf..6b48cc447e131 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPGridValues.h
@@ -28,29 +28,30 @@ namespace omp {
 ///        use the new array name.
 ///
 /// Example usage in clang:
-///   const unsigned slot_size = ctx.GetTargetInfo().getGridValue(GV_Warp_Size);
+///   const unsigned slot_size =
+///   ctx.GetTargetInfo().getGridValue(llvm::omp::GVIDX::GV_Warp_Size);
 ///
 /// Example usage in libomptarget/deviceRTLs:
-///   #include "OMPGridValues.h"
+///   #include "llvm/Frontend/OpenMP/OMPGridValues.h"
 ///   #ifdef __AMDGPU__
 ///     #define GRIDVAL AMDGPUGpuGridValues
 ///   #else
 ///     #define GRIDVAL NVPTXGpuGridValues
 ///   #endif
 ///   ... Then use this reference for GV_Warp_Size in the deviceRTL source.
-///   GRIDVAL[GV_Warp_Size]
+///   llvm::omp::GRIDVAL[llvm::omp::GVIDX::GV_Warp_Size]
 ///
 /// Example usage in libomptarget hsa plugin:
-///   #include "OMPGridValues.h"
+///   #include "llvm/Frontend/OpenMP/OMPGridValues.h"
 ///   #define GRIDVAL AMDGPUGpuGridValues
 ///   ... Then use this reference to access GV_Warp_Size in the hsa plugin.
-///   GRIDVAL[GV_Warp_Size]
+///   llvm::omp::GRIDVAL[llvm::omp::GVIDX::GV_Warp_Size]
 ///
 /// Example usage in libomptarget cuda plugin:
-///    #include "OMPGridValues.h"
+///    #include "llvm/Frontend/OpenMP/OMPGridValues.h"
 ///    #define GRIDVAL NVPTXGpuGridValues
 ///   ... Then use this reference to access GV_Warp_Size in the cuda plugin.
-///    GRIDVAL[GV_Warp_Size]
+///    llvm::omp::GRIDVAL[llvm::omp::GVIDX::GV_Warp_Size]
 ///
 enum GVIDX {
   /// The maximum number of workers in a kernel.
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
index 3228c1bd538bc..e814de615866b 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPIRBuilder.h
@@ -226,6 +226,9 @@ class OpenMPIRBuilder {
                           omp::IdentFlag Flags = omp::IdentFlag(0),
                           unsigned Reserve2Flags = 0);
 
+  // Get the type corresponding to __kmpc_impl_lanemask_t from the deviceRTL
+  Type *getLanemaskType();
+
   /// Generate control flow and cleanup for cancellation.
   ///
   /// \param CancelFlag Flag indicating if the cancellation is performed.
diff --git a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
index 337e942955b11..b7a1558d6721b 100644
--- a/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
+++ b/llvm/include/llvm/Frontend/OpenMP/OMPKinds.def
@@ -153,6 +153,7 @@ __OMP_TYPE(Int32Ptr)
 __OMP_TYPE(Int64Ptr)
 
 OMP_TYPE(SizeTy, M.getDataLayout().getIntPtrType(Ctx))
+OMP_TYPE(LanemaskTy, getLanemaskType())
 
 #define __OMP_PTR_TYPE(NAME, BASE) OMP_TYPE(NAME, BASE->getPointerTo())
 
@@ -553,8 +554,9 @@ __OMP_RTL(__kmpc_get_team_static_memory, false, Void, Int16, VoidPtr, SizeTy,
           Int16, VoidPtrPtr)
 __OMP_RTL(__kmpc_restore_team_static_memory, false, Void, Int16, Int16)
 __OMP_RTL(__kmpc_barrier_simple_spmd, false, Void, IdentPtr, Int32)
-__OMP_RTL(__kmpc_warp_active_thread_mask, false, Int32, )
-__OMP_RTL(__kmpc_syncwarp, false, Void, Int32)
+
+__OMP_RTL(__kmpc_warp_active_thread_mask, false, LanemaskTy,)
+__OMP_RTL(__kmpc_syncwarp, false, Void, LanemaskTy)
 
 __OMP_RTL(__last, false, Void, )
 
diff --git a/llvm/include/llvm/IR/IRBuilder.h b/llvm/include/llvm/IR/IRBuilder.h
index 8d11643da03a0..e988eecfefbbe 100644
--- a/llvm/include/llvm/IR/IRBuilder.h
+++ b/llvm/include/llvm/IR/IRBuilder.h
@@ -814,7 +814,7 @@ class IRBuilderBase {
   /// start a new statepoint sequence.
   CallInst *CreateGCStatepointCall(uint64_t ID, uint32_t NumPatchBytes,
                                    Value *ActualCallee, uint32_t Flags,
-                                   ArrayRef<Use> CallArgs,
+                                   ArrayRef<Value *> CallArgs,
                                    Optional<ArrayRef<Use>> TransitionArgs,
                                    Optional<ArrayRef<Use>> DeoptArgs,
                                    ArrayRef<Value *> GCArgs,
@@ -843,7 +843,7 @@ class IRBuilderBase {
   InvokeInst *CreateGCStatepointInvoke(
       uint64_t ID, uint32_t NumPatchBytes, Value *ActualInvokee,
       BasicBlock *NormalDest, BasicBlock *UnwindDest, uint32_t Flags,
-      ArrayRef<Use> InvokeArgs, Optional<ArrayRef<Use>> TransitionArgs,
+      ArrayRef<Value *> InvokeArgs, Optional<ArrayRef<Use>> TransitionArgs,
       Optional<ArrayRef<Use>> DeoptArgs, ArrayRef<Value *> GCArgs,
       const Twine &Name = "");
 
diff --git a/llvm/include/llvm/IR/IntrinsicsX86.td b/llvm/include/llvm/IR/IntrinsicsX86.td
index 014fb71b726d3..d426d866f576b 100644
--- a/llvm/include/llvm/IR/IntrinsicsX86.td
+++ b/llvm/include/llvm/IR/IntrinsicsX86.td
@@ -5043,3 +5043,17 @@ let TargetPrefix = "x86" in {
                         [ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<1>>,
                          ImmArg<ArgIndex<2>>]>;
 }
+
+//===----------------------------------------------------------------------===//
+// UINTR - User Level Interrupt
+
+let TargetPrefix = "x86" in {
+  def int_x86_clui : GCCBuiltin<"__builtin_ia32_clui">,
+              Intrinsic<[], [], []>;
+  def int_x86_stui : GCCBuiltin<"__builtin_ia32_stui">,
+              Intrinsic<[], [], []>;
+  def int_x86_testui : GCCBuiltin<"__builtin_ia32_testui">,
+              Intrinsic<[llvm_i8_ty], [], []>;
+  def int_x86_senduipi : GCCBuiltin<"__builtin_ia32_senduipi">,
+              Intrinsic<[], [llvm_i64_ty], []>;
+}
diff --git a/llvm/include/llvm/InitializePasses.h b/llvm/include/llvm/InitializePasses.h
index f34f361eef510..2d89b8650f266 100644
--- a/llvm/include/llvm/InitializePasses.h
+++ b/llvm/include/llvm/InitializePasses.h
@@ -83,7 +83,7 @@ void initializeBasicBlockSectionsPass(PassRegistry &);
 void initializeBDCELegacyPassPass(PassRegistry&);
 void initializeBarrierNoopPass(PassRegistry&);
 void initializeBasicAAWrapperPassPass(PassRegistry&);
-void initializeBlockExtractorPass(PassRegistry &);
+void initializeBlockExtractorLegacyPassPass(PassRegistry &);
 void initializeBlockFrequencyInfoWrapperPassPass(PassRegistry&);
 void initializeBoundsCheckingLegacyPassPass(PassRegistry&);
 void initializeBranchFolderPassPass(PassRegistry&);
diff --git a/llvm/include/llvm/LTO/Config.h b/llvm/include/llvm/LTO/Config.h
index 0a3e523164601..a305778a305b0 100644
--- a/llvm/include/llvm/LTO/Config.h
+++ b/llvm/include/llvm/LTO/Config.h
@@ -1,4 +1,4 @@
-//===-Config.h - LLVM Link Time Optimizer Configuration -------------------===//
+//===-Config.h - LLVM Link Time Optimizer Configuration ---------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
diff --git a/llvm/include/llvm/Support/MachineValueType.h b/llvm/include/llvm/Support/MachineValueType.h
index 33b0fe6750c62..011926e39a251 100644
--- a/llvm/include/llvm/Support/MachineValueType.h
+++ b/llvm/include/llvm/Support/MachineValueType.h
@@ -425,6 +425,36 @@ namespace llvm {
               SimpleTy == MVT::iPTRAny);
     }
 
+    /// Return a vector with the same number of elements as this vector, but
+    /// with the element type converted to an integer type with the same
+    /// bitwidth.
+    MVT changeVectorElementTypeToInteger() const {
+      MVT EltTy = getVectorElementType();
+      MVT IntTy = MVT::getIntegerVT(EltTy.getSizeInBits());
+      MVT VecTy = MVT::getVectorVT(IntTy, getVectorElementCount());
+      assert(VecTy.SimpleTy != MVT::INVALID_SIMPLE_VALUE_TYPE &&
+             "Simple vector VT not representable by simple integer vector VT!");
+      return VecTy;
+    }
+
+    /// Return a VT for a vector type whose attributes match ourselves
+    /// with the exception of the element type that is chosen by the caller.
+    MVT changeVectorElementType(MVT EltVT) const {
+      MVT VecTy = MVT::getVectorVT(EltVT, getVectorElementCount());
+      assert(VecTy.SimpleTy != MVT::INVALID_SIMPLE_VALUE_TYPE &&
+             "Simple vector VT not representable by simple integer vector VT!");
+      return VecTy;
+    }
+
+    /// Return the type converted to an equivalently sized integer or vector
+    /// with integer element type. Similar to changeVectorElementTypeToInteger,
+    /// but also handles scalars.
+    MVT changeTypeToInteger() {
+      if (isVector())
+        return changeVectorElementTypeToInteger();
+      return MVT::getIntegerVT(getSizeInBits());
+    }
+
     /// Return a VT for a vector type with the same element type but
     /// half the number of elements.
     MVT getHalfNumVectorElementsVT() const {
diff --git a/llvm/include/llvm/Support/X86TargetParser.def b/llvm/include/llvm/Support/X86TargetParser.def
index 162a7f3a10d02..4a0ba26f49f98 100644
--- a/llvm/include/llvm/Support/X86TargetParser.def
+++ b/llvm/include/llvm/Support/X86TargetParser.def
@@ -178,6 +178,7 @@ X86_FEATURE       (SHA,             "sha")
 X86_FEATURE       (SHSTK,           "shstk")
 X86_FEATURE       (TBM,             "tbm")
 X86_FEATURE       (TSXLDTRK,        "tsxldtrk")
+X86_FEATURE       (UINTR,           "uintr")
 X86_FEATURE       (VAES,            "vaes")
 X86_FEATURE       (VZEROUPPER,      "vzeroupper")
 X86_FEATURE       (WAITPKG,         "waitpkg")
diff --git a/llvm/include/llvm/TableGen/Record.h b/llvm/include/llvm/TableGen/Record.h
index c0eb88b56be30..c8cd00204172a 100644
--- a/llvm/include/llvm/TableGen/Record.h
+++ b/llvm/include/llvm/TableGen/Record.h
@@ -420,7 +420,7 @@ inline raw_ostream &operator<<(raw_ostream &OS, const Init &I) {
 }
 
 /// This is the common superclass of types that have a specific,
-/// explicit, type, stored in ValueTy.
+/// explicit type, stored in ValueTy.
 class TypedInit : public Init {
   RecTy *ValueTy;
 
@@ -437,6 +437,7 @@ class TypedInit : public Init {
            I->getKind() <= IK_LastTypedInit;
   }
 
+  /// Get the type of the Init as a RecTy.
   RecTy *getType() const { return ValueTy; }
 
   Init *getCastTo(RecTy *Ty) const override;
@@ -451,7 +452,7 @@ class TypedInit : public Init {
   RecTy *getFieldType(StringInit *FieldName) const override;
 };
 
-/// '?' - Represents an uninitialized value
+/// '?' - Represents an uninitialized value.
 class UnsetInit : public Init {
   UnsetInit() : Init(IK_UnsetInit) {}
 
@@ -463,6 +464,7 @@ class UnsetInit : public Init {
     return I->getKind() == IK_UnsetInit;
   }
 
+  /// Get the singleton unset Init.
   static UnsetInit *get();
 
   Init *getCastTo(RecTy *Ty) const override;
@@ -472,8 +474,12 @@ class UnsetInit : public Init {
     return const_cast<UnsetInit*>(this);
   }
 
+  /// Is this a complete value with no unset (uninitialized) subvalues?
   bool isComplete() const override { return false; }
+
   bool isConcrete() const override { return true; }
+
+  /// Get the string representation of the Init.
   std::string getAsString() const override { return "?"; }
 };
 
@@ -1395,6 +1401,8 @@ class DagInit final : public TypedInit, public FoldingSetNode,
 //  High-Level Classes
 //===----------------------------------------------------------------------===//
 
+/// This class represents a field in a record, including its name, type,
+/// value, and source location.
 class RecordVal {
   friend class Record;
 
@@ -1407,22 +1415,37 @@ class RecordVal {
   RecordVal(Init *N, RecTy *T, bool P);
   RecordVal(Init *N, SMLoc Loc, RecTy *T, bool P);
 
+  /// Get the name of the field as a StringRef.
   StringRef getName() const;
+
+  /// Get the name of the field as an Init.
   Init *getNameInit() const { return Name; }
 
+  /// Get the name of the field as a std::string.
   std::string getNameInitAsString() const {
     return getNameInit()->getAsUnquotedString();
   }
 
+  /// Get the source location of the point where the field was defined.
   const SMLoc &getLoc() const { return Loc; }
+
   bool getPrefix() const { return TyAndPrefix.getInt(); }
+
+  /// Get the type of the field value as a RecTy.
   RecTy *getType() const { return TyAndPrefix.getPointer(); }
+
+  /// Get the value of the field as an Init.
   Init *getValue() const { return Value; }
 
+  /// Set the value of the field from an Init.
   bool setValue(Init *V);
+
+  /// Set the value and source location of the field.
   bool setValue(Init *V, SMLoc NewLoc);
 
   void dump() const;
+
+  /// Print the value to an output stream, possibly with a semicolon.
   void print(raw_ostream &OS, bool PrintSem = true) const;
 };
 
diff --git a/llvm/include/llvm/Target/TargetOptions.h b/llvm/include/llvm/Target/TargetOptions.h
index 049f5eff25e05..4c27ce0e09b8b 100644
--- a/llvm/include/llvm/Target/TargetOptions.h
+++ b/llvm/include/llvm/Target/TargetOptions.h
@@ -73,6 +73,12 @@ namespace llvm {
     None    // Do not use Basic Block Sections.
   };
 
+  enum class StackProtectorGuards {
+    None,
+    TLS,
+    Global
+  };
+
   enum class EABI {
     Unknown,
     Default, // Default means not specified
@@ -307,6 +313,16 @@ namespace llvm {
     /// Emit XRay Function Index section
     unsigned XRayOmitFunctionIndex : 1;
 
+    /// Stack protector guard offset to use.
+    unsigned StackProtectorGuardOffset : 32;
+
+    /// Stack protector guard mode to use, e.g. tls, global.
+    StackProtectorGuards StackProtectorGuard =
+                                         StackProtectorGuards::None;
+
+    /// Stack protector guard reg to use, e.g. usually fs or gs in X86.
+    std::string StackProtectorGuardReg = "None";
+
     /// FloatABIType - This setting is set by -float-abi=xxx option is specfied
     /// on the command line. This setting may either be Default, Soft, or Hard.
     /// Default selects the target's default behavior. Soft selects the ABI for
diff --git a/llvm/include/llvm/Transforms/IPO/BlockExtractor.h b/llvm/include/llvm/Transforms/IPO/BlockExtractor.h
new file mode 100644
index 0000000000000..deeb5ebe23d9a
--- /dev/null
+++ b/llvm/include/llvm/Transforms/IPO/BlockExtractor.h
@@ -0,0 +1,25 @@
+//===- BlockExtractor.h - Extracts blocks into their own functions --------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This pass extracts the specified basic blocks from the module into their
+// own functions.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_TRANSFORMS_IPO_BLOCKEXTRACTOR_H
+#define LLVM_TRANSFORMS_IPO_BLOCKEXTRACTOR_H
+
+#include "llvm/IR/PassManager.h"
+
+namespace llvm {
+struct BlockExtractorPass : PassInfoMixin<BlockExtractorPass> {
+  PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
+};
+} // namespace llvm
+
+#endif // LLVM_TRANSFORMS_IPO_BLOCKEXTRACTOR_H
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index b730eb33960ad..afa4af977c95b 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -2849,6 +2849,42 @@ static Value *simplifyICmpWithBinOpOnLHS(
   return nullptr;
 }
 
+
+// If only one of the icmp's operands has NSW flags, try to prove that:
+//
+//   icmp slt (x + C1), (x +nsw C2)
+//
+// is equivalent to:
+//
+//   icmp slt C1, C2
+//
+// which is true if x + C2 has the NSW flags set and:
+// *) C1 < C2 && C1 >= 0, or
+// *) C2 < C1 && C1 <= 0.
+//
+static bool trySimplifyICmpWithAdds(CmpInst::Predicate Pred, Value *LHS,
+                                    Value *RHS) {
+  // TODO: only support icmp slt for now.
+  if (Pred != CmpInst::ICMP_SLT)
+    return false;
+
+  // Canonicalize nsw add as RHS.
+  if (!match(RHS, m_NSWAdd(m_Value(), m_Value())))
+    std::swap(LHS, RHS);
+  if (!match(RHS, m_NSWAdd(m_Value(), m_Value())))
+    return false;
+
+  Value *X;
+  const APInt *C1, *C2;
+  if (!match(LHS, m_c_Add(m_Value(X), m_APInt(C1))) ||
+      !match(RHS, m_c_Add(m_Specific(X), m_APInt(C2))))
+    return false;
+
+  return (C1->slt(*C2) && C1->isNonNegative()) ||
+         (C2->slt(*C1) && C1->isNonPositive());
+}
+
+
 /// TODO: A large part of this logic is duplicated in InstCombine's
 /// foldICmpBinOp(). We should be able to share that and avoid the code
 /// duplication.
@@ -2898,8 +2934,9 @@ static Value *simplifyICmpWithBinOp(CmpInst::Predicate Pred, Value *LHS,
         return V;
 
     // icmp (X+Y), (X+Z) -> icmp Y,Z for equalities or if there is no overflow.
-    if (A && C && (A == C || A == D || B == C || B == D) && NoLHSWrapProblem &&
-        NoRHSWrapProblem) {
+    bool CanSimplify = (NoLHSWrapProblem && NoRHSWrapProblem) ||
+                       trySimplifyICmpWithAdds(Pred, LHS, RHS);
+    if (A && C && (A == C || A == D || B == C || B == D) && CanSimplify) {
       // Determine Y and Z in the form icmp (X+Y), (X+Z).
       Value *Y, *Z;
       if (A == C) {
diff --git a/llvm/lib/Analysis/RegionPass.cpp b/llvm/lib/Analysis/RegionPass.cpp
index ace682465abba..a73607dbef61b 100644
--- a/llvm/lib/Analysis/RegionPass.cpp
+++ b/llvm/lib/Analysis/RegionPass.cpp
@@ -31,8 +31,6 @@ char RGPassManager::ID = 0;
 
 RGPassManager::RGPassManager()
   : FunctionPass(ID), PMDataManager() {
-  skipThisRegion = false;
-  redoThisRegion = false;
   RI = nullptr;
   CurrentRegion = nullptr;
 }
@@ -76,8 +74,6 @@ bool RGPassManager::runOnFunction(Function &F) {
   while (!RQ.empty()) {
 
     CurrentRegion  = RQ.back();
-    skipThisRegion = false;
-    redoThisRegion = false;
 
     // Run all passes on the current Region.
     for (unsigned Index = 0; Index < getNumContainedPasses(); ++Index) {
@@ -115,54 +111,36 @@ bool RGPassManager::runOnFunction(Function &F) {
       if (isPassDebuggingExecutionsOrMore()) {
         if (LocalChanged)
           dumpPassInfo(P, MODIFICATION_MSG, ON_REGION_MSG,
-                       skipThisRegion ? "<deleted>" :
                                       CurrentRegion->getNameStr());
         dumpPreservedSet(P);
       }
 
-      if (!skipThisRegion) {
-        // Manually check that this region is still healthy. This is done
-        // instead of relying on RegionInfo::verifyRegion since RegionInfo
-        // is a function pass and it's really expensive to verify every
-        // Region in the function every time. That level of checking can be
-        // enabled with the -verify-region-info option.
-        {
-          TimeRegion PassTimer(getPassTimer(P));
-          CurrentRegion->verifyRegion();
-        }
-
-        // Then call the regular verifyAnalysis functions.
-        verifyPreservedAnalysis(P);
+      // Manually check that this region is still healthy. This is done
+      // instead of relying on RegionInfo::verifyRegion since RegionInfo
+      // is a function pass and it's really expensive to verify every
+      // Region in the function every time. That level of checking can be
+      // enabled with the -verify-region-info option.
+      {
+        TimeRegion PassTimer(getPassTimer(P));
+        CurrentRegion->verifyRegion();
       }
 
+      // Then call the regular verifyAnalysis functions.
+      verifyPreservedAnalysis(P);
+
       if (LocalChanged)
         removeNotPreservedAnalysis(P);
       recordAvailableAnalysis(P);
       removeDeadPasses(P,
-                       (!isPassDebuggingExecutionsOrMore() || skipThisRegion) ?
-                       "<deleted>" :  CurrentRegion->getNameStr(),
+                       (!isPassDebuggingExecutionsOrMore())
+                           ? "<deleted>"
+                           : CurrentRegion->getNameStr(),
                        ON_REGION_MSG);
-
-      if (skipThisRegion)
-        // Do not run other passes on this region.
-        break;
     }
 
-    // If the region was deleted, release all the region passes. This frees up
-    // some memory, and avoids trouble with the pass manager trying to call
-    // verifyAnalysis on them.
-    if (skipThisRegion)
-      for (unsigned Index = 0; Index < getNumContainedPasses(); ++Index) {
-        Pass *P = getContainedPass(Index);
-        freePass(P, "<deleted>", ON_REGION_MSG);
-      }
-
     // Pop the region from queue after running all passes.
     RQ.pop_back();
 
-    if (redoThisRegion)
-      RQ.push_back(CurrentRegion);
-
     // Free all region nodes created in region passes.
     RI->clearNodeCache();
   }
diff --git a/llvm/lib/Analysis/ScalarEvolution.cpp b/llvm/lib/Analysis/ScalarEvolution.cpp
index b9c6cf8f13380..55d759568b4ec 100644
--- a/llvm/lib/Analysis/ScalarEvolution.cpp
+++ b/llvm/lib/Analysis/ScalarEvolution.cpp
@@ -5527,17 +5527,6 @@ ScalarEvolution::getRangeRef(const SCEV *S,
         ConservativeResult =
             ConservativeResult.intersectWith(RangeFromFactoring, RangeType);
       }
-
-      // Now try symbolic BE count and more powerful methods.
-      MaxBECount = computeMaxBackedgeTakenCount(AddRec->getLoop());
-      if (!isa<SCEVCouldNotCompute>(MaxBECount) &&
-          getTypeSizeInBits(MaxBECount->getType()) <= BitWidth &&
-          AddRec->hasNoSelfWrap()) {
-        auto RangeFromAffineNew = getRangeForAffineNoSelfWrappingAR(
-            AddRec, MaxBECount, BitWidth, SignHint);
-        ConservativeResult =
-            ConservativeResult.intersectWith(RangeFromAffineNew, RangeType);
-      }
     }
 
     return setRange(AddRec, SignHint, std::move(ConservativeResult));
@@ -5707,70 +5696,6 @@ ConstantRange ScalarEvolution::getRangeForAffineAR(const SCEV *Start,
   return SR.intersectWith(UR, ConstantRange::Smallest);
 }
 
-ConstantRange ScalarEvolution::getRangeForAffineNoSelfWrappingAR(
-    const SCEVAddRecExpr *AddRec, const SCEV *MaxBECount, unsigned BitWidth,
-    ScalarEvolution::RangeSignHint SignHint) {
-  assert(AddRec->isAffine() && "Non-affine AddRecs are not suppored!\n");
-  assert(AddRec->hasNoSelfWrap() &&
-         "This only works for non-self-wrapping AddRecs!");
-  const bool IsSigned = SignHint == HINT_RANGE_SIGNED;
-  const SCEV *Step = AddRec->getStepRecurrence(*this);
-  // Let's make sure that we can prove that we do not self-wrap during
-  // MaxBECount iterations. We need this because MaxBECount is a maximum
-  // iteration count estimate, and we might infer nw from some exit for which we
-  // do not know max exit count (or any other side reasoning).
-  // TODO: Turn into assert at some point.
-  MaxBECount = getNoopOrZeroExtend(MaxBECount, AddRec->getType());
-  const SCEV *RangeWidth = getNegativeSCEV(getOne(AddRec->getType()));
-  const SCEV *StepAbs = getUMinExpr(Step, getNegativeSCEV(Step));
-  const SCEV *MaxItersWithoutWrap = getUDivExpr(RangeWidth, StepAbs);
-  if (!isKnownPredicate(ICmpInst::ICMP_ULE, MaxBECount, MaxItersWithoutWrap))
-    return ConstantRange::getFull(BitWidth);
-
-  ICmpInst::Predicate LEPred =
-      IsSigned ? ICmpInst::ICMP_SLE : ICmpInst::ICMP_ULE;
-  ICmpInst::Predicate GEPred =
-      IsSigned ? ICmpInst::ICMP_SGE : ICmpInst::ICMP_UGE;
-  const SCEV *Start = AddRec->getStart();
-  const SCEV *End = AddRec->evaluateAtIteration(MaxBECount, *this);
-
-  // We know that there is no self-wrap. Let's take Start and End values and
-  // look at all intermediate values V1, V2, ..., Vn that IndVar takes during
-  // the iteration. They either lie inside the range [Min(Start, End),
-  // Max(Start, End)] or outside it:
-  //
-  // Case 1:   RangeMin    ...    Start V1 ... VN End ...           RangeMax;
-  // Case 2:   RangeMin Vk ... V1 Start    ...    End Vn ... Vk + 1 RangeMax;
-  //
-  // No self wrap flag guarantees that the intermediate values cannot be BOTH
-  // outside and inside the range [Min(Start, End), Max(Start, End)]. Using that
-  // knowledge, let's try to prove that we are dealing with Case 1. It is so if
-  // Start <= End and step is positive, or Start >= End and step is negative.
-  ConstantRange StartRange =
-      IsSigned ? getSignedRange(Start) : getUnsignedRange(Start);
-  ConstantRange EndRange =
-      IsSigned ? getSignedRange(End) : getUnsignedRange(End);
-  ConstantRange RangeBetween = StartRange.unionWith(EndRange);
-  // If they already cover full iteration space, we will know nothing useful
-  // even if we prove what we want to prove.
-  if (RangeBetween.isFullSet())
-    return RangeBetween;
-  // Only deal with ranges that do not wrap (i.e. RangeMin < RangeMax).
-  bool IsWrappingRange =
-      IsSigned ? RangeBetween.getLower().sge(RangeBetween.getUpper())
-               : RangeBetween.getLower().uge(RangeBetween.getUpper());
-  if (IsWrappingRange)
-    return ConstantRange::getFull(BitWidth);
-
-  if (isKnownPositive(Step) &&
-      isKnownPredicateViaConstantRanges(LEPred, Start, End))
-    return RangeBetween;
-  else if (isKnownNegative(Step) &&
-           isKnownPredicateViaConstantRanges(GEPred, Start, End))
-    return RangeBetween;
-  return ConstantRange::getFull(BitWidth);
-}
-
 ConstantRange ScalarEvolution::getRangeViaFactoring(const SCEV *Start,
                                                     const SCEV *Step,
                                                     const SCEV *MaxBECount,
@@ -6585,7 +6510,7 @@ const SCEV *ScalarEvolution::getExitCount(const Loop *L,
   case Exact:
     return getBackedgeTakenInfo(L).getExact(ExitingBlock, this);
   case ConstantMaximum:
-    return getBackedgeTakenInfo(L).getMax(ExitingBlock, this);
+    return getBackedgeTakenInfo(L).getConstantMax(ExitingBlock, this);
   };
   llvm_unreachable("Invalid ExitCountKind!");
 }
@@ -6602,13 +6527,13 @@ const SCEV *ScalarEvolution::getBackedgeTakenCount(const Loop *L,
   case Exact:
     return getBackedgeTakenInfo(L).getExact(L, this);
   case ConstantMaximum:
-    return getBackedgeTakenInfo(L).getMax(this);
+    return getBackedgeTakenInfo(L).getConstantMax(this);
   };
   llvm_unreachable("Invalid ExitCountKind!");
 }
 
 bool ScalarEvolution::isBackedgeTakenCountMaxOrZero(const Loop *L) {
-  return getBackedgeTakenInfo(L).isMaxOrZero(this);
+  return getBackedgeTakenInfo(L).isConstantMaxOrZero(this);
 }
 
 /// Push PHI nodes in the header of the given loop onto the given Worklist.
@@ -6662,12 +6587,11 @@ ScalarEvolution::getBackedgeTakenInfo(const Loop *L) {
   const SCEV *BEExact = Result.getExact(L, this);
   if (BEExact != getCouldNotCompute()) {
     assert(isLoopInvariant(BEExact, L) &&
-           isLoopInvariant(Result.getMax(this), L) &&
+           isLoopInvariant(Result.getConstantMax(this), L) &&
            "Computed backedge-taken count isn't loop invariant for loop!");
     ++NumTripCountsComputed;
-  }
-  else if (Result.getMax(this) == getCouldNotCompute() &&
-           isa<PHINode>(L->getHeader()->begin())) {
+  } else if (Result.getConstantMax(this) == getCouldNotCompute() &&
+             isa<PHINode>(L->getHeader()->begin())) {
     // Only count loops that have phi nodes as not being computable.
     ++NumTripCountsNotComputed;
   }
@@ -6917,9 +6841,8 @@ ScalarEvolution::BackedgeTakenInfo::getExact(const BasicBlock *ExitingBlock,
   return SE->getCouldNotCompute();
 }
 
-const SCEV *
-ScalarEvolution::BackedgeTakenInfo::getMax(const BasicBlock *ExitingBlock,
-                                           ScalarEvolution *SE) const {
+const SCEV *ScalarEvolution::BackedgeTakenInfo::getConstantMax(
+    const BasicBlock *ExitingBlock, ScalarEvolution *SE) const {
   for (auto &ENT : ExitNotTaken)
     if (ENT.ExitingBlock == ExitingBlock && ENT.hasAlwaysTruePredicate())
       return ENT.MaxNotTaken;
@@ -6929,20 +6852,22 @@ ScalarEvolution::BackedgeTakenInfo::getMax(const BasicBlock *ExitingBlock,
 
 /// getMax - Get the max backedge taken count for the loop.
 const SCEV *
-ScalarEvolution::BackedgeTakenInfo::getMax(ScalarEvolution *SE) const {
+ScalarEvolution::BackedgeTakenInfo::getConstantMax(ScalarEvolution *SE) const {
   auto PredicateNotAlwaysTrue = [](const ExitNotTakenInfo &ENT) {
     return !ENT.hasAlwaysTruePredicate();
   };
 
-  if (any_of(ExitNotTaken, PredicateNotAlwaysTrue) || !getMax())
+  if (any_of(ExitNotTaken, PredicateNotAlwaysTrue) || !getConstantMax())
     return SE->getCouldNotCompute();
 
-  assert((isa<SCEVCouldNotCompute>(getMax()) || isa<SCEVConstant>(getMax())) &&
+  assert((isa<SCEVCouldNotCompute>(getConstantMax()) ||
+          isa<SCEVConstant>(getConstantMax())) &&
          "No point in having a non-constant max backedge taken count!");
-  return getMax();
+  return getConstantMax();
 }
 
-bool ScalarEvolution::BackedgeTakenInfo::isMaxOrZero(ScalarEvolution *SE) const {
+bool ScalarEvolution::BackedgeTakenInfo::isConstantMaxOrZero(
+    ScalarEvolution *SE) const {
   auto PredicateNotAlwaysTrue = [](const ExitNotTakenInfo &ENT) {
     return !ENT.hasAlwaysTruePredicate();
   };
@@ -6951,8 +6876,8 @@ bool ScalarEvolution::BackedgeTakenInfo::isMaxOrZero(ScalarEvolution *SE) const
 
 bool ScalarEvolution::BackedgeTakenInfo::hasOperand(const SCEV *S,
                                                     ScalarEvolution *SE) const {
-  if (getMax() && getMax() != SE->getCouldNotCompute() &&
-      SE->hasOperand(getMax(), S))
+  if (getConstantMax() && getConstantMax() != SE->getCouldNotCompute() &&
+      SE->hasOperand(getConstantMax(), S))
     return true;
 
   for (auto &ENT : ExitNotTaken)
@@ -7005,10 +6930,9 @@ ScalarEvolution::ExitLimit::ExitLimit(const SCEV *E, const SCEV *M,
 /// Allocate memory for BackedgeTakenInfo and copy the not-taken count of each
 /// computable exit into a persistent ExitNotTakenInfo array.
 ScalarEvolution::BackedgeTakenInfo::BackedgeTakenInfo(
-    ArrayRef<ScalarEvolution::BackedgeTakenInfo::EdgeExitInfo>
-        ExitCounts,
-    bool Complete, const SCEV *MaxCount, bool MaxOrZero)
-    : MaxAndComplete(MaxCount, Complete), MaxOrZero(MaxOrZero) {
+    ArrayRef<ScalarEvolution::BackedgeTakenInfo::EdgeExitInfo> ExitCounts,
+    bool IsComplete, const SCEV *ConstantMax, bool MaxOrZero)
+    : ConstantMax(ConstantMax), IsComplete(IsComplete), MaxOrZero(MaxOrZero) {
   using EdgeExitInfo = ScalarEvolution::BackedgeTakenInfo::EdgeExitInfo;
 
   ExitNotTaken.reserve(ExitCounts.size());
@@ -7028,7 +6952,8 @@ ScalarEvolution::BackedgeTakenInfo::BackedgeTakenInfo(
         return ExitNotTakenInfo(ExitBB, EL.ExactNotTaken, EL.MaxNotTaken,
                                 std::move(Predicate));
       });
-  assert((isa<SCEVCouldNotCompute>(MaxCount) || isa<SCEVConstant>(MaxCount)) &&
+  assert((isa<SCEVCouldNotCompute>(ConstantMax) ||
+          isa<SCEVConstant>(ConstantMax)) &&
          "No point in having a non-constant max backedge taken count!");
 }
 
diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 4dd536e76d065..9ecc5ce85e653 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1358,24 +1358,32 @@ static void computeKnownBitsFromOperator(const Operator *I,
   case Instruction::GetElementPtr: {
     // Analyze all of the subscripts of this getelementptr instruction
     // to determine if we can prove known low zero bits.
-    KnownBits LocalKnown(BitWidth);
-    computeKnownBits(I->getOperand(0), LocalKnown, Depth + 1, Q);
-    unsigned TrailZ = LocalKnown.countMinTrailingZeros();
+    computeKnownBits(I->getOperand(0), Known, Depth + 1, Q);
+    // Accumulate the constant indices in a separate variable
+    // to minimize the number of calls to computeForAddSub.
+    APInt AccConstIndices(BitWidth, 0, /*IsSigned*/ true);
 
     gep_type_iterator GTI = gep_type_begin(I);
+    // If the inbounds keyword is not present, the offsets are added to the
+    // base address with silently-wrapping two’s complement arithmetic.
+    bool IsInBounds = cast<GEPOperator>(I)->isInBounds();
     for (unsigned i = 1, e = I->getNumOperands(); i != e; ++i, ++GTI) {
       // TrailZ can only become smaller, short-circuit if we hit zero.
-      if (TrailZ == 0)
+      if (Known.isUnknown())
         break;
 
       Value *Index = I->getOperand(i);
+
+      // Handle case when index is zero.
+      Constant *CIndex = dyn_cast<Constant>(Index);
+      if (CIndex && CIndex->isZeroValue())
+        continue;
+
       if (StructType *STy = GTI.getStructTypeOrNull()) {
         // Handle struct member offset arithmetic.
 
-        // Handle case when index is vector zeroinitializer
-        Constant *CIndex = cast<Constant>(Index);
-        if (CIndex->isZeroValue())
-          continue;
+        assert(CIndex &&
+               "Access to structure field must be known at compile time");
 
         if (CIndex->getType()->isVectorTy())
           Index = CIndex->getSplatValue();
@@ -1383,26 +1391,58 @@ static void computeKnownBitsFromOperator(const Operator *I,
         unsigned Idx = cast<ConstantInt>(Index)->getZExtValue();
         const StructLayout *SL = Q.DL.getStructLayout(STy);
         uint64_t Offset = SL->getElementOffset(Idx);
-        TrailZ = std::min<unsigned>(TrailZ,
-                                    countTrailingZeros(Offset));
+        AccConstIndices += Offset;
+        continue;
+      }
+
+      // Handle array index arithmetic.
+      Type *IndexedTy = GTI.getIndexedType();
+      if (!IndexedTy->isSized()) {
+        Known.resetAll();
+        break;
+      }
+
+      unsigned IndexBitWidth = Index->getType()->getScalarSizeInBits();
+      KnownBits IndexBits(IndexBitWidth);
+      computeKnownBits(Index, IndexBits, Depth + 1, Q);
+      TypeSize IndexTypeSize = Q.DL.getTypeAllocSize(IndexedTy);
+      uint64_t TypeSizeInBytes = IndexTypeSize.getKnownMinSize();
+      KnownBits ScalingFactor(IndexBitWidth);
+      // Multiply by current sizeof type.
+      // &A[i] == A + i * sizeof(*A[i]).
+      if (IndexTypeSize.isScalable()) {
+        // For scalable types the only thing we know about sizeof is
+        // that this is a multiple of the minimum size.
+        ScalingFactor.Zero.setLowBits(countTrailingZeros(TypeSizeInBytes));
+      } else if (IndexBits.isConstant()) {
+        APInt IndexConst = IndexBits.getConstant();
+        APInt ScalingFactor(IndexBitWidth, TypeSizeInBytes);
+        IndexConst *= ScalingFactor;
+        AccConstIndices += IndexConst.sextOrTrunc(BitWidth);
+        continue;
       } else {
-        // Handle array index arithmetic.
-        Type *IndexedTy = GTI.getIndexedType();
-        if (!IndexedTy->isSized()) {
-          TrailZ = 0;
-          break;
-        }
-        unsigned GEPOpiBits = Index->getType()->getScalarSizeInBits();
-        uint64_t TypeSize = Q.DL.getTypeAllocSize(IndexedTy).getKnownMinSize();
-        LocalKnown.Zero = LocalKnown.One = APInt(GEPOpiBits, 0);
-        computeKnownBits(Index, LocalKnown, Depth + 1, Q);
-        TrailZ = std::min(TrailZ,
-                          unsigned(countTrailingZeros(TypeSize) +
-                                   LocalKnown.countMinTrailingZeros()));
+        ScalingFactor.Zero = ~TypeSizeInBytes;
+        ScalingFactor.One = TypeSizeInBytes;
       }
-    }
+      IndexBits = KnownBits::computeForMul(IndexBits, ScalingFactor);
+
+      // If the offsets have a different width from the pointer, according
+      // to the language reference we need to sign-extend or truncate them
+      // to the width of the pointer.
+      IndexBits = IndexBits.sextOrTrunc(BitWidth);
 
-    Known.Zero.setLowBits(TrailZ);
+      Known = KnownBits::computeForAddSub(
+          /*Add=*/true,
+          /*NSW=*/IsInBounds, Known, IndexBits);
+    }
+    if (!Known.isUnknown() && !AccConstIndices.isNullValue()) {
+      KnownBits Index(BitWidth);
+      Index.Zero = ~AccConstIndices;
+      Index.One = AccConstIndices;
+      Known = KnownBits::computeForAddSub(
+          /*Add=*/true,
+          /*NSW=*/IsInBounds, Known, Index);
+    }
     break;
   }
   case Instruction::PHI: {
diff --git a/llvm/lib/CodeGen/AsmPrinter/DbgEntityHistoryCalculator.cpp b/llvm/lib/CodeGen/AsmPrinter/DbgEntityHistoryCalculator.cpp
index df1a9617b87d9..86b8047746968 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DbgEntityHistoryCalculator.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DbgEntityHistoryCalculator.cpp
@@ -417,7 +417,7 @@ void llvm::calculateDbgEntityHistory(const MachineFunction *MF,
                                      DbgValueHistoryMap &DbgValues,
                                      DbgLabelInstrMap &DbgLabels) {
   const TargetLowering *TLI = MF->getSubtarget().getTargetLowering();
-  unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
+  Register SP = TLI->getStackPointerRegisterToSaveRestore();
   Register FrameReg = TRI->getFrameRegister(*MF);
   RegDescribedVarsMap RegVars;
   DbgValueEntriesMap LiveEntries;
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
index a315a7e2ca021..9c95ed414f2f8 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -695,7 +695,7 @@ static void interpretValues(const MachineInstr *CurMI,
                              ForwardedRegWorklist[ParamFwdReg], Params);
       } else if (ParamValue->first.isReg()) {
         Register RegLoc = ParamValue->first.getReg();
-        unsigned SP = TLI.getStackPointerRegisterToSaveRestore();
+        Register SP = TLI.getStackPointerRegisterToSaveRestore();
         Register FP = TRI.getFrameRegister(*MF);
         bool IsSPorFP = (RegLoc == SP) || (RegLoc == FP);
         if (TRI.isCalleeSavedPhysReg(RegLoc, *MF) || IsSPorFP) {
diff --git a/llvm/lib/CodeGen/BreakFalseDeps.cpp b/llvm/lib/CodeGen/BreakFalseDeps.cpp
index 69755358a71eb..b11db3e657703 100644
--- a/llvm/lib/CodeGen/BreakFalseDeps.cpp
+++ b/llvm/lib/CodeGen/BreakFalseDeps.cpp
@@ -171,8 +171,8 @@ bool BreakFalseDeps::pickBestRegisterForUndef(MachineInstr *MI, unsigned OpIdx,
 
 bool BreakFalseDeps::shouldBreakDependence(MachineInstr *MI, unsigned OpIdx,
                                            unsigned Pref) {
-  Register reg = MI->getOperand(OpIdx).getReg();
-  unsigned Clearance = RDA->getClearance(MI, reg);
+  MCRegister Reg = MI->getOperand(OpIdx).getReg().asMCReg();
+  unsigned Clearance = RDA->getClearance(MI, Reg);
   LLVM_DEBUG(dbgs() << "Clearance: " << Clearance << ", want " << Pref);
 
   if (Pref > Clearance) {
diff --git a/llvm/lib/CodeGen/CommandFlags.cpp b/llvm/lib/CodeGen/CommandFlags.cpp
index a9093586d696f..bb15ff286d7c9 100644
--- a/llvm/lib/CodeGen/CommandFlags.cpp
+++ b/llvm/lib/CodeGen/CommandFlags.cpp
@@ -76,6 +76,9 @@ CGOPT_EXP(bool, DataSections)
 CGOPT_EXP(bool, FunctionSections)
 CGOPT(bool, IgnoreXCOFFVisibility)
 CGOPT(std::string, BBSections)
+CGOPT(std::string, StackProtectorGuard)
+CGOPT(unsigned, StackProtectorGuardOffset)
+CGOPT(std::string, StackProtectorGuardReg)
 CGOPT(unsigned, TLSSize)
 CGOPT(bool, EmulatedTLS)
 CGOPT(bool, UniqueSectionNames)
@@ -348,6 +351,21 @@ codegen::RegisterCodeGenFlags::RegisterCodeGenFlags() {
       cl::init("none"));
   CGBINDOPT(BBSections);
 
+  static cl::opt<std::string> StackProtectorGuard(
+      "stack-protector-guard", cl::desc("Stack protector guard mode"),
+      cl::init("none"));
+  CGBINDOPT(StackProtectorGuard);
+
+  static cl::opt<std::string> StackProtectorGuardReg(
+      "stack-protector-guard-reg", cl::desc("Stack protector guard register"),
+      cl::init("none"));
+  CGBINDOPT(StackProtectorGuardReg);
+
+  static cl::opt<unsigned> StackProtectorGuardOffset(
+      "stack-protector-guard-offset", cl::desc("Stack protector guard offset"),
+      cl::init((unsigned)-1));
+  CGBINDOPT(StackProtectorGuardOffset);
+
   static cl::opt<unsigned> TLSSize(
       "tls-size", cl::desc("Bit size of immediate TLS offsets"), cl::init(0));
   CGBINDOPT(TLSSize);
@@ -459,6 +477,24 @@ codegen::getBBSectionsMode(llvm::TargetOptions &Options) {
   }
 }
 
+llvm::StackProtectorGuards
+codegen::getStackProtectorGuardMode(llvm::TargetOptions &Options) {
+  if (getStackProtectorGuard() == "tls")
+    return StackProtectorGuards::TLS;
+  if (getStackProtectorGuard() == "global")
+    return StackProtectorGuards::Global;
+  if (getStackProtectorGuard() != "none") {
+    ErrorOr<std::unique_ptr<MemoryBuffer>> MBOrErr =
+        MemoryBuffer::getFile(getStackProtectorGuard());
+    if (!MBOrErr)
+      errs() << "error illegal stack protector guard mode: "
+             << MBOrErr.getError().message() << "\n";
+    else
+      Options.BBSectionsFuncListBuf = std::move(*MBOrErr);
+  }
+  return StackProtectorGuards::None;
+}
+
 // Common utility function tightly tied to the options listed here. Initializes
 // a TargetOptions object with CodeGen flags and returns it.
 TargetOptions
@@ -493,6 +529,9 @@ codegen::InitTargetOptionsFromCodeGenFlags(const Triple &TheTriple) {
   Options.BBSections = getBBSectionsMode(Options);
   Options.UniqueSectionNames = getUniqueSectionNames();
   Options.UniqueBasicBlockSectionNames = getUniqueBasicBlockSectionNames();
+  Options.StackProtectorGuard = getStackProtectorGuardMode(Options);
+  Options.StackProtectorGuardOffset = getStackProtectorGuardOffset();
+  Options.StackProtectorGuardReg = getStackProtectorGuardReg();
   Options.TLSSize = getTLSSize();
   Options.EmulatedTLS = getEmulatedTLS();
   Options.ExplicitEmulatedTLS = EmulatedTLSView->getNumOccurrences() > 0;
diff --git a/llvm/lib/CodeGen/IfConversion.cpp b/llvm/lib/CodeGen/IfConversion.cpp
index 1a5c5d6850172..d149f8c3a139e 100644
--- a/llvm/lib/CodeGen/IfConversion.cpp
+++ b/llvm/lib/CodeGen/IfConversion.cpp
@@ -751,7 +751,7 @@ bool IfConverter::CountDuplicatedInstructions(
     // A pred-clobbering instruction in the shared portion prevents
     // if-conversion.
     std::vector<MachineOperand> PredDefs;
-    if (TII->DefinesPredicate(*TIB, PredDefs))
+    if (TII->ClobbersPredicate(*TIB, PredDefs, false))
       return false;
     // If we get all the way to the branch instructions, don't count them.
     if (!TIB->isBranch())
@@ -1146,7 +1146,7 @@ void IfConverter::ScanInstructions(BBInfo &BBI,
     // FIXME: Make use of PredDefs? e.g. ADDC, SUBC sets predicates but are
     // still potentially predicable.
     std::vector<MachineOperand> PredDefs;
-    if (TII->DefinesPredicate(MI, PredDefs))
+    if (TII->ClobbersPredicate(MI, PredDefs, true))
       BBI.ClobbersPred = true;
 
     if (!TII->isPredicable(MI)) {
diff --git a/llvm/lib/CodeGen/LiveDebugValues/InstrRefBasedImpl.cpp b/llvm/lib/CodeGen/LiveDebugValues/InstrRefBasedImpl.cpp
index 5c05fbd62949e..1713700081b63 100644
--- a/llvm/lib/CodeGen/LiveDebugValues/InstrRefBasedImpl.cpp
+++ b/llvm/lib/CodeGen/LiveDebugValues/InstrRefBasedImpl.cpp
@@ -497,7 +497,7 @@ class MLocTracker {
     // Always track SP. This avoids the implicit clobbering caused by regmasks
     // from affectings its values. (LiveDebugValues disbelieves calls and
     // regmasks that claim to clobber SP).
-    unsigned SP = TLI.getStackPointerRegisterToSaveRestore();
+    Register SP = TLI.getStackPointerRegisterToSaveRestore();
     if (SP) {
       unsigned ID = getLocID(SP, false);
       (void)lookupOrTrackRegister(ID);
@@ -506,8 +506,8 @@ class MLocTracker {
 
   /// Produce location ID number for indexing LocIDToLocIdx. Takes the register
   /// or spill number, and flag for whether it's a spill or not.
-  unsigned getLocID(unsigned RegOrSpill, bool isSpill) {
-    return (isSpill) ? RegOrSpill + NumRegs - 1 : RegOrSpill;
+  unsigned getLocID(Register RegOrSpill, bool isSpill) {
+    return (isSpill) ? RegOrSpill.id() + NumRegs - 1 : RegOrSpill.id();
   }
 
   /// Accessor for reading the value at Idx.
@@ -640,7 +640,7 @@ class MLocTracker {
   /// later.
   void writeRegMask(const MachineOperand *MO, unsigned CurBB, unsigned InstID) {
     // Ensure SP exists, so that we don't override it later.
-    unsigned SP = TLI.getStackPointerRegisterToSaveRestore();
+    Register SP = TLI.getStackPointerRegisterToSaveRestore();
 
     // Def any register we track have that isn't preserved. The regmask
     // terminates the liveness of a register, meaning its value can't be
@@ -1555,7 +1555,7 @@ void InstrRefBasedLDV::transferRegisterDef(MachineInstr &MI) {
 
   MachineFunction *MF = MI.getMF();
   const TargetLowering *TLI = MF->getSubtarget().getTargetLowering();
-  unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
+  Register SP = TLI->getStackPointerRegisterToSaveRestore();
 
   // Find the regs killed by MI, and find regmasks of preserved regs.
   // Max out the number of statically allocated elements in `DeadRegs`, as this
@@ -1994,7 +1994,7 @@ void InstrRefBasedLDV::produceTransferFunctions(
 
   // Compute a bitvector of all the registers that are tracked in this block.
   const TargetLowering *TLI = MF.getSubtarget().getTargetLowering();
-  unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
+  Register SP = TLI->getStackPointerRegisterToSaveRestore();
   BitVector UsedRegs(TRI->getNumRegs());
   for (auto Location : MTracker->locations()) {
     unsigned ID = MTracker->LocIdxToLocID[Location.Idx];
diff --git a/llvm/lib/CodeGen/LiveDebugVariables.cpp b/llvm/lib/CodeGen/LiveDebugVariables.cpp
index bd7024e8f483c..f4238bd381261 100644
--- a/llvm/lib/CodeGen/LiveDebugVariables.cpp
+++ b/llvm/lib/CodeGen/LiveDebugVariables.cpp
@@ -395,6 +395,11 @@ class LDVImpl {
   LiveIntervals *LIS;
   const TargetRegisterInfo *TRI;
 
+  using StashedInstrRef =
+      std::tuple<unsigned, unsigned, const DILocalVariable *,
+                 const DIExpression *, DebugLoc>;
+  std::map<SlotIndex, std::vector<StashedInstrRef>> StashedInstrReferences;
+
   /// Whether emitDebugValues is called.
   bool EmitDone = false;
 
@@ -431,6 +436,16 @@ class LDVImpl {
   /// \returns True if the DBG_VALUE instruction should be deleted.
   bool handleDebugValue(MachineInstr &MI, SlotIndex Idx);
 
+  /// Track a DBG_INSTR_REF. This needs to be removed from the MachineFunction
+  /// during regalloc -- but there's no need to maintain live ranges, as we
+  /// refer to a value rather than a location.
+  ///
+  /// \param MI DBG_INSTR_REF instruction
+  /// \param Idx Last valid SlotIndex before instruction
+  ///
+  /// \returns True if the DBG_VALUE instruction should be deleted.
+  bool handleDebugInstrRef(MachineInstr &MI, SlotIndex Idx);
+
   /// Add DBG_LABEL instruction to UserLabel.
   ///
   /// \param MI DBG_LABEL instruction
@@ -459,6 +474,7 @@ class LDVImpl {
   /// Release all memory.
   void clear() {
     MF = nullptr;
+    StashedInstrReferences.clear();
     userValues.clear();
     userLabels.clear();
     virtRegToEqClass.clear();
@@ -666,6 +682,19 @@ bool LDVImpl::handleDebugValue(MachineInstr &MI, SlotIndex Idx) {
   return true;
 }
 
+bool LDVImpl::handleDebugInstrRef(MachineInstr &MI, SlotIndex Idx) {
+  assert(MI.isDebugRef());
+  unsigned InstrNum = MI.getOperand(0).getImm();
+  unsigned OperandNum = MI.getOperand(1).getImm();
+  auto *Var = MI.getDebugVariable();
+  auto *Expr = MI.getDebugExpression();
+  auto &DL = MI.getDebugLoc();
+  StashedInstrRef Stashed =
+      std::make_tuple(InstrNum, OperandNum, Var, Expr, DL);
+  StashedInstrReferences[Idx].push_back(Stashed);
+  return true;
+}
+
 bool LDVImpl::handleDebugLabel(MachineInstr &MI, SlotIndex Idx) {
   // DBG_LABEL label
   if (MI.getNumOperands() != 1 || !MI.getOperand(0).isMetadata()) {
@@ -713,6 +742,7 @@ bool LDVImpl::collectDebugValues(MachineFunction &mf) {
         // Only handle DBG_VALUE in handleDebugValue(). Skip all other
         // kinds of debug instructions.
         if ((MBBI->isDebugValue() && handleDebugValue(*MBBI, Idx)) ||
+            (MBBI->isDebugRef() && handleDebugInstrRef(*MBBI, Idx)) ||
             (MBBI->isDebugLabel() && handleDebugLabel(*MBBI, Idx))) {
           MBBI = MBB->erase(MBBI);
           Changed = true;
@@ -1435,6 +1465,28 @@ void LDVImpl::emitDebugValues(VirtRegMap *VRM) {
     LLVM_DEBUG(userLabel->print(dbgs(), TRI));
     userLabel->emitDebugLabel(*LIS, *TII);
   }
+
+  LLVM_DEBUG(dbgs() << "********** EMITTING INSTR REFERENCES **********\n");
+
+  // Re-insert any DBG_INSTR_REFs back in the position they were. Ordering
+  // is preserved by vector.
+  auto Slots = LIS->getSlotIndexes();
+  const MCInstrDesc &RefII = TII->get(TargetOpcode::DBG_INSTR_REF);
+  for (auto &P : StashedInstrReferences) {
+    const SlotIndex &Idx = P.first;
+    auto *MBB = Slots->getMBBFromIndex(Idx);
+    MachineBasicBlock::iterator insertPos = findInsertLocation(MBB, Idx, *LIS);
+    for (auto &Stashed : P.second) {
+      auto MIB = BuildMI(*MF, std::get<4>(Stashed), RefII);
+      MIB.addImm(std::get<0>(Stashed));
+      MIB.addImm(std::get<1>(Stashed));
+      MIB.addMetadata(std::get<2>(Stashed));
+      MIB.addMetadata(std::get<3>(Stashed));
+      MachineInstr *New = MIB;
+      MBB->insert(insertPos, New);
+    }
+  }
+
   EmitDone = true;
 }
 
diff --git a/llvm/lib/CodeGen/MachineFunction.cpp b/llvm/lib/CodeGen/MachineFunction.cpp
index 2568448f3c92c..a7edc274dd237 100644
--- a/llvm/lib/CodeGen/MachineFunction.cpp
+++ b/llvm/lib/CodeGen/MachineFunction.cpp
@@ -955,7 +955,8 @@ void MachineFunction::makeDebugValueSubstitution(DebugInstrOperandPair A,
 }
 
 void MachineFunction::substituteDebugValuesForInst(const MachineInstr &Old,
-                                                   MachineInstr &New) {
+                                                   MachineInstr &New,
+                                                   unsigned MaxOperand) {
   // If the Old instruction wasn't tracked at all, there is no work to do.
   unsigned OldInstrNum = Old.peekDebugInstrNum();
   if (!OldInstrNum)
@@ -965,12 +966,16 @@ void MachineFunction::substituteDebugValuesForInst(const MachineInstr &Old,
   // Avoid creating new instr numbers unless we create a new substitution.
   // While this has no functional effect, it risks confusing someone reading
   // MIR output.
+  // Examine all the operands, or the first N specified by the caller.
+  MaxOperand = std::min(MaxOperand, Old.getNumOperands());
   for (unsigned int I = 0; I < Old.getNumOperands(); ++I) {
     const auto &OldMO = Old.getOperand(I);
+    auto &NewMO = New.getOperand(I);
+    (void)NewMO;
 
     if (!OldMO.isReg() || !OldMO.isDef())
       continue;
-    assert(Old.getOperand(I).isDef());
+    assert(NewMO.isDef());
 
     unsigned NewInstrNum = New.getDebugInstrNum();
     makeDebugValueSubstitution(std::make_pair(OldInstrNum, I),
diff --git a/llvm/lib/CodeGen/MachineVerifier.cpp b/llvm/lib/CodeGen/MachineVerifier.cpp
index cf48d292d9320..dc25146d7868c 100644
--- a/llvm/lib/CodeGen/MachineVerifier.cpp
+++ b/llvm/lib/CodeGen/MachineVerifier.cpp
@@ -102,10 +102,10 @@ namespace {
     bool isFunctionRegBankSelected;
     bool isFunctionSelected;
 
-    using RegVector = SmallVector<unsigned, 16>;
+    using RegVector = SmallVector<Register, 16>;
     using RegMaskVector = SmallVector<const uint32_t *, 4>;
-    using RegSet = DenseSet<unsigned>;
-    using RegMap = DenseMap<unsigned, const MachineInstr *>;
+    using RegSet = DenseSet<Register>;
+    using RegMap = DenseMap<Register, const MachineInstr *>;
     using BlockSet = SmallPtrSet<const MachineBasicBlock *, 8>;
 
     const MachineInstr *FirstNonPHI;
@@ -120,10 +120,10 @@ namespace {
     SlotIndex lastIndex;
 
     // Add Reg and any sub-registers to RV
-    void addRegWithSubRegs(RegVector &RV, unsigned Reg) {
+    void addRegWithSubRegs(RegVector &RV, Register Reg) {
       RV.push_back(Reg);
-      if (Register::isPhysicalRegister(Reg))
-        for (const MCPhysReg &SubReg : TRI->subregs(Reg))
+      if (Reg.isPhysical())
+        for (const MCPhysReg &SubReg : TRI->subregs(Reg.asMCReg()))
           RV.push_back(SubReg);
     }
 
@@ -159,8 +159,8 @@ namespace {
 
       // Add register to vregsRequired if it belongs there. Return true if
       // anything changed.
-      bool addRequired(unsigned Reg) {
-        if (!Register::isVirtualRegister(Reg))
+      bool addRequired(Register Reg) {
+        if (!Reg.isVirtual())
           return false;
         if (regsLiveOut.count(Reg))
           return false;
@@ -170,7 +170,7 @@ namespace {
       // Same for a full set.
       bool addRequired(const RegSet &RS) {
         bool Changed = false;
-        for (unsigned Reg : RS)
+        for (Register Reg : RS)
           Changed |= addRequired(Reg);
         return Changed;
       }
@@ -184,7 +184,7 @@ namespace {
       }
 
       // Live-out registers are either in regsLiveOut or vregsPassed.
-      bool isLiveOut(unsigned Reg) const {
+      bool isLiveOut(Register Reg) const {
         return regsLiveOut.count(Reg) || vregsPassed.count(Reg);
       }
     };
@@ -192,13 +192,13 @@ namespace {
     // Extra register info per MBB.
     DenseMap<const MachineBasicBlock*, BBInfo> MBBInfoMap;
 
-    bool isReserved(unsigned Reg) {
-      return Reg < regsReserved.size() && regsReserved.test(Reg);
+    bool isReserved(Register Reg) {
+      return Reg.id() < regsReserved.size() && regsReserved.test(Reg.id());
     }
 
-    bool isAllocatable(unsigned Reg) const {
-      return Reg < TRI->getNumRegs() && TRI->isInAllocatableClass(Reg) &&
-             !regsReserved.test(Reg);
+    bool isAllocatable(Register Reg) const {
+      return Reg.id() < TRI->getNumRegs() && TRI->isInAllocatableClass(Reg) &&
+             !regsReserved.test(Reg.id());
     }
 
     // Analysis information if available
@@ -226,7 +226,7 @@ namespace {
                 LLT MOVRegType = LLT{});
 
     void report_context(const LiveInterval &LI) const;
-    void report_context(const LiveRange &LR, unsigned VRegUnit,
+    void report_context(const LiveRange &LR, Register VRegUnit,
                         LaneBitmask LaneMask) const;
     void report_context(const LiveRange::Segment &S) const;
     void report_context(const VNInfo &VNI) const;
@@ -234,18 +234,19 @@ namespace {
     void report_context(MCPhysReg PhysReg) const;
     void report_context_liverange(const LiveRange &LR) const;
     void report_context_lanemask(LaneBitmask LaneMask) const;
-    void report_context_vreg(unsigned VReg) const;
-    void report_context_vreg_regunit(unsigned VRegOrUnit) const;
+    void report_context_vreg(Register VReg) const;
+    void report_context_vreg_regunit(Register VRegOrUnit) const;
 
     void verifyInlineAsm(const MachineInstr *MI);
 
     void checkLiveness(const MachineOperand *MO, unsigned MONum);
     void checkLivenessAtUse(const MachineOperand *MO, unsigned MONum,
-                            SlotIndex UseIdx, const LiveRange &LR, unsigned VRegOrUnit,
+                            SlotIndex UseIdx, const LiveRange &LR,
+                            Register VRegOrUnit,
                             LaneBitmask LaneMask = LaneBitmask::getNone());
     void checkLivenessAtDef(const MachineOperand *MO, unsigned MONum,
-                            SlotIndex DefIdx, const LiveRange &LR, unsigned VRegOrUnit,
-                            bool SubRangeCheck = false,
+                            SlotIndex DefIdx, const LiveRange &LR,
+                            Register VRegOrUnit, bool SubRangeCheck = false,
                             LaneBitmask LaneMask = LaneBitmask::getNone());
 
     void markReachable(const MachineBasicBlock *MBB);
@@ -256,12 +257,12 @@ namespace {
     void verifyLiveVariables();
     void verifyLiveIntervals();
     void verifyLiveInterval(const LiveInterval&);
-    void verifyLiveRangeValue(const LiveRange&, const VNInfo*, unsigned,
+    void verifyLiveRangeValue(const LiveRange &, const VNInfo *, Register,
                               LaneBitmask);
-    void verifyLiveRangeSegment(const LiveRange&,
-                                const LiveRange::const_iterator I, unsigned,
+    void verifyLiveRangeSegment(const LiveRange &,
+                                const LiveRange::const_iterator I, Register,
                                 LaneBitmask);
-    void verifyLiveRange(const LiveRange&, unsigned,
+    void verifyLiveRange(const LiveRange &, Register,
                          LaneBitmask LaneMask = LaneBitmask::getNone());
 
     void verifyStackFrame();
@@ -508,7 +509,7 @@ void MachineVerifier::report_context(const LiveInterval &LI) const {
   errs() << "- interval:    " << LI << '\n';
 }
 
-void MachineVerifier::report_context(const LiveRange &LR, unsigned VRegUnit,
+void MachineVerifier::report_context(const LiveRange &LR, Register VRegUnit,
                                      LaneBitmask LaneMask) const {
   report_context_liverange(LR);
   report_context_vreg_regunit(VRegUnit);
@@ -532,11 +533,11 @@ void MachineVerifier::report_context(MCPhysReg PReg) const {
   errs() << "- p. register: " << printReg(PReg, TRI) << '\n';
 }
 
-void MachineVerifier::report_context_vreg(unsigned VReg) const {
+void MachineVerifier::report_context_vreg(Register VReg) const {
   errs() << "- v. register: " << printReg(VReg, TRI) << '\n';
 }
 
-void MachineVerifier::report_context_vreg_regunit(unsigned VRegOrUnit) const {
+void MachineVerifier::report_context_vreg_regunit(Register VRegOrUnit) const {
   if (Register::isVirtualRegister(VRegOrUnit)) {
     report_context_vreg(VRegOrUnit);
   } else {
@@ -1958,8 +1959,10 @@ MachineVerifier::visitMachineOperand(const MachineOperand *MO, unsigned MONum) {
 }
 
 void MachineVerifier::checkLivenessAtUse(const MachineOperand *MO,
-    unsigned MONum, SlotIndex UseIdx, const LiveRange &LR, unsigned VRegOrUnit,
-    LaneBitmask LaneMask) {
+                                         unsigned MONum, SlotIndex UseIdx,
+                                         const LiveRange &LR,
+                                         Register VRegOrUnit,
+                                         LaneBitmask LaneMask) {
   LiveQueryResult LRQ = LR.Query(UseIdx);
   // Check if we have a segment at the use, note however that we only need one
   // live subregister range, the others may be dead.
@@ -1980,8 +1983,11 @@ void MachineVerifier::checkLivenessAtUse(const MachineOperand *MO,
 }
 
 void MachineVerifier::checkLivenessAtDef(const MachineOperand *MO,
-    unsigned MONum, SlotIndex DefIdx, const LiveRange &LR, unsigned VRegOrUnit,
-    bool SubRangeCheck, LaneBitmask LaneMask) {
+                                         unsigned MONum, SlotIndex DefIdx,
+                                         const LiveRange &LR,
+                                         Register VRegOrUnit,
+                                         bool SubRangeCheck,
+                                         LaneBitmask LaneMask) {
   if (const VNInfo *VNI = LR.getVNInfoAt(DefIdx)) {
     assert(VNI && "NULL valno is not allowed");
     if (VNI->def != DefIdx) {
@@ -2025,7 +2031,7 @@ void MachineVerifier::checkLivenessAtDef(const MachineOperand *MO,
 
 void MachineVerifier::checkLiveness(const MachineOperand *MO, unsigned MONum) {
   const MachineInstr *MI = MO->getParent();
-  const unsigned Reg = MO->getReg();
+  const Register Reg = MO->getReg();
 
   // Both use and def operands can read a register.
   if (MO->readsReg()) {
@@ -2043,8 +2049,9 @@ void MachineVerifier::checkLiveness(const MachineOperand *MO, unsigned MONum) {
     if (LiveInts && !LiveInts->isNotInMIMap(*MI)) {
       SlotIndex UseIdx = LiveInts->getInstructionIndex(*MI);
       // Check the cached regunit intervals.
-      if (Register::isPhysicalRegister(Reg) && !isReserved(Reg)) {
-        for (MCRegUnitIterator Units(Reg, TRI); Units.isValid(); ++Units) {
+      if (Reg.isPhysical() && !isReserved(Reg)) {
+        for (MCRegUnitIterator Units(Reg.asMCReg(), TRI); Units.isValid();
+             ++Units) {
           if (MRI->isReservedRegUnit(*Units))
             continue;
           if (const LiveRange *LR = LiveInts->getCachedRegUnit(*Units))
@@ -2190,9 +2197,9 @@ void MachineVerifier::visitMachineBundleAfter(const MachineInstr *MI) {
   // Kill any masked registers.
   while (!regMasks.empty()) {
     const uint32_t *Mask = regMasks.pop_back_val();
-    for (unsigned Reg : regsLive)
-      if (Register::isPhysicalRegister(Reg) &&
-          MachineOperand::clobbersPhysReg(Mask, Reg))
+    for (Register Reg : regsLive)
+      if (Reg.isPhysical() &&
+          MachineOperand::clobbersPhysReg(Mask, Reg.asMCReg()))
         regsDead.push_back(Reg);
   }
   set_subtract(regsLive, regsDead);   regsDead.clear();
@@ -2225,7 +2232,7 @@ struct VRegFilter {
   // Add elements to the filter itself. \pre Input set \p FromRegSet must have
   // no duplicates. Both virtual and physical registers are fine.
   template <typename RegSetT> void add(const RegSetT &FromRegSet) {
-    SmallVector<unsigned, 0> VRegsBuffer;
+    SmallVector<Register, 0> VRegsBuffer;
     filterAndAdd(FromRegSet, VRegsBuffer);
   }
   // Filter \p FromRegSet through the filter and append passed elements into \p
@@ -2233,13 +2240,13 @@ struct VRegFilter {
   // \returns true if anything changed.
   template <typename RegSetT>
   bool filterAndAdd(const RegSetT &FromRegSet,
-                    SmallVectorImpl<unsigned> &ToVRegs) {
+                    SmallVectorImpl<Register> &ToVRegs) {
     unsigned SparseUniverse = Sparse.size();
     unsigned NewSparseUniverse = SparseUniverse;
     unsigned NewDenseSize = Dense.size();
     size_t Begin = ToVRegs.size();
-    for (unsigned Reg : FromRegSet) {
-      if (!Register::isVirtualRegister(Reg))
+    for (Register Reg : FromRegSet) {
+      if (!Reg.isVirtual())
         continue;
       unsigned Index = Register::virtReg2Index(Reg);
       if (Index < SparseUniverseMax) {
@@ -2263,7 +2270,7 @@ struct VRegFilter {
     Sparse.resize(NewSparseUniverse);
     Dense.reserve(NewDenseSize);
     for (unsigned I = Begin; I < End; ++I) {
-      unsigned Reg = ToVRegs[I];
+      Register Reg = ToVRegs[I];
       unsigned Index = Register::virtReg2Index(Reg);
       if (Index < SparseUniverseMax)
         Sparse.set(Index);
@@ -2296,7 +2303,7 @@ struct VRegFilter {
 // universe). filter_b implicitly contains all physical registers at all times.
 class FilteringVRegSet {
   VRegFilter Filter;
-  SmallVector<unsigned, 0> VRegs;
+  SmallVector<Register, 0> VRegs;
 
 public:
   // Set-up the filter_b. \pre Input register set \p RS must have no duplicates.
@@ -2474,7 +2481,7 @@ void MachineVerifier::visitMachineFunctionAfter() {
   // Check for killed virtual registers that should be live out.
   for (const auto &MBB : *MF) {
     BBInfo &MInfo = MBBInfoMap[&MBB];
-    for (unsigned VReg : MInfo.vregsRequired)
+    for (Register VReg : MInfo.vregsRequired)
       if (MInfo.regsKilled.count(VReg)) {
         report("Virtual register killed in block, but needed live out.", &MBB);
         errs() << "Virtual register " << printReg(VReg)
@@ -2484,7 +2491,7 @@ void MachineVerifier::visitMachineFunctionAfter() {
 
   if (!MF->empty()) {
     BBInfo &MInfo = MBBInfoMap[&MF->front()];
-    for (unsigned VReg : MInfo.vregsRequired) {
+    for (Register VReg : MInfo.vregsRequired) {
       report("Virtual register defs don't dominate all uses.", MF);
       report_context_vreg(VReg);
     }
@@ -2543,8 +2550,8 @@ void MachineVerifier::visitMachineFunctionAfter() {
 
 void MachineVerifier::verifyLiveVariables() {
   assert(LiveVars && "Don't call verifyLiveVariables without LiveVars");
-  for (unsigned i = 0, e = MRI->getNumVirtRegs(); i != e; ++i) {
-    unsigned Reg = Register::index2VirtReg(i);
+  for (unsigned I = 0, E = MRI->getNumVirtRegs(); I != E; ++I) {
+    Register Reg = Register::index2VirtReg(I);
     LiveVariables::VarInfo &VI = LiveVars->getVarInfo(Reg);
     for (const auto &MBB : *MF) {
       BBInfo &MInfo = MBBInfoMap[&MBB];
@@ -2569,8 +2576,8 @@ void MachineVerifier::verifyLiveVariables() {
 
 void MachineVerifier::verifyLiveIntervals() {
   assert(LiveInts && "Don't call verifyLiveIntervals without LiveInts");
-  for (unsigned i = 0, e = MRI->getNumVirtRegs(); i != e; ++i) {
-    unsigned Reg = Register::index2VirtReg(i);
+  for (unsigned I = 0, E = MRI->getNumVirtRegs(); I != E; ++I) {
+    Register Reg = Register::index2VirtReg(I);
 
     // Spilling and splitting may leave unused registers around. Skip them.
     if (MRI->reg_nodbg_empty(Reg))
@@ -2594,7 +2601,7 @@ void MachineVerifier::verifyLiveIntervals() {
 }
 
 void MachineVerifier::verifyLiveRangeValue(const LiveRange &LR,
-                                           const VNInfo *VNI, unsigned Reg,
+                                           const VNInfo *VNI, Register Reg,
                                            LaneBitmask LaneMask) {
   if (VNI->isUnused())
     return;
@@ -2687,8 +2694,8 @@ void MachineVerifier::verifyLiveRangeValue(const LiveRange &LR,
 
 void MachineVerifier::verifyLiveRangeSegment(const LiveRange &LR,
                                              const LiveRange::const_iterator I,
-                                             unsigned Reg, LaneBitmask LaneMask)
-{
+                                             Register Reg,
+                                             LaneBitmask LaneMask) {
   const LiveRange::Segment &S = *I;
   const VNInfo *VNI = S.valno;
   assert(VNI && "Live segment has no valno");
@@ -2899,7 +2906,7 @@ void MachineVerifier::verifyLiveRangeSegment(const LiveRange &LR,
   }
 }
 
-void MachineVerifier::verifyLiveRange(const LiveRange &LR, unsigned Reg,
+void MachineVerifier::verifyLiveRange(const LiveRange &LR, Register Reg,
                                       LaneBitmask LaneMask) {
   for (const VNInfo *VNI : LR.valnos)
     verifyLiveRangeValue(LR, VNI, Reg, LaneMask);
@@ -2909,7 +2916,7 @@ void MachineVerifier::verifyLiveRange(const LiveRange &LR, unsigned Reg,
 }
 
 void MachineVerifier::verifyLiveInterval(const LiveInterval &LI) {
-  unsigned Reg = LI.reg();
+  Register Reg = LI.reg();
   assert(Register::isVirtualRegister(Reg));
   verifyLiveRange(LI, Reg);
 
diff --git a/llvm/lib/CodeGen/RDFRegisters.cpp b/llvm/lib/CodeGen/RDFRegisters.cpp
index 6c2c889ba64ad..8760ba118934f 100644
--- a/llvm/lib/CodeGen/RDFRegisters.cpp
+++ b/llvm/lib/CodeGen/RDFRegisters.cpp
@@ -84,10 +84,10 @@ PhysicalRegisterInfo::PhysicalRegisterInfo(const TargetRegisterInfo &tri,
   for (uint32_t M = 1, NM = RegMasks.size(); M <= NM; ++M) {
     BitVector PU(TRI.getNumRegUnits());
     const uint32_t *MB = RegMasks.get(M);
-    for (unsigned i = 1, e = TRI.getNumRegs(); i != e; ++i) {
-      if (!(MB[i/32] & (1u << (i%32))))
+    for (unsigned I = 1, E = TRI.getNumRegs(); I != E; ++I) {
+      if (!(MB[I / 32] & (1u << (I % 32))))
         continue;
-      for (MCRegUnitIterator U(i, &TRI); U.isValid(); ++U)
+      for (MCRegUnitIterator U(MCRegister::from(I), &TRI); U.isValid(); ++U)
         PU.set(*U);
     }
     MaskInfos[M].Units = PU.flip();
diff --git a/llvm/lib/CodeGen/ReachingDefAnalysis.cpp b/llvm/lib/CodeGen/ReachingDefAnalysis.cpp
index 9f6da2aed1b01..a15229942d638 100644
--- a/llvm/lib/CodeGen/ReachingDefAnalysis.cpp
+++ b/llvm/lib/CodeGen/ReachingDefAnalysis.cpp
@@ -29,7 +29,7 @@ static bool isValidRegUse(const MachineOperand &MO) {
   return isValidReg(MO) && MO.isUse();
 }
 
-static bool isValidRegUseOf(const MachineOperand &MO, int PhysReg) {
+static bool isValidRegUseOf(const MachineOperand &MO, MCRegister PhysReg) {
   return isValidRegUse(MO) && MO.getReg() == PhysReg;
 }
 
@@ -37,7 +37,7 @@ static bool isValidRegDef(const MachineOperand &MO) {
   return isValidReg(MO) && MO.isDef();
 }
 
-static bool isValidRegDefOf(const MachineOperand &MO, int PhysReg) {
+static bool isValidRegDefOf(const MachineOperand &MO, MCRegister PhysReg) {
   return isValidRegDef(MO) && MO.getReg() == PhysReg;
 }
 
@@ -121,7 +121,8 @@ void ReachingDefAnalysis::processDefs(MachineInstr *MI) {
   for (auto &MO : MI->operands()) {
     if (!isValidRegDef(MO))
       continue;
-    for (MCRegUnitIterator Unit(MO.getReg(), TRI); Unit.isValid(); ++Unit) {
+    for (MCRegUnitIterator Unit(MO.getReg().asMCReg(), TRI); Unit.isValid();
+         ++Unit) {
       // This instruction explicitly defines the current reg unit.
       LLVM_DEBUG(dbgs() << printReg(*Unit, TRI) << ":\t" << CurInstr
                         << '\t' << *MI);
@@ -252,7 +253,8 @@ void ReachingDefAnalysis::traverse() {
 #endif
 }
 
-int ReachingDefAnalysis::getReachingDef(MachineInstr *MI, int PhysReg) const {
+int ReachingDefAnalysis::getReachingDef(MachineInstr *MI,
+                                        MCRegister PhysReg) const {
   assert(InstIds.count(MI) && "Unexpected machine instuction.");
   int InstId = InstIds.lookup(MI);
   int DefRes = ReachingDefDefaultVal;
@@ -271,15 +273,16 @@ int ReachingDefAnalysis::getReachingDef(MachineInstr *MI, int PhysReg) const {
   return LatestDef;
 }
 
-MachineInstr* ReachingDefAnalysis::getReachingLocalMIDef(MachineInstr *MI,
-                                                         int PhysReg) const {
+MachineInstr *
+ReachingDefAnalysis::getReachingLocalMIDef(MachineInstr *MI,
+                                           MCRegister PhysReg) const {
   return hasLocalDefBefore(MI, PhysReg)
     ? getInstFromId(MI->getParent(), getReachingDef(MI, PhysReg))
     : nullptr;
 }
 
 bool ReachingDefAnalysis::hasSameReachingDef(MachineInstr *A, MachineInstr *B,
-                                             int PhysReg) const {
+                                             MCRegister PhysReg) const {
   MachineBasicBlock *ParentA = A->getParent();
   MachineBasicBlock *ParentB = B->getParent();
   if (ParentA != ParentB)
@@ -307,18 +310,19 @@ MachineInstr *ReachingDefAnalysis::getInstFromId(MachineBasicBlock *MBB,
   return nullptr;
 }
 
-int
-ReachingDefAnalysis::getClearance(MachineInstr *MI, MCPhysReg PhysReg) const {
+int ReachingDefAnalysis::getClearance(MachineInstr *MI,
+                                      MCRegister PhysReg) const {
   assert(InstIds.count(MI) && "Unexpected machine instuction.");
   return InstIds.lookup(MI) - getReachingDef(MI, PhysReg);
 }
 
-bool
-ReachingDefAnalysis::hasLocalDefBefore(MachineInstr *MI, int PhysReg) const {
+bool ReachingDefAnalysis::hasLocalDefBefore(MachineInstr *MI,
+                                            MCRegister PhysReg) const {
   return getReachingDef(MI, PhysReg) >= 0;
 }
 
-void ReachingDefAnalysis::getReachingLocalUses(MachineInstr *Def, int PhysReg,
+void ReachingDefAnalysis::getReachingLocalUses(MachineInstr *Def,
+                                               MCRegister PhysReg,
                                                InstSet &Uses) const {
   MachineBasicBlock *MBB = Def->getParent();
   MachineBasicBlock::iterator MI = MachineBasicBlock::iterator(Def);
@@ -342,9 +346,9 @@ void ReachingDefAnalysis::getReachingLocalUses(MachineInstr *Def, int PhysReg,
   }
 }
 
-bool
-ReachingDefAnalysis::getLiveInUses(MachineBasicBlock *MBB, int PhysReg,
-                                   InstSet &Uses) const {
+bool ReachingDefAnalysis::getLiveInUses(MachineBasicBlock *MBB,
+                                        MCRegister PhysReg,
+                                        InstSet &Uses) const {
   for (MachineInstr &MI :
        instructionsWithoutDebug(MBB->instr_begin(), MBB->instr_end())) {
     for (auto &MO : MI.operands()) {
@@ -361,9 +365,8 @@ ReachingDefAnalysis::getLiveInUses(MachineBasicBlock *MBB, int PhysReg,
   return isReachingDefLiveOut(&*Last, PhysReg);
 }
 
-void
-ReachingDefAnalysis::getGlobalUses(MachineInstr *MI, int PhysReg,
-                                   InstSet &Uses) const {
+void ReachingDefAnalysis::getGlobalUses(MachineInstr *MI, MCRegister PhysReg,
+                                        InstSet &Uses) const {
   MachineBasicBlock *MBB = MI->getParent();
 
   // Collect the uses that each def touches within the block.
@@ -391,9 +394,9 @@ ReachingDefAnalysis::getGlobalUses(MachineInstr *MI, int PhysReg,
   }
 }
 
-void
-ReachingDefAnalysis::getGlobalReachingDefs(MachineInstr *MI, int PhysReg,
-                                           InstSet &Defs) const {
+void ReachingDefAnalysis::getGlobalReachingDefs(MachineInstr *MI,
+                                                MCRegister PhysReg,
+                                                InstSet &Defs) const {
   if (auto *Def = getUniqueReachingMIDef(MI, PhysReg)) {
     Defs.insert(Def);
     return;
@@ -403,15 +406,15 @@ ReachingDefAnalysis::getGlobalReachingDefs(MachineInstr *MI, int PhysReg,
     getLiveOuts(MBB, PhysReg, Defs);
 }
 
-void ReachingDefAnalysis::getLiveOuts(MachineBasicBlock *MBB, int PhysReg,
-                                      InstSet &Defs) const {
+void ReachingDefAnalysis::getLiveOuts(MachineBasicBlock *MBB,
+                                      MCRegister PhysReg, InstSet &Defs) const {
   SmallPtrSet<MachineBasicBlock*, 2> VisitedBBs;
   getLiveOuts(MBB, PhysReg, Defs, VisitedBBs);
 }
 
-void
-ReachingDefAnalysis::getLiveOuts(MachineBasicBlock *MBB, int PhysReg,
-                                 InstSet &Defs, BlockSet &VisitedBBs) const {
+void ReachingDefAnalysis::getLiveOuts(MachineBasicBlock *MBB,
+                                      MCRegister PhysReg, InstSet &Defs,
+                                      BlockSet &VisitedBBs) const {
   if (VisitedBBs.count(MBB))
     return;
 
@@ -428,8 +431,9 @@ ReachingDefAnalysis::getLiveOuts(MachineBasicBlock *MBB, int PhysReg,
       getLiveOuts(Pred, PhysReg, Defs, VisitedBBs);
 }
 
-MachineInstr *ReachingDefAnalysis::getUniqueReachingMIDef(MachineInstr *MI,
-                                                          int PhysReg) const {
+MachineInstr *
+ReachingDefAnalysis::getUniqueReachingMIDef(MachineInstr *MI,
+                                            MCRegister PhysReg) const {
   // If there's a local def before MI, return it.
   MachineInstr *LocalDef = getReachingLocalMIDef(MI, PhysReg);
   if (LocalDef && InstIds.lookup(LocalDef) < InstIds.lookup(MI))
@@ -460,7 +464,8 @@ MachineInstr *ReachingDefAnalysis::getMIOperand(MachineInstr *MI,
   return getUniqueReachingMIDef(MI, MO.getReg());
 }
 
-bool ReachingDefAnalysis::isRegUsedAfter(MachineInstr *MI, int PhysReg) const {
+bool ReachingDefAnalysis::isRegUsedAfter(MachineInstr *MI,
+                                         MCRegister PhysReg) const {
   MachineBasicBlock *MBB = MI->getParent();
   LivePhysRegs LiveRegs(*TRI);
   LiveRegs.addLiveOuts(*MBB);
@@ -481,7 +486,7 @@ bool ReachingDefAnalysis::isRegUsedAfter(MachineInstr *MI, int PhysReg) const {
 }
 
 bool ReachingDefAnalysis::isRegDefinedAfter(MachineInstr *MI,
-                                            int PhysReg) const {
+                                            MCRegister PhysReg) const {
   MachineBasicBlock *MBB = MI->getParent();
   auto Last = MBB->getLastNonDebugInstr();
   if (Last != MBB->end() &&
@@ -494,8 +499,8 @@ bool ReachingDefAnalysis::isRegDefinedAfter(MachineInstr *MI,
   return false;
 }
 
-bool
-ReachingDefAnalysis::isReachingDefLiveOut(MachineInstr *MI, int PhysReg) const {
+bool ReachingDefAnalysis::isReachingDefLiveOut(MachineInstr *MI,
+                                               MCRegister PhysReg) const {
   MachineBasicBlock *MBB = MI->getParent();
   LivePhysRegs LiveRegs(*TRI);
   LiveRegs.addLiveOuts(*MBB);
@@ -515,8 +520,9 @@ ReachingDefAnalysis::isReachingDefLiveOut(MachineInstr *MI, int PhysReg) const {
   return true;
 }
 
-MachineInstr* ReachingDefAnalysis::getLocalLiveOutMIDef(MachineBasicBlock *MBB,
-                                                        int PhysReg) const {
+MachineInstr *
+ReachingDefAnalysis::getLocalLiveOutMIDef(MachineBasicBlock *MBB,
+                                          MCRegister PhysReg) const {
   LivePhysRegs LiveRegs(*TRI);
   LiveRegs.addLiveOuts(*MBB);
   if (!LiveRegs.contains(PhysReg))
@@ -640,7 +646,7 @@ ReachingDefAnalysis::isSafeToRemove(MachineInstr *MI, InstSet &Visited,
 void ReachingDefAnalysis::collectKilledOperands(MachineInstr *MI,
                                                 InstSet &Dead) const {
   Dead.insert(MI);
-  auto IsDead = [this, &Dead](MachineInstr *Def, int PhysReg) {
+  auto IsDead = [this, &Dead](MachineInstr *Def, MCRegister PhysReg) {
     if (mayHaveSideEffects(*Def))
       return false;
 
@@ -673,12 +679,12 @@ void ReachingDefAnalysis::collectKilledOperands(MachineInstr *MI,
 }
 
 bool ReachingDefAnalysis::isSafeToDefRegAt(MachineInstr *MI,
-                                           int PhysReg) const {
+                                           MCRegister PhysReg) const {
   SmallPtrSet<MachineInstr*, 1> Ignore;
   return isSafeToDefRegAt(MI, PhysReg, Ignore);
 }
 
-bool ReachingDefAnalysis::isSafeToDefRegAt(MachineInstr *MI, int PhysReg,
+bool ReachingDefAnalysis::isSafeToDefRegAt(MachineInstr *MI, MCRegister PhysReg,
                                            InstSet &Ignore) const {
   // Check for any uses of the register after MI.
   if (isRegUsedAfter(MI, PhysReg)) {
diff --git a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
index c47c8d744a01d..5899da777fe95 100644
--- a/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
+++ b/llvm/lib/CodeGen/ScheduleDAGInstrs.cpp
@@ -199,7 +199,10 @@ void ScheduleDAGInstrs::exitRegion() {
 }
 
 void ScheduleDAGInstrs::addSchedBarrierDeps() {
-  MachineInstr *ExitMI = RegionEnd != BB->end() ? &*RegionEnd : nullptr;
+  MachineInstr *ExitMI =
+      RegionEnd != BB->end()
+          ? &*skipDebugInstructionsBackward(RegionEnd, RegionBegin)
+          : nullptr;
   ExitSU.setInstr(ExitMI);
   // Add dependencies on the defs and uses of the instruction.
   if (ExitMI) {
@@ -511,6 +514,8 @@ void ScheduleDAGInstrs::addVRegDefDeps(SUnit *SU, unsigned OperIdx) {
 /// TODO: Handle ExitSU "uses" properly.
 void ScheduleDAGInstrs::addVRegUseDeps(SUnit *SU, unsigned OperIdx) {
   const MachineInstr *MI = SU->getInstr();
+  assert(!MI->isDebugInstr());
+
   const MachineOperand &MO = MI->getOperand(OperIdx);
   Register Reg = MO.getReg();
 
@@ -802,7 +807,7 @@ void ScheduleDAGInstrs::buildSchedGraph(AAResults *AA,
       DbgMI = nullptr;
     }
 
-    if (MI.isDebugValue()) {
+    if (MI.isDebugValue() || MI.isDebugRef()) {
       DbgMI = &MI;
       continue;
     }
diff --git a/llvm/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp b/llvm/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
index fd821ed5c9692..669bca966a7d8 100644
--- a/llvm/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/FunctionLoweringInfo.cpp
@@ -197,7 +197,7 @@ void FunctionLoweringInfo::set(const Function &fn, MachineFunction &mf,
       // Look for inline asm that clobbers the SP register.
       if (auto *Call = dyn_cast<CallBase>(&I)) {
         if (Call->isInlineAsm()) {
-          unsigned SP = TLI->getStackPointerRegisterToSaveRestore();
+          Register SP = TLI->getStackPointerRegisterToSaveRestore();
           const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
           std::vector<TargetLowering::AsmOperandInfo> Ops =
               TLI->ParseConstraints(Fn->getParent()->getDataLayout(), TRI,
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index d2487c9c6c107..6328626868d03 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -1616,7 +1616,7 @@ SDValue SelectionDAGLegalize::ExpandFABS(SDNode *Node) const {
 
 void SelectionDAGLegalize::ExpandDYNAMIC_STACKALLOC(SDNode* Node,
                                            SmallVectorImpl<SDValue> &Results) {
-  unsigned SPReg = TLI.getStackPointerRegisterToSaveRestore();
+  Register SPReg = TLI.getStackPointerRegisterToSaveRestore();
   assert(SPReg && "Target cannot require DYNAMIC_STACKALLOC expansion and"
           " not tell us which reg is the stack pointer!");
   SDLoc dl(Node);
@@ -2460,12 +2460,19 @@ SDValue SelectionDAGLegalize::ExpandLegalINT_TO_FP(SDNode *Node,
   assert(!isSigned && "Legalize cannot Expand SINT_TO_FP for i64 yet");
 
   // TODO: Generalize this for use with other types.
-  if ((SrcVT == MVT::i32 || SrcVT == MVT::i64) && DestVT == MVT::f32) {
-    LLVM_DEBUG(dbgs() << "Converting unsigned i32/i64 to f32\n");
+  if (((SrcVT == MVT::i32 || SrcVT == MVT::i64) && DestVT == MVT::f32) ||
+      (SrcVT == MVT::i64 && DestVT == MVT::f64)) {
+    LLVM_DEBUG(dbgs() << "Converting unsigned i32/i64 to f32/f64\n");
     // For unsigned conversions, convert them to signed conversions using the
     // algorithm from the x86_64 __floatundisf in compiler_rt. That method
     // should be valid for i32->f32 as well.
 
+    // More generally this transform should be valid if there are 3 more bits
+    // in the integer type than the significand. Rounding uses the first bit
+    // after the width of the significand and the OR of all bits after that. So
+    // we need to be able to OR the shifted out bit into one of the bits that
+    // participate in the OR.
+
     // TODO: This really should be implemented using a branch rather than a
     // select.  We happen to get lucky and machinesink does the right
     // thing most of the time.  This would be a good candidate for a
@@ -3245,7 +3252,7 @@ bool SelectionDAGLegalize::ExpandNode(SDNode *Node) {
   case ISD::STACKSAVE:
     // Expand to CopyFromReg if the target set
     // StackPointerRegisterToSaveRestore.
-    if (unsigned SP = TLI.getStackPointerRegisterToSaveRestore()) {
+    if (Register SP = TLI.getStackPointerRegisterToSaveRestore()) {
       Results.push_back(DAG.getCopyFromReg(Node->getOperand(0), dl, SP,
                                            Node->getValueType(0)));
       Results.push_back(Results[0].getValue(1));
@@ -3257,7 +3264,7 @@ bool SelectionDAGLegalize::ExpandNode(SDNode *Node) {
   case ISD::STACKRESTORE:
     // Expand to CopyToReg if the target set
     // StackPointerRegisterToSaveRestore.
-    if (unsigned SP = TLI.getStackPointerRegisterToSaveRestore()) {
+    if (Register SP = TLI.getStackPointerRegisterToSaveRestore()) {
       Results.push_back(DAG.getCopyToReg(Node->getOperand(0), dl, SP,
                                          Node->getOperand(1)));
     } else {
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 5fe26a6d2abdf..24c32d367283a 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -2045,7 +2045,14 @@ SDValue SelectionDAG::CreateStackTemporary(EVT VT, unsigned minAlign) {
 }
 
 SDValue SelectionDAG::CreateStackTemporary(EVT VT1, EVT VT2) {
-  TypeSize Bytes = std::max(VT1.getStoreSize(), VT2.getStoreSize());
+  TypeSize VT1Size = VT1.getStoreSize();
+  TypeSize VT2Size = VT2.getStoreSize();
+  assert(VT1Size.isScalable() == VT2Size.isScalable() &&
+         "Don't know how to choose the maximum size when creating a stack "
+         "temporary");
+  TypeSize Bytes =
+      VT1Size.getKnownMinSize() > VT2Size.getKnownMinSize() ? VT1Size : VT2Size;
+
   Type *Ty1 = VT1.getTypeForEVT(*getContext());
   Type *Ty2 = VT2.getTypeForEVT(*getContext());
   const DataLayout &DL = getDataLayout();
@@ -5581,8 +5588,8 @@ SDValue SelectionDAG::getNode(unsigned Opcode, const SDLoc &DL, EVT VT,
 
     // EXTRACT_SUBVECTOR of CONCAT_VECTOR can be simplified if the pieces of
     // the concat have the same type as the extract.
-    if (N2C && N1.getOpcode() == ISD::CONCAT_VECTORS &&
-        N1.getNumOperands() > 0 && VT == N1.getOperand(0).getValueType()) {
+    if (N1.getOpcode() == ISD::CONCAT_VECTORS && N1.getNumOperands() > 0 &&
+        VT == N1.getOperand(0).getValueType()) {
       unsigned Factor = VT.getVectorMinNumElements();
       return N1.getOperand(N2C->getZExtValue() / Factor);
     }
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index e1d3ae7ba7e39..a89cbc7e5fa77 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -957,7 +957,7 @@ void RegsForValue::AddInlineAsmOperands(unsigned Code, bool HasMatching,
     // shouldn't try to apply any sort of splitting logic to them.
     assert(Regs.size() == RegVTs.size() && Regs.size() == ValueVTs.size() &&
            "No 1:1 mapping from clobbers to regs?");
-    unsigned SP = TLI.getStackPointerRegisterToSaveRestore();
+    Register SP = TLI.getStackPointerRegisterToSaveRestore();
     (void)SP;
     for (unsigned I = 0, E = ValueVTs.size(); I != E; ++I) {
       Ops.push_back(DAG.getRegister(Regs[I], RegVTs[I]));
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
index 3e6c11a0c1299..58a731afee36b 100644
--- a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -3981,7 +3981,8 @@ SDValue TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
     EVT ShValTy = N0.getValueType();
 
     // Fold bit comparisons when we can.
-    if ((Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
+    if (getBooleanContents(N0.getValueType()) == ZeroOrOneBooleanContent &&
+        (Cond == ISD::SETEQ || Cond == ISD::SETNE) &&
         (VT == ShValTy || (isTypeLegal(VT) && VT.bitsLE(ShValTy))) &&
         N0.getOpcode() == ISD::AND) {
       if (auto *AndRHS = dyn_cast<ConstantSDNode>(N0.getOperand(1))) {
@@ -6527,8 +6528,13 @@ bool TargetLowering::expandFP_TO_UINT(SDNode *Node, SDValue &Result,
 bool TargetLowering::expandUINT_TO_FP(SDNode *Node, SDValue &Result,
                                       SDValue &Chain,
                                       SelectionDAG &DAG) const {
-  unsigned OpNo = Node->isStrictFPOpcode() ? 1 : 0;
-  SDValue Src = Node->getOperand(OpNo);
+  // This transform is not correct for converting 0 when rounding mode is set
+  // to round toward negative infinity which will produce -0.0. So disable under
+  // strictfp.
+  if (Node->isStrictFPOpcode())
+    return false;
+
+  SDValue Src = Node->getOperand(0);
   EVT SrcVT = Src.getValueType();
   EVT DstVT = Node->getValueType(0);
 
@@ -6547,9 +6553,10 @@ bool TargetLowering::expandUINT_TO_FP(SDNode *Node, SDValue &Result,
   EVT ShiftVT = getShiftAmountTy(SrcVT, DAG.getDataLayout());
 
   // Implementation of unsigned i64 to f64 following the algorithm in
-  // __floatundidf in compiler_rt. This implementation has the advantage
-  // of performing rounding correctly, both in the default rounding mode
-  // and in all alternate rounding modes.
+  // __floatundidf in compiler_rt.  This implementation performs rounding
+  // correctly in all rounding modes with the exception of converting 0
+  // when rounding toward negative infinity. In that case the fsub will produce
+  // -0.0. This will be added to +0.0 and produce -0.0 which is incorrect.
   SDValue TwoP52 = DAG.getConstant(UINT64_C(0x4330000000000000), dl, SrcVT);
   SDValue TwoP84PlusTwoP52 = DAG.getConstantFP(
       BitsToDouble(UINT64_C(0x4530000000100000)), dl, DstVT);
@@ -6563,18 +6570,9 @@ bool TargetLowering::expandUINT_TO_FP(SDNode *Node, SDValue &Result,
   SDValue HiOr = DAG.getNode(ISD::OR, dl, SrcVT, Hi, TwoP84);
   SDValue LoFlt = DAG.getBitcast(DstVT, LoOr);
   SDValue HiFlt = DAG.getBitcast(DstVT, HiOr);
-  if (Node->isStrictFPOpcode()) {
-    SDValue HiSub =
-        DAG.getNode(ISD::STRICT_FSUB, dl, {DstVT, MVT::Other},
-                    {Node->getOperand(0), HiFlt, TwoP84PlusTwoP52});
-    Result = DAG.getNode(ISD::STRICT_FADD, dl, {DstVT, MVT::Other},
-                         {HiSub.getValue(1), LoFlt, HiSub});
-    Chain = Result.getValue(1);
-  } else {
-    SDValue HiSub =
-        DAG.getNode(ISD::FSUB, dl, DstVT, HiFlt, TwoP84PlusTwoP52);
-    Result = DAG.getNode(ISD::FADD, dl, DstVT, LoFlt, HiSub);
-  }
+  SDValue HiSub =
+      DAG.getNode(ISD::FSUB, dl, DstVT, HiFlt, TwoP84PlusTwoP52);
+  Result = DAG.getNode(ISD::FADD, dl, DstVT, LoFlt, HiSub);
   return true;
 }
 
diff --git a/llvm/lib/CodeGen/ShrinkWrap.cpp b/llvm/lib/CodeGen/ShrinkWrap.cpp
index ce43fb1fbd4b0..f89069e9f7285 100644
--- a/llvm/lib/CodeGen/ShrinkWrap.cpp
+++ b/llvm/lib/CodeGen/ShrinkWrap.cpp
@@ -144,7 +144,7 @@ class ShrinkWrap : public MachineFunctionPass {
   unsigned FrameDestroyOpcode;
 
   /// Stack pointer register, used by llvm.{savestack,restorestack}
-  unsigned SP;
+  Register SP;
 
   /// Entry block.
   const MachineBasicBlock *Entry;
@@ -331,11 +331,7 @@ void ShrinkWrap::updateSaveRestorePoints(MachineBasicBlock &MBB,
     Save = &MBB;
   else
     Save = MDT->findNearestCommonDominator(Save, &MBB);
-
-  if (!Save) {
-    LLVM_DEBUG(dbgs() << "Found a block that is not reachable from Entry\n");
-    return;
-  }
+  assert(Save);
 
   if (!Restore)
     Restore = &MBB;
@@ -381,7 +377,7 @@ void ShrinkWrap::updateSaveRestorePoints(MachineBasicBlock &MBB,
   // C. Save and Restore are in the same loop.
   bool SaveDominatesRestore = false;
   bool RestorePostDominatesSave = false;
-  while (Save && Restore &&
+  while (Restore &&
          (!(SaveDominatesRestore = MDT->dominates(Save, Restore)) ||
           !(RestorePostDominatesSave = MPDT->dominates(Restore, Save)) ||
           // Post-dominance is not enough in loops to ensure that all uses/defs
@@ -412,8 +408,7 @@ void ShrinkWrap::updateSaveRestorePoints(MachineBasicBlock &MBB,
       Restore = MPDT->findNearestCommonDominator(Restore, Save);
 
     // Fix (C).
-    if (Save && Restore &&
-        (MLI->getLoopFor(Save) || MLI->getLoopFor(Restore))) {
+    if (Restore && (MLI->getLoopFor(Save) || MLI->getLoopFor(Restore))) {
       if (MLI->getLoopDepth(Save) > MLI->getLoopDepth(Restore)) {
         // Push Save outside of this loop if immediate dominator is different
         // from save block. If immediate dominator is not different, bail out.
diff --git a/llvm/lib/CodeGen/StackProtector.cpp b/llvm/lib/CodeGen/StackProtector.cpp
index 3d961af8ec3e3..8074793bc6273 100644
--- a/llvm/lib/CodeGen/StackProtector.cpp
+++ b/llvm/lib/CodeGen/StackProtector.cpp
@@ -381,7 +381,10 @@ bool StackProtector::RequiresStackProtector() {
 static Value *getStackGuard(const TargetLoweringBase *TLI, Module *M,
                             IRBuilder<> &B,
                             bool *SupportsSelectionDAGSP = nullptr) {
-  if (Value *Guard = TLI->getIRStackGuard(B))
+  Value *Guard = TLI->getIRStackGuard(B);
+  auto GuardMode = TLI->getTargetMachine().Options.StackProtectorGuard;
+  if ((GuardMode == llvm::StackProtectorGuards::TLS ||
+       GuardMode == llvm::StackProtectorGuards::None) && Guard)
     return B.CreateLoad(B.getInt8PtrTy(), Guard, true, "StackGuard");
 
   // Use SelectionDAG SSP handling, since there isn't an IR guard.
diff --git a/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp b/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp
index 4a727902bdd26..17ad51a5d6ede 100644
--- a/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp
+++ b/llvm/lib/CodeGen/TwoAddressInstructionPass.cpp
@@ -606,6 +606,24 @@ TwoAddressInstructionPass::convertInstTo3Addr(MachineBasicBlock::iterator &mi,
   if (LIS)
     LIS->ReplaceMachineInstrInMaps(*mi, *NewMI);
 
+  // If the old instruction is debug value tracked, an update is required.
+  if (auto OldInstrNum = mi->peekDebugInstrNum()) {
+    // Sanity check.
+    assert(mi->getNumExplicitDefs() == 1);
+    assert(NewMI->getNumExplicitDefs() == 1);
+
+    // Find the old and new def location.
+    auto OldIt = mi->defs().begin();
+    auto NewIt = NewMI->defs().begin();
+    unsigned OldIdx = mi->getOperandNo(OldIt);
+    unsigned NewIdx = NewMI->getOperandNo(NewIt);
+
+    // Record that one def has been replaced by the other.
+    unsigned NewInstrNum = NewMI->getDebugInstrNum();
+    MF->makeDebugValueSubstitution(std::make_pair(OldInstrNum, OldIdx),
+                                   std::make_pair(NewInstrNum, NewIdx));
+  }
+
   MBB->erase(mi); // Nuke the old inst.
 
   DistanceMap.insert(std::make_pair(NewMI, Dist));
diff --git a/llvm/lib/DebugInfo/Symbolize/Symbolize.cpp b/llvm/lib/DebugInfo/Symbolize/Symbolize.cpp
index ed53ea9712b6d..383e2c0d3054c 100644
--- a/llvm/lib/DebugInfo/Symbolize/Symbolize.cpp
+++ b/llvm/lib/DebugInfo/Symbolize/Symbolize.cpp
@@ -68,8 +68,7 @@ LLVMSymbolizer::symbolizeCode(const ObjectFile &Obj,
   if (I != Modules.end())
     return symbolizeCodeCommon(I->second.get(), ModuleOffset);
 
-  std::unique_ptr<DIContext> Context =
-      DWARFContext::create(Obj, nullptr, "", Opts.RecoverableErrorHandler);
+  std::unique_ptr<DIContext> Context = DWARFContext::create(Obj);
   Expected<SymbolizableModule *> InfoOrErr =
                      createModuleInfo(&Obj, std::move(Context), ModuleName);
   if (!InfoOrErr)
@@ -573,8 +572,7 @@ LLVMSymbolizer::getOrCreateModuleInfo(const std::string &ModuleName) {
     }
   }
   if (!Context)
-    Context = DWARFContext::create(*Objects.second, nullptr, Opts.DWPName,
-                                   Opts.RecoverableErrorHandler);
+    Context = DWARFContext::create(*Objects.second, nullptr, Opts.DWPName);
   return createModuleInfo(Objects.first, std::move(Context), ModuleName);
 }
 
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index f5b7e4ebd4396..dafa45c367862 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -16,6 +16,7 @@
 
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/StringSwitch.h"
+#include "llvm/ADT/Triple.h"
 #include "llvm/IR/CFG.h"
 #include "llvm/IR/DebugInfo.h"
 #include "llvm/IR/IRBuilder.h"
@@ -217,6 +218,14 @@ Value *OpenMPIRBuilder::getOrCreateIdent(Constant *SrcLocStr,
   return Ident;
 }
 
+Type *OpenMPIRBuilder::getLanemaskType() {
+  LLVMContext &Ctx = M.getContext();
+  Triple triple(M.getTargetTriple());
+
+  // This test is adequate until deviceRTL has finer grained lane widths
+  return triple.isAMDGCN() ? Type::getInt64Ty(Ctx) : Type::getInt32Ty(Ctx);
+}
+
 Constant *OpenMPIRBuilder::getOrCreateSrcLocStr(StringRef LocStr) {
   Constant *&SrcLocStr = SrcLocStrMap[LocStr];
   if (!SrcLocStr) {
diff --git a/llvm/lib/IR/IRBuilder.cpp b/llvm/lib/IR/IRBuilder.cpp
index 982a158ff5c04..c0e4451f52003 100644
--- a/llvm/lib/IR/IRBuilder.cpp
+++ b/llvm/lib/IR/IRBuilder.cpp
@@ -653,10 +653,10 @@ CallInst *IRBuilderBase::CreateGCStatepointCall(
 
 CallInst *IRBuilderBase::CreateGCStatepointCall(
     uint64_t ID, uint32_t NumPatchBytes, Value *ActualCallee, uint32_t Flags,
-    ArrayRef<Use> CallArgs, Optional<ArrayRef<Use>> TransitionArgs,
+    ArrayRef<Value *> CallArgs, Optional<ArrayRef<Use>> TransitionArgs,
     Optional<ArrayRef<Use>> DeoptArgs, ArrayRef<Value *> GCArgs,
     const Twine &Name) {
-  return CreateGCStatepointCallCommon<Use, Use, Use, Value *>(
+  return CreateGCStatepointCallCommon<Value *, Use, Use, Value *>(
       this, ID, NumPatchBytes, ActualCallee, Flags, CallArgs, TransitionArgs,
       DeoptArgs, GCArgs, Name);
 }
@@ -711,9 +711,9 @@ InvokeInst *IRBuilderBase::CreateGCStatepointInvoke(
 InvokeInst *IRBuilderBase::CreateGCStatepointInvoke(
     uint64_t ID, uint32_t NumPatchBytes, Value *ActualInvokee,
     BasicBlock *NormalDest, BasicBlock *UnwindDest, uint32_t Flags,
-    ArrayRef<Use> InvokeArgs, Optional<ArrayRef<Use>> TransitionArgs,
+    ArrayRef<Value *> InvokeArgs, Optional<ArrayRef<Use>> TransitionArgs,
     Optional<ArrayRef<Use>> DeoptArgs, ArrayRef<Value *> GCArgs, const Twine &Name) {
-  return CreateGCStatepointInvokeCommon<Use, Use, Use, Value *>(
+  return CreateGCStatepointInvokeCommon<Value *, Use, Use, Value *>(
       this, ID, NumPatchBytes, ActualInvokee, NormalDest, UnwindDest, Flags,
       InvokeArgs, TransitionArgs, DeoptArgs, GCArgs, Name);
 }
diff --git a/llvm/lib/Passes/PassBuilder.cpp b/llvm/lib/Passes/PassBuilder.cpp
index 6db7f3c88dd97..ab4b305f250a4 100644
--- a/llvm/lib/Passes/PassBuilder.cpp
+++ b/llvm/lib/Passes/PassBuilder.cpp
@@ -84,6 +84,7 @@
 #include "llvm/Transforms/IPO/AlwaysInliner.h"
 #include "llvm/Transforms/IPO/ArgumentPromotion.h"
 #include "llvm/Transforms/IPO/Attributor.h"
+#include "llvm/Transforms/IPO/BlockExtractor.h"
 #include "llvm/Transforms/IPO/CalledValuePropagation.h"
 #include "llvm/Transforms/IPO/ConstantMerge.h"
 #include "llvm/Transforms/IPO/CrossDSOCFI.h"
@@ -676,7 +677,7 @@ PassBuilder::buildFunctionSimplificationPipeline(OptimizationLevel Level,
   LPM1.addPass(LoopInstSimplifyPass());
   LPM1.addPass(LoopSimplifyCFGPass());
 
-  // Rotate Loop - disable header duplication at -Oz
+  // Disable header duplication in loop rotation at -Oz.
   LPM1.addPass(LoopRotatePass(Level != OptimizationLevel::Oz));
   // TODO: Investigate promotion cap for O1.
   LPM1.addPass(LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap));
@@ -750,6 +751,12 @@ PassBuilder::buildFunctionSimplificationPipeline(OptimizationLevel Level,
   // redo DCE, etc.
   FPM.addPass(JumpThreadingPass());
   FPM.addPass(CorrelatedValuePropagationPass());
+
+  // Finally, do an expensive DCE pass to catch all the dead code exposed by
+  // the simplifications and basic cleanup after all the simplifications.
+  // TODO: Investigate if this is too expensive.
+  FPM.addPass(ADCEPass());
+
   FPM.addPass(DSEPass());
   FPM.addPass(createFunctionToLoopPassAdaptor(
       LICMPass(PTO.LicmMssaOptCap, PTO.LicmMssaNoAccForPromotionCap),
@@ -761,10 +768,6 @@ PassBuilder::buildFunctionSimplificationPipeline(OptimizationLevel Level,
   for (auto &C : ScalarOptimizerLateEPCallbacks)
     C(FPM, Level);
 
-  // Finally, do an expensive DCE pass to catch all the dead code exposed by
-  // the simplifications and basic cleanup after all the simplifications.
-  // TODO: Investigate if this is too expensive.
-  FPM.addPass(ADCEPass());
   FPM.addPass(SimplifyCFGPass());
   FPM.addPass(InstCombinePass());
   invokePeepholeEPCallbacks(FPM, Level);
@@ -830,8 +833,9 @@ void PassBuilder::addPGOInstrPasses(ModulePassManager &MPM, bool DebugLogging,
   MPM.addPass(PGOInstrumentationGen(IsCS));
 
   FunctionPassManager FPM;
+  // Disable header duplication in loop rotation at -Oz.
   FPM.addPass(createFunctionToLoopPassAdaptor(
-      LoopRotatePass(), EnableMSSALoopDependency,
+      LoopRotatePass(Level != OptimizationLevel::Oz), EnableMSSALoopDependency,
       /*UseBlockFrequencyInfo=*/false, DebugLogging));
   MPM.addPass(createModuleToFunctionPassAdaptor(std::move(FPM)));
 
@@ -1161,8 +1165,9 @@ ModulePassManager PassBuilder::buildModuleOptimizationPipeline(
     C(OptimizePM, Level);
 
   // First rotate loops that may have been un-rotated by prior passes.
+  // Disable header duplication at -Oz.
   OptimizePM.addPass(createFunctionToLoopPassAdaptor(
-      LoopRotatePass(), EnableMSSALoopDependency,
+      LoopRotatePass(Level != OptimizationLevel::Oz), EnableMSSALoopDependency,
       /*UseBlockFrequencyInfo=*/false, DebugLogging));
 
   // Distribute loops to allow partial vectorization.  I.e. isolate dependences
diff --git a/llvm/lib/Passes/PassRegistry.def b/llvm/lib/Passes/PassRegistry.def
index 5a0b37738c09a..0cac7c4607522 100644
--- a/llvm/lib/Passes/PassRegistry.def
+++ b/llvm/lib/Passes/PassRegistry.def
@@ -50,6 +50,7 @@ MODULE_PASS("constmerge", ConstantMergePass())
 MODULE_PASS("cross-dso-cfi", CrossDSOCFIPass())
 MODULE_PASS("deadargelim", DeadArgumentEliminationPass())
 MODULE_PASS("elim-avail-extern", EliminateAvailableExternallyPass())
+MODULE_PASS("extract-blocks", BlockExtractorPass())
 MODULE_PASS("forceattrs", ForceFunctionAttrsPass())
 MODULE_PASS("function-import", FunctionImportPass())
 MODULE_PASS("globaldce", GlobalDCEPass())
diff --git a/llvm/lib/Support/Host.cpp b/llvm/lib/Support/Host.cpp
index ccd19b384c908..e30ccfc2867f8 100644
--- a/llvm/lib/Support/Host.cpp
+++ b/llvm/lib/Support/Host.cpp
@@ -1475,6 +1475,7 @@ bool sys::getHostCPUFeatures(StringMap<bool> &Features) {
   Features["movdir64b"]       = HasLeaf7 && ((ECX >> 28) & 1);
   Features["enqcmd"]          = HasLeaf7 && ((ECX >> 29) & 1);
 
+  Features["uintr"]           = HasLeaf7 && ((EDX >> 5) & 1);
   Features["avx512vp2intersect"] =
       HasLeaf7 && ((EDX >> 8) & 1) && HasAVX512Save;
   Features["serialize"]       = HasLeaf7 && ((EDX >> 14) & 1);
diff --git a/llvm/lib/Support/Path.cpp b/llvm/lib/Support/Path.cpp
index bbc02a246a50a..ef223ae5ac1d6 100644
--- a/llvm/lib/Support/Path.cpp
+++ b/llvm/lib/Support/Path.cpp
@@ -354,10 +354,9 @@ StringRef root_path(StringRef path, Style style) {
       if ((++pos != e) && is_separator((*pos)[0], style)) {
         // {C:/,//net/}, so get the first two components.
         return path.substr(0, b->size() + pos->size());
-      } else {
-        // just {C:,//net}, return the first component.
-        return *b;
       }
+      // just {C:,//net}, return the first component.
+      return *b;
     }
 
     // POSIX style root directory.
@@ -467,8 +466,7 @@ StringRef parent_path(StringRef path, Style style) {
   size_t end_pos = parent_path_end(path, style);
   if (end_pos == StringRef::npos)
     return StringRef();
-  else
-    return path.substr(0, end_pos);
+  return path.substr(0, end_pos);
 }
 
 void remove_filename(SmallVectorImpl<char> &path, Style style) {
@@ -581,12 +579,10 @@ StringRef stem(StringRef path, Style style) {
   size_t pos = fname.find_last_of('.');
   if (pos == StringRef::npos)
     return fname;
-  else
-    if ((fname.size() == 1 && fname == ".") ||
-        (fname.size() == 2 && fname == ".."))
-      return fname;
-    else
-      return fname.substr(0, pos);
+  if ((fname.size() == 1 && fname == ".") ||
+      (fname.size() == 2 && fname == ".."))
+    return fname;
+  return fname.substr(0, pos);
 }
 
 StringRef extension(StringRef path, Style style) {
@@ -594,12 +590,10 @@ StringRef extension(StringRef path, Style style) {
   size_t pos = fname.find_last_of('.');
   if (pos == StringRef::npos)
     return StringRef();
-  else
-    if ((fname.size() == 1 && fname == ".") ||
-        (fname.size() == 2 && fname == ".."))
-      return StringRef();
-    else
-      return fname.substr(pos);
+  if ((fname.size() == 1 && fname == ".") ||
+      (fname.size() == 2 && fname == ".."))
+    return StringRef();
+  return fname.substr(pos);
 }
 
 bool is_separator(char value, Style style) {
@@ -1299,7 +1293,7 @@ Expected<TempFile> TempFile::create(const Twine &Model, unsigned Mode) {
 #endif
   return std::move(Ret);
 }
-}
+} // namespace fs
 
-} // end namsspace sys
-} // end namespace llvm
+} // namespace sys
+} // namespace llvm
diff --git a/llvm/lib/Support/X86TargetParser.cpp b/llvm/lib/Support/X86TargetParser.cpp
index 698dc4a5e9d61..7b1e0c0bc6859 100644
--- a/llvm/lib/Support/X86TargetParser.cpp
+++ b/llvm/lib/Support/X86TargetParser.cpp
@@ -204,7 +204,8 @@ constexpr FeatureBitset FeaturesSapphireRapids =
     FeaturesICLServer | FeatureAMX_TILE | FeatureAMX_INT8 | FeatureAMX_BF16 |
     FeatureAVX512BF16 | FeatureAVX512VP2INTERSECT | FeatureCLDEMOTE |
     FeatureENQCMD | FeatureMOVDIR64B | FeatureMOVDIRI | FeaturePTWRITE |
-    FeatureSERIALIZE | FeatureSHSTK | FeatureTSXLDTRK | FeatureWAITPKG;
+    FeatureSERIALIZE | FeatureSHSTK | FeatureTSXLDTRK | FeatureUINTR |
+    FeatureWAITPKG;
 
 // Intel Atom processors.
 // Bonnell has feature parity with Core2 and adds MOVBE.
@@ -481,6 +482,7 @@ constexpr FeatureBitset ImpliedFeaturesSGX = {};
 constexpr FeatureBitset ImpliedFeaturesSHSTK = {};
 constexpr FeatureBitset ImpliedFeaturesTBM = {};
 constexpr FeatureBitset ImpliedFeaturesTSXLDTRK = {};
+constexpr FeatureBitset ImpliedFeaturesUINTR = {};
 constexpr FeatureBitset ImpliedFeaturesWAITPKG = {};
 constexpr FeatureBitset ImpliedFeaturesWBNOINVD = {};
 constexpr FeatureBitset ImpliedFeaturesVZEROUPPER = {};
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
index 6238fb9664c65..c76d292acd6a8 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp
@@ -993,7 +993,9 @@ void AMDGPUAsmPrinter::getSIProgramInfo(SIProgramInfo &ProgInfo,
   ProgInfo.FlatUsed = Info.UsesFlatScratch;
   ProgInfo.DynamicCallStack = Info.HasDynamicallySizedStack || Info.HasRecursion;
 
-  if (!isUInt<32>(ProgInfo.ScratchSize)) {
+  const uint64_t MaxScratchPerWorkitem =
+      GCNSubtarget::MaxWaveScratchSize / STM.getWavefrontSize();
+  if (ProgInfo.ScratchSize > MaxScratchPerWorkitem) {
     DiagnosticInfoStackSize DiagStackSize(MF.getFunction(),
                                           ProgInfo.ScratchSize, DS_Error);
     MF.getFunction().getContext().diagnose(DiagStackSize);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td b/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
index 7c83b6dcb44b9..5cbd65b95dff8 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
+++ b/llvm/lib/Target/AMDGPU/AMDGPUCallingConv.td
@@ -125,6 +125,8 @@ def CSR_AMDGPU_HighRegs : CalleeSavedRegs<
   (add CSR_AMDGPU_VGPRs, CSR_AMDGPU_SGPRs_32_105)
 >;
 
+def CSR_AMDGPU_NoRegs : CalleeSavedRegs<(add)>;
+
 // Calling convention for leaf functions
 def CC_AMDGPU_Func : CallingConv<[
   CCIfByVal<CCPassByVal<4, 4>>,
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index 3e1184f4586ef..b84191d3c03f1 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -397,6 +397,9 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
   setOperationAction(ISD::ROTL, MVT::i64, Expand);
   setOperationAction(ISD::ROTR, MVT::i64, Expand);
 
+  setOperationAction(ISD::MULHU, MVT::i16, Expand);
+  setOperationAction(ISD::MULHS, MVT::i16, Expand);
+
   setOperationAction(ISD::MUL, MVT::i64, Expand);
   setOperationAction(ISD::MULHU, MVT::i64, Expand);
   setOperationAction(ISD::MULHS, MVT::i64, Expand);
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUPropagateAttributes.cpp b/llvm/lib/Target/AMDGPU/AMDGPUPropagateAttributes.cpp
index 982aae3748849..dcbe4270e8a92 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUPropagateAttributes.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUPropagateAttributes.cpp
@@ -32,6 +32,7 @@
 #include "Utils/AMDGPUBaseInfo.h"
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/CodeGen/TargetPassConfig.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/Module.h"
 #include "llvm/Target/TargetMachine.h"
@@ -56,8 +57,10 @@ static constexpr const FeatureBitset TargetFeatures = {
 };
 
 // Attributes to propagate.
+// TODO: Support conservative min/max merging instead of cloning.
 static constexpr const char* AttributeNames[] = {
-  "amdgpu-waves-per-eu"
+  "amdgpu-waves-per-eu",
+  "amdgpu-flat-work-group-size"
 };
 
 static constexpr unsigned NumAttr =
@@ -371,15 +374,28 @@ AMDGPUPropagateAttributes::getFeatureString(const FeatureBitset &Features) const
 }
 
 bool AMDGPUPropagateAttributesEarly::runOnFunction(Function &F) {
-  if (!TM || !AMDGPU::isEntryFunctionCC(F.getCallingConv()))
+  if (!TM) {
+    auto *TPC = getAnalysisIfAvailable<TargetPassConfig>();
+    if (!TPC)
+      return false;
+
+    TM = &TPC->getTM<TargetMachine>();
+  }
+
+  if (!AMDGPU::isEntryFunctionCC(F.getCallingConv()))
     return false;
 
   return AMDGPUPropagateAttributes(TM, false).process(F);
 }
 
 bool AMDGPUPropagateAttributesLate::runOnModule(Module &M) {
-  if (!TM)
-    return false;
+  if (!TM) {
+    auto *TPC = getAnalysisIfAvailable<TargetPassConfig>();
+    if (!TPC)
+      return false;
+
+    TM = &TPC->getTM<TargetMachine>();
+  }
 
   return AMDGPUPropagateAttributes(TM, true).process(M);
 }
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
index 568180e20ebdb..17021fc06ca51 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPUSubtarget.h
@@ -422,10 +422,10 @@ class GCNSubtarget : public AMDGPUGenSubtargetInfo,
   SITargetLowering TLInfo;
   SIFrameLowering FrameLowering;
 
+public:
   // See COMPUTE_TMPRING_SIZE.WAVESIZE, 13-bit field in units of 256-dword.
   static const unsigned MaxWaveScratchSize = (256 * 4) * ((1 << 13) - 1);
 
-public:
   GCNSubtarget(const Triple &TT, StringRef GPU, StringRef FS,
                const GCNTargetMachine &TM);
   ~GCNSubtarget() override;
diff --git a/llvm/lib/Target/AMDGPU/BUFInstructions.td b/llvm/lib/Target/AMDGPU/BUFInstructions.td
index e1c9f1609a02a..763b2f0ef80e5 100644
--- a/llvm/lib/Target/AMDGPU/BUFInstructions.td
+++ b/llvm/lib/Target/AMDGPU/BUFInstructions.td
@@ -183,15 +183,15 @@ class getMTBUFAsmOps<int addrKind> {
 }
 
 class MTBUF_SetupAddr<int addrKind> {
-  bits<1> offen  = !if(!eq(addrKind, BUFAddrKind.OffEn), 1,
-                   !if(!eq(addrKind, BUFAddrKind.BothEn), 1 , 0));
+  bits<1> offen  = !or(!eq(addrKind, BUFAddrKind.OffEn),
+                       !eq(addrKind, BUFAddrKind.BothEn));
 
-  bits<1> idxen  = !if(!eq(addrKind, BUFAddrKind.IdxEn), 1,
-                   !if(!eq(addrKind, BUFAddrKind.BothEn), 1 , 0));
+  bits<1> idxen  = !or(!eq(addrKind, BUFAddrKind.IdxEn),
+                       !eq(addrKind, BUFAddrKind.BothEn));
 
-  bits<1> addr64 = !if(!eq(addrKind, BUFAddrKind.Addr64), 1, 0);
+  bits<1> addr64 = !eq(addrKind, BUFAddrKind.Addr64);
 
-  bits<1> has_vaddr = !if(!eq(addrKind, BUFAddrKind.Offset), 0, 1);
+  bits<1> has_vaddr = !ne(addrKind, BUFAddrKind.Offset);
 }
 
 class MTBUF_Load_Pseudo <string opName,
@@ -462,15 +462,15 @@ class getMUBUFAsmOps<int addrKind> {
 }
 
 class MUBUF_SetupAddr<int addrKind> {
-  bits<1> offen  = !if(!eq(addrKind, BUFAddrKind.OffEn), 1,
-                   !if(!eq(addrKind, BUFAddrKind.BothEn), 1 , 0));
+  bits<1> offen  = !or(!eq(addrKind, BUFAddrKind.OffEn),
+                       !eq(addrKind, BUFAddrKind.BothEn));
 
-  bits<1> idxen  = !if(!eq(addrKind, BUFAddrKind.IdxEn), 1,
-                   !if(!eq(addrKind, BUFAddrKind.BothEn), 1 , 0));
+  bits<1> idxen  = !or(!eq(addrKind, BUFAddrKind.IdxEn),
+                       !eq(addrKind, BUFAddrKind.BothEn));
 
-  bits<1> addr64 = !if(!eq(addrKind, BUFAddrKind.Addr64), 1, 0);
+  bits<1> addr64 = !eq(addrKind, BUFAddrKind.Addr64);
 
-  bits<1> has_vaddr = !if(!eq(addrKind, BUFAddrKind.Offset), 0, 1);
+  bits<1> has_vaddr = !ne(addrKind, BUFAddrKind.Offset);
 }
 
 class MUBUF_Load_Pseudo <string opName,
@@ -498,7 +498,7 @@ class MUBUF_Load_Pseudo <string opName,
   let mayStore = 0;
   let maybeAtomic = 1;
   let Uses = !if(isLds, [EXEC, M0], [EXEC]);
-  let has_tfe = !if(isLds, 0, 1);
+  let has_tfe = !not(isLds);
   let lds = isLds;
   let elements = getMUBUFElements<vdata_vt>.ret;
 }
@@ -1857,7 +1857,7 @@ class Base_MUBUF_Real_gfx6_gfx7_gfx10<bits<7> op, MUBUF_Pseudo ps, int ef> :
   let Inst{12}    = ps.offen;
   let Inst{13}    = ps.idxen;
   let Inst{14}    = !if(ps.has_glc, glc, ps.glc_value);
-  let Inst{16}    = !if(ps.lds, 1, 0);
+  let Inst{16}    = ps.lds;
   let Inst{24-18} = op;
   let Inst{31-26} = 0x38;
   let Inst{39-32} = !if(ps.has_vaddr, vaddr, ?);
@@ -2208,7 +2208,7 @@ class MUBUF_Real_vi <bits<7> op, MUBUF_Pseudo ps> :
   let Inst{12}    = ps.offen;
   let Inst{13}    = ps.idxen;
   let Inst{14}    = !if(ps.has_glc, glc, ps.glc_value);
-  let Inst{16}    = !if(ps.lds, 1, 0);
+  let Inst{16}    = ps.lds;
   let Inst{17}    = !if(ps.has_slc, slc, ?);
   let Inst{24-18} = op;
   let Inst{31-26} = 0x38; //encoding
@@ -2258,7 +2258,7 @@ class MUBUF_Real_gfx80 <bits<7> op, MUBUF_Pseudo ps> :
   let Inst{12}    = ps.offen;
   let Inst{13}    = ps.idxen;
   let Inst{14}    = !if(ps.has_glc, glc, ps.glc_value);
-  let Inst{16}    = !if(ps.lds, 1, 0);
+  let Inst{16}    = ps.lds;
   let Inst{17}    = !if(ps.has_slc, slc, ?);
   let Inst{24-18} = op;
   let Inst{31-26} = 0x38; //encoding
diff --git a/llvm/lib/Target/AMDGPU/DSInstructions.td b/llvm/lib/Target/AMDGPU/DSInstructions.td
index 96345d07c95dc..6f6cd23647030 100644
--- a/llvm/lib/Target/AMDGPU/DSInstructions.td
+++ b/llvm/lib/Target/AMDGPU/DSInstructions.td
@@ -166,12 +166,12 @@ class DS_1A1D_RET <string opName, RegisterClass rc = VGPR_32>
 multiclass DS_1A1D_RET_mc <string opName, RegisterClass rc = VGPR_32,
                            string NoRetOp = ""> {
   def "" : DS_1A1D_RET<opName, rc>,
-    AtomicNoRet<NoRetOp, !if(!eq(NoRetOp, ""), 0, 1)>;
+    AtomicNoRet<NoRetOp, !ne(NoRetOp, "")>;
 
   let has_m0_read = 0 in {
     def _gfx9 : DS_1A1D_RET<opName, rc>,
       AtomicNoRet<!if(!eq(NoRetOp, ""), "", NoRetOp#"_gfx9"),
-                  !if(!eq(NoRetOp, ""), 0, 1)>;
+                  !ne(NoRetOp, "")>;
   }
 }
 
@@ -191,11 +191,11 @@ multiclass DS_1A2D_RET_mc<string opName,
                           string NoRetOp = "",
                           RegisterClass src = rc> {
   def "" : DS_1A2D_RET<opName, rc, src>,
-    AtomicNoRet<NoRetOp, !if(!eq(NoRetOp, ""), 0, 1)>;
+    AtomicNoRet<NoRetOp, !ne(NoRetOp, "")>;
 
   let has_m0_read = 0 in {
     def _gfx9 : DS_1A2D_RET<opName, rc, src>,
-      AtomicNoRet<NoRetOp#"_gfx9", !if(!eq(NoRetOp, ""), 0, 1)>;
+      AtomicNoRet<NoRetOp#"_gfx9", !ne(NoRetOp, "")>;
   }
 }
 
diff --git a/llvm/lib/Target/AMDGPU/FLATInstructions.td b/llvm/lib/Target/AMDGPU/FLATInstructions.td
index e7f6b0ca3b553..29a350d5d2a3a 100644
--- a/llvm/lib/Target/AMDGPU/FLATInstructions.td
+++ b/llvm/lib/Target/AMDGPU/FLATInstructions.td
@@ -63,9 +63,9 @@ class FLAT_Pseudo<string opName, dag outs, dag ins,
   // Buffer instruction; so, they increment both VM_CNT and LGKM_CNT
   // and are not considered done until both have been decremented.
   let VM_CNT = 1;
-  let LGKM_CNT = !if(!or(is_flat_global, is_flat_scratch), 0, 1);
+  let LGKM_CNT = !not(!or(is_flat_global, is_flat_scratch));
 
-  let IsNonFlatSeg = !if(!or(is_flat_global, is_flat_scratch), 1, 0);
+  let IsNonFlatSeg = !or(is_flat_global, is_flat_scratch);
 }
 
 class FLAT_Real <bits<7> op, FLAT_Pseudo ps> :
diff --git a/llvm/lib/Target/AMDGPU/MIMGInstructions.td b/llvm/lib/Target/AMDGPU/MIMGInstructions.td
index f56b8728e64c0..e9ee87283553e 100644
--- a/llvm/lib/Target/AMDGPU/MIMGInstructions.td
+++ b/llvm/lib/Target/AMDGPU/MIMGInstructions.td
@@ -148,7 +148,7 @@ class MIMG_Base <dag outs, string dns = "">
   let hasSideEffects = 0; // XXX ????
 
   let DecoderNamespace = dns;
-  let isAsmParserOnly = !if(!eq(dns,""), 1, 0);
+  let isAsmParserOnly = !eq(dns, "");
 }
 
 class MIMG <dag outs, string dns = "">
@@ -308,13 +308,13 @@ multiclass MIMG_NoSampler_Src_Helper <bits<8> op, string asm,
 multiclass MIMG_NoSampler <bits<8> op, string asm, bit has_d16, bit mip = 0,
                            bit isResInfo = 0> {
   def "" : MIMGBaseOpcode {
-    let Coordinates = !if(isResInfo, 0, 1);
+    let Coordinates = !not(isResInfo);
     let LodOrClampOrMip = mip;
     let HasD16 = has_d16;
   }
 
   let BaseOpcode = !cast<MIMGBaseOpcode>(NAME),
-      mayLoad = !if(isResInfo, 0, 1) in {
+      mayLoad = !not(isResInfo) in {
     let VDataDwords = 1 in
     defm _V1 : MIMG_NoSampler_Src_Helper <op, asm, VGPR_32, 1>;
     let VDataDwords = 2 in
@@ -665,12 +665,12 @@ multiclass MIMG_Sampler <bits<8> op, AMDGPUSampleVariant sample, bit wqm = 0,
                          bit isG16 = 0, bit isGetLod = 0,
                          string asm = "image_sample"#sample.LowerCaseMod#!if(isG16, "_g16", "")> {
   def "" : MIMG_Sampler_BaseOpcode<sample> {
-    let HasD16 = !if(isGetLod, 0, 1);
+    let HasD16 = !not(isGetLod);
     let G16 = isG16;
   }
 
   let BaseOpcode = !cast<MIMGBaseOpcode>(NAME), WQM = wqm,
-      mayLoad = !if(isGetLod, 0, 1) in {
+      mayLoad = !not(isGetLod) in {
     let VDataDwords = 1 in
     defm _V1 : MIMG_Sampler_Src_Helper<op, asm, sample, VGPR_32, 1>;
     let VDataDwords = 2 in
@@ -712,8 +712,8 @@ class MIMG_IntersectRay_gfx10<int op, string opcode, RegisterClass AddrRC, bit A
     : MIMG_gfx10<op, (outs VReg_128:$vdata), "AMDGPU"> {
 
   let InOperandList = !con((ins AddrRC:$vaddr0, SReg_128:$srsrc),
-                           !if(!eq(A16,1), (ins GFX10A16:$a16), (ins)));
-  let AsmString = opcode#" $vdata, $vaddr0, $srsrc"#!if(!eq(A16,1), "$a16", "");
+                           !if(A16, (ins GFX10A16:$a16), (ins)));
+  let AsmString = opcode#" $vdata, $vaddr0, $srsrc"#!if(A16, "$a16", "");
 
   let nsa = 0;
 }
@@ -722,15 +722,15 @@ class MIMG_IntersectRay_nsa_gfx10<int op, string opcode, int num_addrs, bit A16>
     : MIMG_nsa_gfx10<op, (outs VReg_128:$vdata), num_addrs, "AMDGPU"> {
   let InOperandList = !con(nsah.AddrIns,
                            (ins SReg_128:$srsrc),
-                           !if(!eq(A16,1), (ins GFX10A16:$a16), (ins)));
-  let AsmString = opcode#" $vdata, "#nsah.AddrAsm#", $srsrc"#!if(!eq(A16,1), "$a16", "");
+                           !if(A16, (ins GFX10A16:$a16), (ins)));
+  let AsmString = opcode#" $vdata, "#nsah.AddrAsm#", $srsrc"#!if(A16, "$a16", "");
 }
 
 multiclass MIMG_IntersectRay<int op, string opcode, int num_addrs, bit A16> {
   def "" : MIMGBaseOpcode;
   let SubtargetPredicate = HasGFX10_BEncoding,
       AssemblerPredicate = HasGFX10_BEncoding,
-      AsmMatchConverter = !if(!eq(A16,1), "cvtIntersectRay", ""),
+      AsmMatchConverter = !if(A16, "cvtIntersectRay", ""),
       dmask = 0xf,
       unorm = 1,
       d16 = 0,
diff --git a/llvm/lib/Target/AMDGPU/R600InstrInfo.cpp b/llvm/lib/Target/AMDGPU/R600InstrInfo.cpp
index cf5791f39e562..e3439e6c9e884 100644
--- a/llvm/lib/Target/AMDGPU/R600InstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/R600InstrInfo.cpp
@@ -963,8 +963,9 @@ R600InstrInfo::reverseBranchCondition(SmallVectorImpl<MachineOperand> &Cond) con
   return false;
 }
 
-bool R600InstrInfo::DefinesPredicate(MachineInstr &MI,
-                                     std::vector<MachineOperand> &Pred) const {
+bool R600InstrInfo::ClobbersPredicate(MachineInstr &MI,
+                                      std::vector<MachineOperand> &Pred,
+                                      bool SkipDead) const {
   return isPredicateSetter(MI.getOpcode());
 }
 
diff --git a/llvm/lib/Target/AMDGPU/R600InstrInfo.h b/llvm/lib/Target/AMDGPU/R600InstrInfo.h
index 873ee08470cb3..1e249c6348f14 100644
--- a/llvm/lib/Target/AMDGPU/R600InstrInfo.h
+++ b/llvm/lib/Target/AMDGPU/R600InstrInfo.h
@@ -194,8 +194,8 @@ class R600InstrInfo final : public R600GenInstrInfo {
                            unsigned NumFCycles, unsigned ExtraFCycles,
                            BranchProbability Probability) const override;
 
-  bool DefinesPredicate(MachineInstr &MI,
-                        std::vector<MachineOperand> &Pred) const override;
+  bool ClobbersPredicate(MachineInstr &MI, std::vector<MachineOperand> &Pred,
+                         bool SkipDead) const override;
 
   bool isProfitableToUnpredicate(MachineBasicBlock &TMBB,
                                  MachineBasicBlock &FMBB) const override;
diff --git a/llvm/lib/Target/AMDGPU/SIInstrFormats.td b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
index 428c21c896d50..ca1cfc65c94a0 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrFormats.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrFormats.td
@@ -203,11 +203,11 @@ class InstSI <dag outs, dag ins, string asm = "",
   field bits<1> DisableVIDecoder = 0;
   field bits<1> DisableDecoder = 0;
 
-  let isAsmParserOnly = !if(!eq(DisableDecoder{0}, {0}), 0, 1);
+  let isAsmParserOnly = !ne(DisableDecoder{0}, {0});
   let AsmVariantName = AMDGPUAsmVariants.Default;
 
   // Avoid changing source registers in a way that violates constant bus read limitations.
-  let hasExtraSrcRegAllocReq = !if(VOP1,1,!if(VOP2,1,!if(VOP3,1,!if(VOPC,1,!if(SDWA,1, !if(VALU,1,0))))));
+  let hasExtraSrcRegAllocReq = !or(VOP1, VOP2, VOP3, VOPC, SDWA, VALU);
 }
 
 class PseudoInstSI<dag outs, dag ins, list<dag> pattern = [], string asm = "">
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index 1aa4be8aa2865..4c0e244ceb633 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -259,31 +259,25 @@ def SIdenorm_mode : SDNode<"AMDGPUISD::DENORM_MODE",
 // Returns 1 if the source arguments have modifiers, 0 if they do not.
 // XXX - do f16 instructions?
 class isFloatType<ValueType SrcVT> {
-  bit ret =
-    !if(!eq(SrcVT.Value, f16.Value), 1,
-    !if(!eq(SrcVT.Value, f32.Value), 1,
-    !if(!eq(SrcVT.Value, f64.Value), 1,
-    !if(!eq(SrcVT.Value, v2f16.Value), 1,
-    !if(!eq(SrcVT.Value, v4f16.Value), 1,
-    !if(!eq(SrcVT.Value, v2f32.Value), 1,
-    !if(!eq(SrcVT.Value, v2f64.Value), 1,
-    0)))))));
+  bit ret = !or(!eq(SrcVT.Value, f16.Value),
+                !eq(SrcVT.Value, f32.Value),
+                !eq(SrcVT.Value, f64.Value),
+                !eq(SrcVT.Value, v2f16.Value),
+                !eq(SrcVT.Value, v4f16.Value),
+                !eq(SrcVT.Value, v2f32.Value),
+                !eq(SrcVT.Value, v2f64.Value));
 }
 
 class isIntType<ValueType SrcVT> {
-  bit ret =
-    !if(!eq(SrcVT.Value, i16.Value), 1,
-    !if(!eq(SrcVT.Value, i32.Value), 1,
-    !if(!eq(SrcVT.Value, i64.Value), 1,
-    0)));
+  bit ret = !or(!eq(SrcVT.Value, i16.Value),
+                !eq(SrcVT.Value, i32.Value),
+                !eq(SrcVT.Value, i64.Value));
 }
 
 class isPackedType<ValueType SrcVT> {
-  bit ret =
-    !if(!eq(SrcVT.Value, v2i16.Value), 1,
-      !if(!eq(SrcVT.Value, v2f16.Value), 1,
-        !if(!eq(SrcVT.Value, v4f16.Value), 1, 0)
-    ));
+  bit ret = !or(!eq(SrcVT.Value, v2i16.Value),
+                !eq(SrcVT.Value, v2f16.Value),
+                !eq(SrcVT.Value, v4f16.Value));
 }
 
 //===----------------------------------------------------------------------===//
@@ -1393,8 +1387,8 @@ def HWREG {
 
 class getHwRegImm<int Reg, int Offset = 0, int Size = 32> {
   int ret = !and(!or(Reg,
-                  !or(!shl(Offset, 6),
-                      !shl(!add(Size, -1), 11))), 65535);
+                     !shl(Offset, 6),
+                     !shl(!add(Size, -1), 11)), 65535);
 }
 
 //===----------------------------------------------------------------------===//
@@ -1590,13 +1584,11 @@ class getVOP3SrcForVT<ValueType VT> {
 
 // Float or packed int
 class isModifierType<ValueType SrcVT> {
-  bit ret =
-    !if(!eq(SrcVT.Value, f16.Value), 1,
-    !if(!eq(SrcVT.Value, f32.Value), 1,
-    !if(!eq(SrcVT.Value, f64.Value), 1,
-    !if(!eq(SrcVT.Value, v2f16.Value), 1,
-    !if(!eq(SrcVT.Value, v2i16.Value), 1,
-    0)))));
+  bit ret = !or(!eq(SrcVT.Value, f16.Value),
+                !eq(SrcVT.Value, f32.Value),
+                !eq(SrcVT.Value, f64.Value),
+                !eq(SrcVT.Value, v2f16.Value),
+                !eq(SrcVT.Value, v2i16.Value));
 }
 
 // Return type of input modifiers operand for specified input operand
@@ -2114,14 +2106,6 @@ class getHasDPP <int NumSrcArgs, ValueType DstVT = i32, ValueType Src0VT = i32,
                 getHasExt<NumSrcArgs, DstVT, Src0VT, Src1VT>.ret);
 }
 
-class BitOr<bit a, bit b> {
-  bit ret = !if(a, 1, !if(b, 1, 0));
-}
-
-class BitAnd<bit a, bit b> {
-  bit ret = !if(a, !if(b, 1, 0), 0);
-}
-
 def PatGenMode {
   int NoPattern = 0;
   int Pattern   = 1;
@@ -2159,18 +2143,18 @@ class VOPProfile <list<ValueType> _ArgVT, bit _EnableF32SrcMods = 0,
   field Operand Src1ModSDWA = getSrcModSDWA<Src1VT>.ret;
 
 
-  field bit HasDst = !if(!eq(DstVT.Value, untyped.Value), 0, 1);
+  field bit HasDst = !ne(DstVT.Value, untyped.Value);
   field bit HasDst32 = HasDst;
   field bit EmitDst = HasDst; // force dst encoding, see v_movreld_b32 special case
   field int NumSrcArgs = getNumSrcArgs<Src0VT, Src1VT, Src2VT>.ret;
-  field bit HasSrc0 = !if(!eq(Src0VT.Value, untyped.Value), 0, 1);
-  field bit HasSrc1 = !if(!eq(Src1VT.Value, untyped.Value), 0, 1);
-  field bit HasSrc2 = !if(!eq(Src2VT.Value, untyped.Value), 0, 1);
+  field bit HasSrc0 = !ne(Src0VT.Value, untyped.Value);
+  field bit HasSrc1 = !ne(Src1VT.Value, untyped.Value);
+  field bit HasSrc2 = !ne(Src2VT.Value, untyped.Value);
 
   // TODO: Modifiers logic is somewhat adhoc here, to be refined later
   // HasModifiers affects the normal and DPP encodings. We take note of EnableF32SrcMods, which
   // enables modifiers for i32 type.
-  field bit HasModifiers = BitOr<isModifierType<Src0VT>.ret, EnableF32SrcMods>.ret;
+  field bit HasModifiers = !or(isModifierType<Src0VT>.ret, EnableF32SrcMods);
 
   // HasSrc*FloatMods affects the SDWA encoding. We ignore EnableF32SrcMods.
   field bit HasSrc0FloatMods = isFloatType<Src0VT>.ret;
@@ -2183,15 +2167,15 @@ class VOPProfile <list<ValueType> _ArgVT, bit _EnableF32SrcMods = 0,
   field bit HasSrc2IntMods = isIntType<Src2VT>.ret;
 
   field bit HasSrc0Mods = HasModifiers;
-  field bit HasSrc1Mods = !if(HasModifiers, BitOr<HasSrc1FloatMods, HasSrc1IntMods>.ret, 0);
-  field bit HasSrc2Mods = !if(HasModifiers, BitOr<HasSrc2FloatMods, HasSrc2IntMods>.ret, 0);
+  field bit HasSrc1Mods = !if(HasModifiers, !or(HasSrc1FloatMods, HasSrc1IntMods), 0);
+  field bit HasSrc2Mods = !if(HasModifiers, !or(HasSrc2FloatMods, HasSrc2IntMods), 0);
 
-  field bit HasClamp = BitOr<isModifierType<Src0VT>.ret, EnableClamp>.ret;
+  field bit HasClamp = !or(isModifierType<Src0VT>.ret, EnableClamp);
   field bit HasSDWAClamp = EmitDst;
-  field bit HasFPClamp = BitAnd<isFloatType<DstVT>.ret, HasClamp>.ret;
+  field bit HasFPClamp = !and(isFloatType<DstVT>.ret, HasClamp);
   field bit HasIntClamp = !if(isFloatType<DstVT>.ret, 0, HasClamp);
   field bit HasClampLo = HasClamp;
-  field bit HasClampHi = BitAnd<isPackedType<DstVT>.ret, HasClamp>.ret;
+  field bit HasClampHi = !and(isPackedType<DstVT>.ret, HasClamp);
   field bit HasHigh = 0;
 
   field bit IsPacked = isPackedType<Src0VT>.ret;
@@ -2573,11 +2557,11 @@ def getVCMPXNoSDstOp : InstrMapping {
 
 // Maps a SOPP to a SOPP with S_NOP
 def getSOPPWithRelaxation : InstrMapping {
-  let FilterClass = "Base_SOPP";
-  let RowFields = ["AsmString"];
-  let ColFields = ["Size"];
-  let KeyCol = ["4"];
-  let ValueCols = [["8"]];
+  let FilterClass = "SOPPRelaxTable";
+  let RowFields = ["KeyName"];
+  let ColFields = ["IsRelaxed"];
+  let KeyCol = ["0"];
+  let ValueCols = [["1"]];
 }
 
 include "SIInstructions.td"
diff --git a/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp b/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
index 708a6dec8fda7..83c60b0fb197b 100644
--- a/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMachineFunctionInfo.cpp
@@ -352,6 +352,8 @@ bool SIMachineFunctionInfo::reserveVGPRforSGPRSpills(MachineFunction &MF) {
 
   Register LaneVGPR = TRI->findUnusedRegister(
       MF.getRegInfo(), &AMDGPU::VGPR_32RegClass, MF, true);
+  if (LaneVGPR == Register())
+    return false;
   SpillVGPRs.push_back(SGPRSpillVGPRCSR(LaneVGPR, None));
   FuncInfo->VGPRReservedForSGPRSpill = LaneVGPR;
   return true;
diff --git a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
index 9c2a492a414c5..b5536cf3664d0 100644
--- a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
@@ -283,16 +283,6 @@ class SICacheControl {
   virtual bool enableNonTemporal(const MachineBasicBlock::iterator &MI)
     const = 0;
 
-  /// Inserts any necessary instructions at position \p Pos relative to
-  /// instruction \p MI to ensure any subsequent memory instructions of this
-  /// thread with address spaces \p AddrSpace will observe the previous memory
-  /// operations by any thread for memory scopes up to memory scope \p Scope .
-  /// Returns true iff any instructions inserted.
-  virtual bool insertAcquire(MachineBasicBlock::iterator &MI,
-                             SIAtomicScope Scope,
-                             SIAtomicAddrSpace AddrSpace,
-                             Position Pos) const = 0;
-
   /// Inserts any necessary instructions at position \p Pos relative
   /// to instruction \p MI to ensure memory instructions before \p Pos of kind
   /// \p Op associated with address spaces \p AddrSpace have completed. Used
@@ -307,6 +297,16 @@ class SICacheControl {
                           bool IsCrossAddrSpaceOrdering,
                           Position Pos) const = 0;
 
+  /// Inserts any necessary instructions at position \p Pos relative to
+  /// instruction \p MI to ensure any subsequent memory instructions of this
+  /// thread with address spaces \p AddrSpace will observe the previous memory
+  /// operations by any thread for memory scopes up to memory scope \p Scope .
+  /// Returns true iff any instructions inserted.
+  virtual bool insertAcquire(MachineBasicBlock::iterator &MI,
+                             SIAtomicScope Scope,
+                             SIAtomicAddrSpace AddrSpace,
+                             Position Pos) const = 0;
+
   /// Inserts any necessary instructions at position \p Pos relative to
   /// instruction \p MI to ensure previous memory instructions by this thread
   /// with address spaces \p AddrSpace have completed and can be observed by
@@ -349,6 +349,13 @@ class SIGfx6CacheControl : public SICacheControl {
 
   bool enableNonTemporal(const MachineBasicBlock::iterator &MI) const override;
 
+  bool insertWait(MachineBasicBlock::iterator &MI,
+                  SIAtomicScope Scope,
+                  SIAtomicAddrSpace AddrSpace,
+                  SIMemOp Op,
+                  bool IsCrossAddrSpaceOrdering,
+                  Position Pos) const override;
+
   bool insertAcquire(MachineBasicBlock::iterator &MI,
                      SIAtomicScope Scope,
                      SIAtomicAddrSpace AddrSpace,
@@ -359,13 +366,6 @@ class SIGfx6CacheControl : public SICacheControl {
                      SIAtomicAddrSpace AddrSpace,
                      bool IsCrossAddrSpaceOrdering,
                      Position Pos) const override;
-
-  bool insertWait(MachineBasicBlock::iterator &MI,
-                  SIAtomicScope Scope,
-                  SIAtomicAddrSpace AddrSpace,
-                  SIMemOp Op,
-                  bool IsCrossAddrSpaceOrdering,
-                  Position Pos) const override;
 };
 
 class SIGfx7CacheControl : public SIGfx6CacheControl {
@@ -399,17 +399,17 @@ class SIGfx10CacheControl : public SIGfx7CacheControl {
 
   bool enableNonTemporal(const MachineBasicBlock::iterator &MI) const override;
 
-  bool insertAcquire(MachineBasicBlock::iterator &MI,
-                     SIAtomicScope Scope,
-                     SIAtomicAddrSpace AddrSpace,
-                     Position Pos) const override;
-
   bool insertWait(MachineBasicBlock::iterator &MI,
                   SIAtomicScope Scope,
                   SIAtomicAddrSpace AddrSpace,
                   SIMemOp Op,
                   bool IsCrossAddrSpaceOrdering,
                   Position Pos) const override;
+
+  bool insertAcquire(MachineBasicBlock::iterator &MI,
+                     SIAtomicScope Scope,
+                     SIAtomicAddrSpace AddrSpace,
+                     Position Pos) const override;
 };
 
 class SIMemoryLegalizer final : public MachineFunctionPass {
@@ -734,51 +734,6 @@ bool SIGfx6CacheControl::enableNonTemporal(
   return Changed;
 }
 
-bool SIGfx6CacheControl::insertAcquire(MachineBasicBlock::iterator &MI,
-                                       SIAtomicScope Scope,
-                                       SIAtomicAddrSpace AddrSpace,
-                                       Position Pos) const {
-  if (!InsertCacheInv)
-    return false;
-
-  bool Changed = false;
-
-  MachineBasicBlock &MBB = *MI->getParent();
-  DebugLoc DL = MI->getDebugLoc();
-
-  if (Pos == Position::AFTER)
-    ++MI;
-
-  if ((AddrSpace & SIAtomicAddrSpace::GLOBAL) != SIAtomicAddrSpace::NONE) {
-    switch (Scope) {
-    case SIAtomicScope::SYSTEM:
-    case SIAtomicScope::AGENT:
-      BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_WBINVL1));
-      Changed = true;
-      break;
-    case SIAtomicScope::WORKGROUP:
-    case SIAtomicScope::WAVEFRONT:
-    case SIAtomicScope::SINGLETHREAD:
-      // No cache to invalidate.
-      break;
-    default:
-      llvm_unreachable("Unsupported synchronization scope");
-    }
-  }
-
-  /// The scratch address space does not need the global memory cache
-  /// to be flushed as all memory operations by the same thread are
-  /// sequentially consistent, and no other thread can access scratch
-  /// memory.
-
-  /// Other address spaces do not have a cache.
-
-  if (Pos == Position::AFTER)
-    --MI;
-
-  return Changed;
-}
-
 bool SIGfx6CacheControl::insertWait(MachineBasicBlock::iterator &MI,
                                     SIAtomicScope Scope,
                                     SIAtomicAddrSpace AddrSpace,
@@ -875,6 +830,51 @@ bool SIGfx6CacheControl::insertWait(MachineBasicBlock::iterator &MI,
   return Changed;
 }
 
+bool SIGfx6CacheControl::insertAcquire(MachineBasicBlock::iterator &MI,
+                                       SIAtomicScope Scope,
+                                       SIAtomicAddrSpace AddrSpace,
+                                       Position Pos) const {
+  if (!InsertCacheInv)
+    return false;
+
+  bool Changed = false;
+
+  MachineBasicBlock &MBB = *MI->getParent();
+  DebugLoc DL = MI->getDebugLoc();
+
+  if (Pos == Position::AFTER)
+    ++MI;
+
+  if ((AddrSpace & SIAtomicAddrSpace::GLOBAL) != SIAtomicAddrSpace::NONE) {
+    switch (Scope) {
+    case SIAtomicScope::SYSTEM:
+    case SIAtomicScope::AGENT:
+      BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_WBINVL1));
+      Changed = true;
+      break;
+    case SIAtomicScope::WORKGROUP:
+    case SIAtomicScope::WAVEFRONT:
+    case SIAtomicScope::SINGLETHREAD:
+      // No cache to invalidate.
+      break;
+    default:
+      llvm_unreachable("Unsupported synchronization scope");
+    }
+  }
+
+  /// The scratch address space does not need the global memory cache
+  /// to be flushed as all memory operations by the same thread are
+  /// sequentially consistent, and no other thread can access scratch
+  /// memory.
+
+  /// Other address spaces do not have a cache.
+
+  if (Pos == Position::AFTER)
+    --MI;
+
+  return Changed;
+}
+
 bool SIGfx6CacheControl::insertRelease(MachineBasicBlock::iterator &MI,
                                        SIAtomicScope Scope,
                                        SIAtomicAddrSpace AddrSpace,
@@ -989,61 +989,6 @@ bool SIGfx10CacheControl::enableNonTemporal(
   return Changed;
 }
 
-bool SIGfx10CacheControl::insertAcquire(MachineBasicBlock::iterator &MI,
-                                        SIAtomicScope Scope,
-                                        SIAtomicAddrSpace AddrSpace,
-                                        Position Pos) const {
-  if (!InsertCacheInv)
-    return false;
-
-  bool Changed = false;
-
-  MachineBasicBlock &MBB = *MI->getParent();
-  DebugLoc DL = MI->getDebugLoc();
-
-  if (Pos == Position::AFTER)
-    ++MI;
-
-  if ((AddrSpace & SIAtomicAddrSpace::GLOBAL) != SIAtomicAddrSpace::NONE) {
-    switch (Scope) {
-    case SIAtomicScope::SYSTEM:
-    case SIAtomicScope::AGENT:
-      BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_GL0_INV));
-      BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_GL1_INV));
-      Changed = true;
-      break;
-    case SIAtomicScope::WORKGROUP:
-      // In WGP mode the waves of a work-group can be executing on either CU of
-      // the WGP. Therefore need to invalidate the L0 which is per CU. Otherwise
-      // in CU mode and all waves of a work-group are on the same CU, and so the
-      // L0 does not need to be invalidated.
-      if (!ST.isCuModeEnabled()) {
-        BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_GL0_INV));
-        Changed = true;
-      }
-      break;
-    case SIAtomicScope::WAVEFRONT:
-    case SIAtomicScope::SINGLETHREAD:
-      // No cache to invalidate.
-      break;
-    default:
-      llvm_unreachable("Unsupported synchronization scope");
-    }
-  }
-
-  /// The scratch address space does not need the global memory cache
-  /// to be flushed as all memory operations by the same thread are
-  /// sequentially consistent, and no other thread can access scratch
-  /// memory.
-
-  /// Other address spaces do not have a cache.
-
-  if (Pos == Position::AFTER)
-    --MI;
-
-  return Changed;
-}
-
 bool SIGfx10CacheControl::insertWait(MachineBasicBlock::iterator &MI,
                                      SIAtomicScope Scope,
                                      SIAtomicAddrSpace AddrSpace,
@@ -1163,6 +1108,61 @@ bool SIGfx10CacheControl::insertWait(MachineBasicBlock::iterator &MI,
   return Changed;
 }
 
+bool SIGfx10CacheControl::insertAcquire(MachineBasicBlock::iterator &MI,
+                                        SIAtomicScope Scope,
+                                        SIAtomicAddrSpace AddrSpace,
+                                        Position Pos) const {
+  if (!InsertCacheInv)
+    return false;
+
+  bool Changed = false;
+
+  MachineBasicBlock &MBB = *MI->getParent();
+  DebugLoc DL = MI->getDebugLoc();
+
+  if (Pos == Position::AFTER)
+    ++MI;
+
+  if ((AddrSpace & SIAtomicAddrSpace::GLOBAL) != SIAtomicAddrSpace::NONE) {
+    switch (Scope) {
+    case SIAtomicScope::SYSTEM:
+    case SIAtomicScope::AGENT:
+      BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_GL0_INV));
+      BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_GL1_INV));
+      Changed = true;
+      break;
+    case SIAtomicScope::WORKGROUP:
+      // In WGP mode the waves of a work-group can be executing on either CU of
+      // the WGP. Therefore need to invalidate the L0 which is per CU. Otherwise
+      // in CU mode and all waves of a work-group are on the same CU, and so the
+      // L0 does not need to be invalidated.
+      if (!ST.isCuModeEnabled()) {
+        BuildMI(MBB, MI, DL, TII->get(AMDGPU::BUFFER_GL0_INV));
+        Changed = true;
+      }
+      break;
+    case SIAtomicScope::WAVEFRONT:
+    case SIAtomicScope::SINGLETHREAD:
+      // No cache to invalidate.
+      break;
+    default:
+      llvm_unreachable("Unsupported synchronization scope");
+    }
+  }
+
+  /// The scratch address space does not need the global memory cache
+  /// to be flushed as all memory operations by the same thread are
+  /// sequentially consistent, and no other thread can access scratch
+  /// memory.
+
+  /// Other address spaces do not have a cache.
+
+  if (Pos == Position::AFTER)
+    --MI;
+
+  return Changed;
+}
+
 bool SIMemoryLegalizer::removeAtomicPseudoMIs() {
   if (AtomicPseudoMIs.empty())
     return false;
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index c7a0121b058be..a4e94448958fc 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -98,9 +98,10 @@ SIRegisterInfo::SIRegisterInfo(const GCNSubtarget &ST)
       Width = SubRegFromChannelTableWidthMap[Width];
       if (Width == 0)
         continue;
-      assert((Width - 1) < SubRegFromChannelTable.size());
-      assert(Offset < SubRegFromChannelTable[Width].size());
-      SubRegFromChannelTable[Width - 1][Offset] = Idx;
+      unsigned TableIdx = Width - 1;
+      assert(TableIdx < SubRegFromChannelTable.size());
+      assert(Offset < SubRegFromChannelTable[TableIdx].size());
+      SubRegFromChannelTable[TableIdx][Offset] = Idx;
     }
   };
 
@@ -151,6 +152,10 @@ const uint32_t *SIRegisterInfo::getCallPreservedMask(const MachineFunction &MF,
   }
 }
 
+const uint32_t *SIRegisterInfo::getNoPreservedMask() const {
+  return CSR_AMDGPU_NoRegs_RegMask;
+}
+
 Register SIRegisterInfo::getFrameRegister(const MachineFunction &MF) const {
   const SIFrameLowering *TFI =
       MF.getSubtarget<GCNSubtarget>().getFrameLowering();
@@ -332,9 +337,8 @@ BitVector SIRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
   for (MCPhysReg Reg : MFI->getVGPRSpillAGPRs())
     reserveRegisterTuples(Reserved, Reg);
 
-  if (MFI->VGPRReservedForSGPRSpill)
-    for (auto SSpill : MFI->getSGPRSpillVGPRs())
-      reserveRegisterTuples(Reserved, SSpill.VGPR);
+  for (auto SSpill : MFI->getSGPRSpillVGPRs())
+    reserveRegisterTuples(Reserved, SSpill.VGPR);
 
   return Reserved;
 }
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.h b/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
index 6cae479a69b58..829656d4f0fc2 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
@@ -41,7 +41,7 @@ class SIRegisterInfo final : public AMDGPUGenRegisterInfo {
   static std::array<std::vector<int16_t>, 16> RegSplitParts;
 
   // Table representing sub reg of given width and offset.
-  // First index is subreg size: 32, 64, 96, 128, 160, 192, 256, 512.
+  // First index is subreg size: 32, 64, 96, 128, 160, 192, 224, 256, 512.
   // Second index is 32 different dword offsets.
   static std::array<std::array<uint16_t, 32>, 9> SubRegFromChannelTable;
 
@@ -68,6 +68,7 @@ class SIRegisterInfo final : public AMDGPUGenRegisterInfo {
   const MCPhysReg *getCalleeSavedRegsViaCopy(const MachineFunction *MF) const;
   const uint32_t *getCallPreservedMask(const MachineFunction &MF,
                                        CallingConv::ID) const override;
+  const uint32_t *getNoPreservedMask() const override;
 
   // Stack access is very expensive. CSRs are also the high registers, and we
   // want to minimize the number of used registers.
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
index ff1f5c4bc49b1..4bbd39c1a8e5e 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
@@ -149,7 +149,7 @@ multiclass SIRegLoHi16 <string n, bits<16> regIdx, bit ArtificialHigh = 1,
                                    !cast<Register>(NAME#"_HI16")]> {
     let Namespace = "AMDGPU";
     let SubRegIndices = [lo16, hi16];
-    let CoveredBySubRegs = !if(ArtificialHigh,0,1);
+    let CoveredBySubRegs = !not(ArtificialHigh);
     let HWEncoding = regIdx;
     let HWEncoding{8} = HWEncodingHigh;
   }
diff --git a/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp b/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
index 94ade095e7a90..29cb1001fcad3 100644
--- a/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
+++ b/llvm/lib/Target/AMDGPU/SIShrinkInstructions.cpp
@@ -437,6 +437,22 @@ getSubRegForIndex(Register Reg, unsigned Sub, unsigned I,
   return TargetInstrInfo::RegSubRegPair(Reg, Sub);
 }
 
+static void dropInstructionKeepingImpDefs(MachineInstr &MI,
+                                          const SIInstrInfo *TII) {
+  for (unsigned i = MI.getDesc().getNumOperands() +
+         MI.getDesc().getNumImplicitUses() +
+         MI.getDesc().getNumImplicitDefs(), e = MI.getNumOperands();
+       i != e; ++i) {
+    const MachineOperand &Op = MI.getOperand(i);
+    if (!Op.isDef())
+      continue;
+    BuildMI(*MI.getParent(), MI.getIterator(), MI.getDebugLoc(),
+            TII->get(AMDGPU::IMPLICIT_DEF), Op.getReg());
+  }
+
+  MI.eraseFromParent();
+}
+
 // Match:
 // mov t, x
 // mov x, y
@@ -476,18 +492,25 @@ static MachineInstr* matchSwap(MachineInstr &MovT, MachineRegisterInfo &MRI,
   if (!TRI.isVGPR(MRI, X))
     return nullptr;
 
+  if (MovT.hasRegisterImplicitUseOperand(AMDGPU::M0))
+    return nullptr;
+
   const unsigned SearchLimit = 16;
   unsigned Count = 0;
+  bool KilledT = false;
   for (auto Iter = std::next(MovT.getIterator()),
             E = MovT.getParent()->instr_end();
-       Iter != E && Count < SearchLimit; ++Iter, ++Count) {
+       Iter != E && Count < SearchLimit && !KilledT; ++Iter, ++Count) {
 
     MachineInstr *MovY = &*Iter;
+    KilledT = MovY->killsRegister(T, &TRI);
+
     if ((MovY->getOpcode() != AMDGPU::V_MOV_B32_e32 &&
          MovY->getOpcode() != AMDGPU::COPY) ||
         !MovY->getOperand(1).isReg()        ||
         MovY->getOperand(1).getReg() != T   ||
-        MovY->getOperand(1).getSubReg() != Tsub)
+        MovY->getOperand(1).getSubReg() != Tsub ||
+        MovY->hasRegisterImplicitUseOperand(AMDGPU::M0))
       continue;
 
     Register Y = MovY->getOperand(0).getReg();
@@ -521,32 +544,53 @@ static MachineInstr* matchSwap(MachineInstr &MovT, MachineRegisterInfo &MRI,
         MovX = nullptr;
         break;
       }
+      // Implicit use of M0 is an indirect move.
+      if (I->hasRegisterImplicitUseOperand(AMDGPU::M0))
+        continue;
+
+      if (Size > 1 && (I->getNumImplicitOperands() > (I->isCopy() ? 0 : 1)))
+        continue;
+
       MovX = &*I;
     }
 
     if (!MovX)
       continue;
 
-    LLVM_DEBUG(dbgs() << "Matched v_swap_b32:\n" << MovT << *MovX << MovY);
+    LLVM_DEBUG(dbgs() << "Matched v_swap_b32:\n" << MovT << *MovX << *MovY);
 
     for (unsigned I = 0; I < Size; ++I) {
       TargetInstrInfo::RegSubRegPair X1, Y1;
       X1 = getSubRegForIndex(X, Xsub, I, TRI, MRI);
       Y1 = getSubRegForIndex(Y, Ysub, I, TRI, MRI);
-      BuildMI(*MovT.getParent(), MovX->getIterator(), MovT.getDebugLoc(),
-                TII->get(AMDGPU::V_SWAP_B32))
+      MachineBasicBlock &MBB = *MovT.getParent();
+      auto MIB = BuildMI(MBB, MovX->getIterator(), MovT.getDebugLoc(),
+                         TII->get(AMDGPU::V_SWAP_B32))
         .addDef(X1.Reg, 0, X1.SubReg)
         .addDef(Y1.Reg, 0, Y1.SubReg)
         .addReg(Y1.Reg, 0, Y1.SubReg)
         .addReg(X1.Reg, 0, X1.SubReg).getInstr();
+      if (MovX->hasRegisterImplicitUseOperand(AMDGPU::EXEC)) {
+        // Drop implicit EXEC.
+        MIB->RemoveOperand(MIB->getNumExplicitOperands());
+        MIB->copyImplicitOps(*MBB.getParent(), *MovX);
+      }
     }
     MovX->eraseFromParent();
-    MovY->eraseFromParent();
+    dropInstructionKeepingImpDefs(*MovY, TII);
     MachineInstr *Next = &*std::next(MovT.getIterator());
-    if (MRI.use_nodbg_empty(T))
-      MovT.eraseFromParent();
-    else
+
+    if (MRI.use_nodbg_empty(T)) {
+      dropInstructionKeepingImpDefs(MovT, TII);
+    } else {
       Xop.setIsKill(false);
+      for (int I = MovT.getNumImplicitOperands() - 1; I >= 0; --I ) {
+        unsigned OpNo = MovT.getNumExplicitOperands() + I;
+        const MachineOperand &Op = MovT.getOperand(OpNo);
+        if (Op.isKill() && TRI.regsOverlap(X, Op.getReg()))
+          MovT.RemoveOperand(OpNo);
+      }
+    }
 
     return Next;
   }
diff --git a/llvm/lib/Target/AMDGPU/SOPInstructions.td b/llvm/lib/Target/AMDGPU/SOPInstructions.td
index 76257ed1584be..f92e51d23ea8c 100644
--- a/llvm/lib/Target/AMDGPU/SOPInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SOPInstructions.td
@@ -359,9 +359,9 @@ class SOP2_Pseudo<string opName, dag outs, dag ins,
   // let Size = 4; // Do we need size here?
 }
 
-class SOP2_Real<bits<7> op, SOP_Pseudo ps> :
+class SOP2_Real<bits<7> op, SOP_Pseudo ps, string real_name = ps.Mnemonic> :
   InstSI <ps.OutOperandList, ps.InOperandList,
-          ps.Mnemonic # " " # ps.AsmOperands, []>,
+          real_name # " " # ps.AsmOperands, []>,
   Enc32 {
   let isPseudo = 0;
   let isCodeGenOnly = 0;
@@ -930,88 +930,101 @@ let SubtargetPredicate = isGFX10Plus in {
 // SOPC Instructions
 //===----------------------------------------------------------------------===//
 
-class SOPCe <bits<7> op> : Enc32 {
-  bits<8> src0;
-  bits<8> src1;
-
-  let Inst{7-0} = src0;
-  let Inst{15-8} = src1;
-  let Inst{22-16} = op;
-  let Inst{31-23} = 0x17e;
-}
-
-class SOPC <bits<7> op, dag outs, dag ins, string asm,
-            list<dag> pattern = []> :
-  InstSI<outs, ins, asm, pattern>, SOPCe <op> {
+class SOPC_Pseudo<string opName, dag outs, dag ins,
+                  string asmOps, list<dag> pattern=[]> :
+  SOP_Pseudo<opName, outs, ins, asmOps, pattern> {
   let mayLoad = 0;
   let mayStore = 0;
   let hasSideEffects = 0;
   let SALU = 1;
   let SOPC = 1;
-  let isCodeGenOnly = 0;
   let Defs = [SCC];
   let SchedRW = [WriteSALU];
   let UseNamedOperandTable = 1;
 }
 
-class SOPC_Base <bits<7> op, RegisterOperand rc0, RegisterOperand rc1,
-                 string opName, list<dag> pattern = []> : SOPC <
-  op, (outs), (ins rc0:$src0, rc1:$src1),
-  opName#" $src0, $src1", pattern > {
-  let Defs = [SCC];
+class SOPC_Real<bits<7> op, SOPC_Pseudo ps, string real_name = ps.Mnemonic> :
+  InstSI <ps.OutOperandList, ps.InOperandList,
+          real_name # " " # ps.AsmOperands, []>,
+  Enc32 {
+  let isPseudo = 0;
+  let isCodeGenOnly = 0;
+
+  // copy relevant pseudo op flags
+  let SubtargetPredicate = ps.SubtargetPredicate;
+  let OtherPredicates    = ps.OtherPredicates;
+  let AsmMatchConverter  = ps.AsmMatchConverter;
+  let UseNamedOperandTable = ps.UseNamedOperandTable;
+  let TSFlags = ps.TSFlags;
+
+  // encoding
+  bits<8> src0;
+  bits<8> src1;
+
+  let Inst{7-0} = src0;
+  let Inst{15-8} = src1;
+  let Inst{22-16} = op;
+  let Inst{31-23} = 0x17e;
+}
+
+class SOPC_Base <RegisterOperand rc0, RegisterOperand rc1,
+                 string opName, list<dag> pattern = []> : SOPC_Pseudo <
+  opName, (outs), (ins rc0:$src0, rc1:$src1),
+  "$src0, $src1", pattern > {
 }
-class SOPC_Helper <bits<7> op, RegisterOperand rc, ValueType vt,
+
+class SOPC_Helper <RegisterOperand rc, ValueType vt,
                     string opName, SDPatternOperator cond> : SOPC_Base <
-  op, rc, rc, opName,
+  rc, rc, opName,
   [(set SCC, (si_setcc_uniform vt:$src0, vt:$src1, cond))] > {
 }
 
-class SOPC_CMP_32<bits<7> op, string opName,
+class SOPC_CMP_32<string opName,
                   SDPatternOperator cond = COND_NULL, string revOp = opName>
-  : SOPC_Helper<op, SSrc_b32, i32, opName, cond>,
+  : SOPC_Helper<SSrc_b32, i32, opName, cond>,
     Commutable_REV<revOp, !eq(revOp, opName)>,
     SOPKInstTable<0, opName> {
   let isCompare = 1;
   let isCommutable = 1;
 }
 
-class SOPC_CMP_64<bits<7> op, string opName,
+class SOPC_CMP_64<string opName,
                   SDPatternOperator cond = COND_NULL, string revOp = opName>
-  : SOPC_Helper<op, SSrc_b64, i64, opName, cond>,
+  : SOPC_Helper<SSrc_b64, i64, opName, cond>,
     Commutable_REV<revOp, !eq(revOp, opName)> {
   let isCompare = 1;
   let isCommutable = 1;
 }
 
-class SOPC_32<bits<7> op, string opName, list<dag> pattern = []>
-  : SOPC_Base<op, SSrc_b32, SSrc_b32, opName, pattern>;
-
-class SOPC_64_32<bits<7> op, string opName, list<dag> pattern = []>
-  : SOPC_Base<op, SSrc_b64, SSrc_b32, opName, pattern>;
-
-def S_CMP_EQ_I32 : SOPC_CMP_32 <0x00, "s_cmp_eq_i32">;
-def S_CMP_LG_I32 : SOPC_CMP_32 <0x01, "s_cmp_lg_i32">;
-def S_CMP_GT_I32 : SOPC_CMP_32 <0x02, "s_cmp_gt_i32", COND_SGT>;
-def S_CMP_GE_I32 : SOPC_CMP_32 <0x03, "s_cmp_ge_i32", COND_SGE>;
-def S_CMP_LT_I32 : SOPC_CMP_32 <0x04, "s_cmp_lt_i32", COND_SLT, "s_cmp_gt_i32">;
-def S_CMP_LE_I32 : SOPC_CMP_32 <0x05, "s_cmp_le_i32", COND_SLE, "s_cmp_ge_i32">;
-def S_CMP_EQ_U32 : SOPC_CMP_32 <0x06, "s_cmp_eq_u32", COND_EQ>;
-def S_CMP_LG_U32 : SOPC_CMP_32 <0x07, "s_cmp_lg_u32", COND_NE>;
-def S_CMP_GT_U32 : SOPC_CMP_32 <0x08, "s_cmp_gt_u32", COND_UGT>;
-def S_CMP_GE_U32 : SOPC_CMP_32 <0x09, "s_cmp_ge_u32", COND_UGE>;
-def S_CMP_LT_U32 : SOPC_CMP_32 <0x0a, "s_cmp_lt_u32", COND_ULT, "s_cmp_gt_u32">;
-def S_CMP_LE_U32 : SOPC_CMP_32 <0x0b, "s_cmp_le_u32", COND_ULE, "s_cmp_ge_u32">;
-
-def S_BITCMP0_B32 : SOPC_32 <0x0c, "s_bitcmp0_b32">;
-def S_BITCMP1_B32 : SOPC_32 <0x0d, "s_bitcmp1_b32">;
-def S_BITCMP0_B64 : SOPC_64_32 <0x0e, "s_bitcmp0_b64">;
-def S_BITCMP1_B64 : SOPC_64_32 <0x0f, "s_bitcmp1_b64">;
+class SOPC_32<string opName, list<dag> pattern = []>
+  : SOPC_Base<SSrc_b32, SSrc_b32, opName, pattern>;
+
+class SOPC_64_32<string opName, list<dag> pattern = []>
+  : SOPC_Base<SSrc_b64, SSrc_b32, opName, pattern>;
+
+def S_CMP_EQ_I32 : SOPC_CMP_32 <"s_cmp_eq_i32">;
+def S_CMP_LG_I32 : SOPC_CMP_32 <"s_cmp_lg_i32">;
+def S_CMP_GT_I32 : SOPC_CMP_32 <"s_cmp_gt_i32", COND_SGT>;
+def S_CMP_GE_I32 : SOPC_CMP_32 <"s_cmp_ge_i32", COND_SGE>;
+def S_CMP_LT_I32 : SOPC_CMP_32 <"s_cmp_lt_i32", COND_SLT, "s_cmp_gt_i32">;
+def S_CMP_LE_I32 : SOPC_CMP_32 <"s_cmp_le_i32", COND_SLE, "s_cmp_ge_i32">;
+def S_CMP_EQ_U32 : SOPC_CMP_32 <"s_cmp_eq_u32", COND_EQ>;
+def S_CMP_LG_U32 : SOPC_CMP_32 <"s_cmp_lg_u32", COND_NE>;
+def S_CMP_GT_U32 : SOPC_CMP_32 <"s_cmp_gt_u32", COND_UGT>;
+def S_CMP_GE_U32 : SOPC_CMP_32 <"s_cmp_ge_u32", COND_UGE>;
+def S_CMP_LT_U32 : SOPC_CMP_32 <"s_cmp_lt_u32", COND_ULT, "s_cmp_gt_u32">;
+def S_CMP_LE_U32 : SOPC_CMP_32 <"s_cmp_le_u32", COND_ULE, "s_cmp_ge_u32">;
+
+def S_BITCMP0_B32 : SOPC_32 <"s_bitcmp0_b32">;
+def S_BITCMP1_B32 : SOPC_32 <"s_bitcmp1_b32">;
+def S_BITCMP0_B64 : SOPC_64_32 <"s_bitcmp0_b64">;
+def S_BITCMP1_B64 : SOPC_64_32 <"s_bitcmp1_b64">;
 let SubtargetPredicate = isGFX6GFX7GFX8GFX9 in
-def S_SETVSKIP : SOPC_32 <0x10, "s_setvskip">;
+def S_SETVSKIP : SOPC_32 <"s_setvskip">;
 
 let SubtargetPredicate = isGFX8Plus in {
-def S_CMP_EQ_U64 : SOPC_CMP_64 <0x12, "s_cmp_eq_u64", COND_EQ>;
-def S_CMP_LG_U64 : SOPC_CMP_64 <0x13, "s_cmp_lg_u64", COND_NE>;
+def S_CMP_EQ_U64 : SOPC_CMP_64 <"s_cmp_eq_u64", COND_EQ>;
+def S_CMP_LG_U64 : SOPC_CMP_64 <"s_cmp_lg_u64", COND_NE>;
 } // End SubtargetPredicate = isGFX8Plus
 
 let SubtargetPredicate = HasVGPRIndexMode in {
@@ -1019,10 +1032,11 @@ let SubtargetPredicate = HasVGPRIndexMode in {
 // register. We don't want to add mode register uses to every
 // instruction, and it's too complicated to deal with anyway. This is
 // modeled just as a side effect.
-def S_SET_GPR_IDX_ON : SOPC <0x11,
+def S_SET_GPR_IDX_ON : SOPC_Pseudo <
+  "s_set_gpr_idx_on" ,
   (outs),
   (ins SSrc_b32:$src0, GPRIdxMode:$src1),
-  "s_set_gpr_idx_on $src0,$src1"> {
+  "$src0,$src1"> {
   let Defs = [M0, MODE]; // No scc def
   let Uses = [M0, MODE]; // Other bits of mode, m0 unmodified.
   let hasSideEffects = 1; // Sets mode.gpr_idx_en
@@ -1034,225 +1048,237 @@ def S_SET_GPR_IDX_ON : SOPC <0x11,
 // SOPP Instructions
 //===----------------------------------------------------------------------===//
 
-class Base_SOPP <string asm> {
-  string AsmString = asm;
-}
-
-class SOPPe <bits<7> op> : Enc32 {
-  bits <16> simm16;
-
-  let Inst{15-0} = simm16;
-  let Inst{22-16} = op;
-  let Inst{31-23} = 0x17f; // encoding
-}
-
-class SOPP <bits<7> op, dag ins, string asm, list<dag> pattern = []> :
-  InstSI <(outs), ins, asm, pattern >, SOPPe <op>, Base_SOPP <asm> {
-
+class SOPP_Pseudo<string opName, dag ins,
+                  string asmOps = "", list<dag> pattern=[], string keyName = opName> :
+  SOP_Pseudo<opName, (outs), ins, asmOps, pattern> {
+  let isPseudo = 1;
+  let isCodeGenOnly = 1;
   let mayLoad = 0;
   let mayStore = 0;
   let hasSideEffects = 0;
   let SALU = 1;
   let SOPP = 1;
-  let Size = 4;
   let SchedRW = [WriteSALU];
-
   let UseNamedOperandTable = 1;
+  bits <16> simm16;
+  bits <1> fixed_imm = 0;
+  string KeyName = keyName;
+}
+
+class SOPPRelaxTable <bit isRelaxed, string keyName, string gfxip> {
+  bit IsRelaxed = isRelaxed;
+  string KeyName = keyName # gfxip;
 }
 
-def S_NOP : SOPP <0x00000000, (ins i16imm:$simm16), "s_nop $simm16">;
+//spaces inserted in realname on instantiation of this record to allow s_endpgm to omit whitespace
+class SOPP_Real<bits<7> op, SOPP_Pseudo ps, string real_name = ps.Mnemonic> :
+  InstSI <ps.OutOperandList, ps.InOperandList,
+          real_name # ps.AsmOperands, []> {
+  let isPseudo = 0;
+  let isCodeGenOnly = 0;
 
-class SOPP_w_nop_e <bits<7> op> : Enc64 {
+  // copy relevant pseudo op flags
+  let SubtargetPredicate = ps.SubtargetPredicate;
+  let OtherPredicates    = ps.OtherPredicates;
+  let AsmMatchConverter  = ps.AsmMatchConverter;
+  let UseNamedOperandTable = ps.UseNamedOperandTable;
+  let TSFlags = ps.TSFlags;
   bits <16> simm16;
+}
 
-  let Inst{15-0} = simm16;
+class SOPP_Real_32 <bits<7> op, SOPP_Pseudo ps, string real_name = ps.Mnemonic> : SOPP_Real<op, ps, real_name>,
+Enc32 {
+  let Inst{15-0} = !if(ps.fixed_imm, ps.simm16, simm16);
   let Inst{22-16} = op;
-  let Inst{31-23} = 0x17f; // encoding
-  let Inst{47-32} = 0x0;
-  let Inst{54-48} = S_NOP.Inst{22-16}; // opcode
-  let Inst{63-55} = S_NOP.Inst{31-23}; // encoding
+  let Inst{31-23} = 0x17f;
 }
 
-class SOPP_w_nop <bits<7> op, dag ins, string asm, list<dag> pattern = []> :
-  InstSI <(outs), ins, asm, pattern >, SOPP_w_nop_e <op>, Base_SOPP <asm> {
-
-  let mayLoad = 0;
-  let mayStore = 0;
-  let hasSideEffects = 0;
-  let SALU = 1;
-  let SOPP = 1;
-  let Size = 8;
-  let SchedRW = [WriteSALU];
-
-  let UseNamedOperandTable = 1;
+class SOPP_Real_64 <bits<7> op, SOPP_Pseudo ps, string real_name = ps.Mnemonic> : SOPP_Real<op, ps, real_name>,
+Enc64 {
+  // encoding
+  let Inst{15-0} = !if(ps.fixed_imm, ps.simm16, simm16);
+  let Inst{22-16} = op;
+  let Inst{31-23} = 0x17f;
+  //effectively a nop
+  let Inst{47-32} = 0x0;
+  let Inst{54-48} = 0x0;
+  let Inst{63-55} = 0x17f;
 }
 
-multiclass SOPP_With_Relaxation <bits<7> op, dag ins, string asm, list<dag> pattern = []> {
-  def "" : SOPP <op, ins, asm, pattern>;
-  def _pad_s_nop : SOPP_w_nop <op, ins, asm, pattern>;
+multiclass SOPP_With_Relaxation <string opName, dag ins,
+                  string asmOps, list<dag> pattern=[]> {
+  def "" : SOPP_Pseudo <opName, ins, asmOps, pattern>;
+  def _pad_s_nop : SOPP_Pseudo <opName # "_pad_s_nop", ins, asmOps, pattern, opName>;
 }
 
-let isTerminator = 1 in {
+def S_NOP : SOPP_Pseudo<"s_nop" , (ins i16imm:$simm16), "$simm16">;
 
-def S_ENDPGM : SOPP <0x00000001, (ins EndpgmImm:$simm16), "s_endpgm$simm16"> {
+let isTerminator = 1 in {
+def S_ENDPGM : SOPP_Pseudo<"s_endpgm", (ins EndpgmImm:$simm16), "$simm16"> {
   let isBarrier = 1;
   let isReturn = 1;
 }
 
-def S_ENDPGM_SAVED : SOPP <0x0000001B, (ins), "s_endpgm_saved"> {
+def S_ENDPGM_SAVED : SOPP_Pseudo<"s_endpgm_saved", (ins)> {
   let SubtargetPredicate = isGFX8Plus;
   let simm16 = 0;
+  let fixed_imm = 1;
   let isBarrier = 1;
   let isReturn = 1;
 }
 
 let SubtargetPredicate = isGFX9Plus in {
-  let isBarrier = 1, isReturn = 1, simm16 = 0 in {
+  let isBarrier = 1, isReturn = 1, simm16 = 0, fixed_imm = 1 in {
     def S_ENDPGM_ORDERED_PS_DONE :
-      SOPP<0x01e, (ins), "s_endpgm_ordered_ps_done">;
-  } // End isBarrier = 1, isReturn = 1, simm16 = 0
+      SOPP_Pseudo<"s_endpgm_ordered_ps_done", (ins)>;
+  } // End isBarrier = 1, isReturn = 1, simm16 = 0, fixed_imm = 1
 } // End SubtargetPredicate = isGFX9Plus
 
 let SubtargetPredicate = isGFX10Plus in {
-  let isBarrier = 1, isReturn = 1, simm16 = 0 in {
+  let isBarrier = 1, isReturn = 1, simm16 = 0, fixed_imm = 1 in {
     def S_CODE_END :
-      SOPP<0x01f, (ins), "s_code_end">;
-  } // End isBarrier = 1, isReturn = 1, simm16 = 0
+      SOPP_Pseudo<"s_code_end", (ins)>;
+  } // End isBarrier = 1, isReturn = 1, simm16 = 0, fixed_imm = 1
 } // End SubtargetPredicate = isGFX10Plus
 
 let isBranch = 1, SchedRW = [WriteBranch] in {
 let isBarrier = 1 in {
-defm S_BRANCH : SOPP_With_Relaxation <
-  0x00000002, (ins sopp_brtarget:$simm16), "s_branch $simm16",
+defm S_BRANCH : SOPP_With_Relaxation<
+  "s_branch" , (ins sopp_brtarget:$simm16), "$simm16",
   [(br bb:$simm16)]>;
 }
 
 let Uses = [SCC] in {
-defm S_CBRANCH_SCC0 : SOPP_With_Relaxation <
-  0x00000004, (ins sopp_brtarget:$simm16),
-  "s_cbranch_scc0 $simm16"
+defm S_CBRANCH_SCC0 : SOPP_With_Relaxation<
+  "s_cbranch_scc0" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 defm S_CBRANCH_SCC1 : SOPP_With_Relaxation <
-  0x00000005, (ins sopp_brtarget:$simm16),
-  "s_cbranch_scc1 $simm16"
+  "s_cbranch_scc1" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 } // End Uses = [SCC]
 
 let Uses = [VCC] in {
 defm S_CBRANCH_VCCZ : SOPP_With_Relaxation <
-  0x00000006, (ins sopp_brtarget:$simm16),
-  "s_cbranch_vccz $simm16"
+  "s_cbranch_vccz" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 defm S_CBRANCH_VCCNZ : SOPP_With_Relaxation <
-  0x00000007, (ins sopp_brtarget:$simm16),
-  "s_cbranch_vccnz $simm16"
+  "s_cbranch_vccnz" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 } // End Uses = [VCC]
 
 let Uses = [EXEC] in {
 defm S_CBRANCH_EXECZ : SOPP_With_Relaxation <
-  0x00000008, (ins sopp_brtarget:$simm16),
-  "s_cbranch_execz $simm16"
+  "s_cbranch_execz" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 defm S_CBRANCH_EXECNZ : SOPP_With_Relaxation <
-  0x00000009, (ins sopp_brtarget:$simm16),
-  "s_cbranch_execnz $simm16"
+  "s_cbranch_execnz" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 } // End Uses = [EXEC]
 
 defm S_CBRANCH_CDBGSYS : SOPP_With_Relaxation <
-  0x00000017, (ins sopp_brtarget:$simm16),
-  "s_cbranch_cdbgsys $simm16"
+  "s_cbranch_cdbgsys" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 
 defm S_CBRANCH_CDBGSYS_AND_USER : SOPP_With_Relaxation <
-  0x0000001A, (ins sopp_brtarget:$simm16),
-  "s_cbranch_cdbgsys_and_user $simm16"
+  "s_cbranch_cdbgsys_and_user" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 
 defm S_CBRANCH_CDBGSYS_OR_USER : SOPP_With_Relaxation <
-  0x00000019, (ins sopp_brtarget:$simm16),
-  "s_cbranch_cdbgsys_or_user $simm16"
+  "s_cbranch_cdbgsys_or_user" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 
 defm S_CBRANCH_CDBGUSER : SOPP_With_Relaxation <
-  0x00000018, (ins sopp_brtarget:$simm16),
-  "s_cbranch_cdbguser $simm16"
+  "s_cbranch_cdbguser" , (ins sopp_brtarget:$simm16),
+  "$simm16"
 >;
 
 } // End isBranch = 1
 } // End isTerminator = 1
 
 let hasSideEffects = 1 in {
-def S_BARRIER : SOPP <0x0000000a, (ins), "s_barrier",
+def S_BARRIER : SOPP_Pseudo <"s_barrier", (ins), "",
   [(int_amdgcn_s_barrier)]> {
   let SchedRW = [WriteBarrier];
   let simm16 = 0;
+  let fixed_imm = 1;
   let isConvergent = 1;
 }
 
-def S_WAKEUP : SOPP <0x00000003, (ins), "s_wakeup"> {
+def S_WAKEUP : SOPP_Pseudo <"s_wakeup", (ins) > {
   let SubtargetPredicate = isGFX8Plus;
   let simm16 = 0;
+  let fixed_imm = 1;
   let mayLoad = 1;
   let mayStore = 1;
 }
 
 let mayLoad = 0, mayStore = 0, hasSideEffects = 1 in
-def S_WAITCNT : SOPP <0x0000000c, (ins WAIT_FLAG:$simm16), "s_waitcnt $simm16",
+def S_WAITCNT : SOPP_Pseudo <"s_waitcnt" , (ins WAIT_FLAG:$simm16), "$simm16",
     [(int_amdgcn_s_waitcnt timm:$simm16)]>;
-def S_SETHALT : SOPP <0x0000000d, (ins i16imm:$simm16), "s_sethalt $simm16">;
-def S_SETKILL : SOPP <0x0000000b, (ins i16imm:$simm16), "s_setkill $simm16">;
+def S_SETHALT : SOPP_Pseudo <"s_sethalt" , (ins i16imm:$simm16), "$simm16">;
+def S_SETKILL : SOPP_Pseudo <"s_setkill" , (ins i16imm:$simm16), "$simm16">;
 
 // On SI the documentation says sleep for approximately 64 * low 2
 // bits, consistent with the reported maximum of 448. On VI the
 // maximum reported is 960 cycles, so 960 / 64 = 15 max, so is the
 // maximum really 15 on VI?
-def S_SLEEP : SOPP <0x0000000e, (ins i32imm:$simm16),
-  "s_sleep $simm16", [(int_amdgcn_s_sleep timm:$simm16)]> {
+def S_SLEEP : SOPP_Pseudo <"s_sleep", (ins i32imm:$simm16),
+  "$simm16", [(int_amdgcn_s_sleep timm:$simm16)]> {
   let hasSideEffects = 1;
   let mayLoad = 0;
   let mayStore = 0;
 }
 
-def S_SETPRIO : SOPP <0x0000000f, (ins i16imm:$simm16), "s_setprio $simm16">;
+def S_SETPRIO : SOPP_Pseudo <"s_setprio" , (ins i16imm:$simm16), "$simm16">;
 
 let Uses = [EXEC, M0] in {
 // FIXME: Should this be mayLoad+mayStore?
-def S_SENDMSG : SOPP <0x00000010, (ins SendMsgImm:$simm16), "s_sendmsg $simm16",
+def S_SENDMSG : SOPP_Pseudo <"s_sendmsg" , (ins SendMsgImm:$simm16), "$simm16",
   [(int_amdgcn_s_sendmsg (i32 timm:$simm16), M0)]>;
 
-def S_SENDMSGHALT : SOPP <0x00000011, (ins SendMsgImm:$simm16), "s_sendmsghalt $simm16",
+def S_SENDMSGHALT : SOPP_Pseudo <"s_sendmsghalt" , (ins SendMsgImm:$simm16), "$simm16",
   [(int_amdgcn_s_sendmsghalt (i32 timm:$simm16), M0)]>;
 
 } // End Uses = [EXEC, M0]
 
-def S_TRAP : SOPP <0x00000012, (ins i16imm:$simm16), "s_trap $simm16"> {
+def S_TRAP : SOPP_Pseudo <"s_trap" , (ins i16imm:$simm16), "$simm16"> {
   let isTrap = 1;
 }
 
-def S_ICACHE_INV : SOPP <0x00000013, (ins), "s_icache_inv"> {
+def S_ICACHE_INV : SOPP_Pseudo <"s_icache_inv", (ins)> {
   let simm16 = 0;
+  let fixed_imm = 1;
 }
-def S_INCPERFLEVEL : SOPP <0x00000014, (ins i32imm:$simm16), "s_incperflevel $simm16",
+def S_INCPERFLEVEL : SOPP_Pseudo <"s_incperflevel", (ins i32imm:$simm16), "$simm16",
   [(int_amdgcn_s_incperflevel timm:$simm16)]> {
   let hasSideEffects = 1;
   let mayLoad = 0;
   let mayStore = 0;
 }
-def S_DECPERFLEVEL : SOPP <0x00000015, (ins i32imm:$simm16), "s_decperflevel $simm16",
+def S_DECPERFLEVEL : SOPP_Pseudo <"s_decperflevel", (ins i32imm:$simm16), "$simm16",
   [(int_amdgcn_s_decperflevel timm:$simm16)]> {
   let hasSideEffects = 1;
   let mayLoad = 0;
   let mayStore = 0;
 }
-def S_TTRACEDATA : SOPP <0x00000016, (ins), "s_ttracedata"> {
+def S_TTRACEDATA : SOPP_Pseudo <"s_ttracedata", (ins)> {
   let simm16 = 0;
+  let fixed_imm = 1;
 }
 
 let SubtargetPredicate = HasVGPRIndexMode in {
-def S_SET_GPR_IDX_OFF : SOPP<0x1c, (ins), "s_set_gpr_idx_off"> {
+def S_SET_GPR_IDX_OFF : SOPP_Pseudo<"s_set_gpr_idx_off", (ins) > {
   let simm16 = 0;
+  let fixed_imm = 1;
   let Defs = [MODE];
   let Uses = [MODE];
 }
@@ -1260,8 +1286,9 @@ def S_SET_GPR_IDX_OFF : SOPP<0x1c, (ins), "s_set_gpr_idx_off"> {
 } // End hasSideEffects
 
 let SubtargetPredicate = HasVGPRIndexMode in {
-def S_SET_GPR_IDX_MODE : SOPP<0x1d, (ins GPRIdxMode:$simm16),
-  "s_set_gpr_idx_mode$simm16"> {
+def S_SET_GPR_IDX_MODE : SOPP_Pseudo<"s_set_gpr_idx_mode", (ins GPRIdxMode:$simm16),
+  "$simm16"> {
+  /*"s_set_gpr_idx_mode$simm16"> {*/
   let Defs = [M0, MODE];
   let Uses = [MODE];
 }
@@ -1269,26 +1296,27 @@ def S_SET_GPR_IDX_MODE : SOPP<0x1d, (ins GPRIdxMode:$simm16),
 
 let SubtargetPredicate = isGFX10Plus in {
   def S_INST_PREFETCH :
-    SOPP<0x020, (ins s16imm:$simm16), "s_inst_prefetch $simm16">;
+    SOPP_Pseudo<"s_inst_prefetch", (ins s16imm:$simm16), "$simm16">;
   def S_CLAUSE :
-    SOPP<0x021, (ins s16imm:$simm16), "s_clause $simm16">;
+    SOPP_Pseudo<"s_clause", (ins s16imm:$simm16), "$simm16">;
   def S_WAITCNT_IDLE :
-    SOPP <0x022, (ins), "s_wait_idle"> {
+    SOPP_Pseudo <"s_wait_idle", (ins), ""> {
       let simm16 = 0;
+      let fixed_imm = 1;
     }
   def S_WAITCNT_DEPCTR :
-    SOPP <0x023, (ins s16imm:$simm16), "s_waitcnt_depctr $simm16">;
+    SOPP_Pseudo <"s_waitcnt_depctr" , (ins s16imm:$simm16), "$simm16">;
 
   let hasSideEffects = 0, Uses = [MODE], Defs = [MODE] in {
     def S_ROUND_MODE :
-      SOPP<0x024, (ins s16imm:$simm16), "s_round_mode $simm16">;
+      SOPP_Pseudo<"s_round_mode", (ins s16imm:$simm16), "$simm16">;
     def S_DENORM_MODE :
-      SOPP<0x025, (ins i32imm:$simm16), "s_denorm_mode $simm16",
+      SOPP_Pseudo<"s_denorm_mode", (ins i32imm:$simm16), "$simm16",
       [(SIdenorm_mode (i32 timm:$simm16))]>;
   }
 
   def S_TTRACEDATA_IMM :
-    SOPP<0x028, (ins s16imm:$simm16), "s_ttracedata_imm $simm16">;
+    SOPP_Pseudo<"s_ttracedata_imm", (ins s16imm:$simm16), "$simm16">;
 } // End SubtargetPredicate = isGFX10Plus
 
 //===----------------------------------------------------------------------===//
@@ -1379,15 +1407,25 @@ def : ScalarNot2Pat<S_ORN2_B64, or, v2i32>;
 // Target-specific instruction encodings.
 //===----------------------------------------------------------------------===//
 
-//===----------------------------------------------------------------------===//
-// SOP1 - GFX10.
-//===----------------------------------------------------------------------===//
-
 class Select_gfx10<string opName> : SIMCInstr<opName, SIEncodingFamily.GFX10> {
   Predicate AssemblerPredicate = isGFX10Plus;
   string DecoderNamespace      = "GFX10";
 }
 
+class Select_vi<string opName> : SIMCInstr<opName, SIEncodingFamily.VI> {
+  Predicate AssemblerPredicate = isGFX8GFX9;
+  string DecoderNamespace = "GFX8";
+}
+
+class Select_gfx6_gfx7<string opName> : SIMCInstr<opName, SIEncodingFamily.SI> {
+  Predicate AssemblerPredicate = isGFX6GFX7;
+  string DecoderNamespace      = "GFX6GFX7";
+}
+
+//===----------------------------------------------------------------------===//
+// SOP1 - GFX10.
+//===----------------------------------------------------------------------===//
+
 multiclass SOP1_Real_gfx10<bits<8> op> {
   def _gfx10 : SOP1_Real<op, !cast<SOP1_Pseudo>(NAME)>,
                Select_gfx10<!cast<SOP1_Pseudo>(NAME).Mnemonic>;
@@ -1416,10 +1454,6 @@ defm S_MOVRELSD_2_B32       : SOP1_Real_gfx10<0x049>;
 // SOP1 - GFX6, GFX7.
 //===----------------------------------------------------------------------===//
 
-class Select_gfx6_gfx7<string opName> : SIMCInstr<opName, SIEncodingFamily.SI> {
-  Predicate AssemblerPredicate = isGFX6GFX7;
-  string DecoderNamespace      = "GFX6GFX7";
-}
 
 multiclass SOP1_Real_gfx6_gfx7<bits<8> op> {
   def _gfx6_gfx7 : SOP1_Real<op, !cast<SOP1_Pseudo>(NAME)>,
@@ -1621,15 +1655,163 @@ defm S_SETREG_B32       : SOPK_Real32_gfx6_gfx7_gfx10<0x013>;
 defm S_SETREG_IMM32_B32 : SOPK_Real64_gfx6_gfx7_gfx10<0x015>;
 
 //===----------------------------------------------------------------------===//
-// GFX8 (VI), GFX9.
+// SOPP - GFX6, GFX7, GFX8, GFX9, GFX10
 //===----------------------------------------------------------------------===//
 
-class Select_vi<string opName> :
-  SIMCInstr<opName, SIEncodingFamily.VI> {
-  Predicate AssemblerPredicate = isGFX8GFX9;
-  string DecoderNamespace = "GFX8";
+multiclass SOPP_Real_32_gfx6_gfx7<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic> {
+  def _gfx6_gfx7 : SOPP_Real_32<op, !cast<SOPP_Pseudo>(NAME), real_name>,
+                   Select_gfx6_gfx7<!cast<SOPP_Pseudo>(NAME).Mnemonic>,
+                   SOPPRelaxTable<0, !cast<SOPP_Pseudo>(NAME).KeyName, "_gfx6_gfx7">;
+}
+
+multiclass SOPP_Real_32_gfx8_gfx9<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> {
+  def _vi : SOPP_Real_32<op, !cast<SOPP_Pseudo>(NAME), real_name>,
+            Select_vi<!cast<SOPP_Pseudo>(NAME).Mnemonic>,
+            SOPPRelaxTable<0, !cast<SOPP_Pseudo>(NAME).KeyName, "_vi">;
+}
+
+multiclass SOPP_Real_32_gfx10<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> {
+  def _gfx10 : SOPP_Real_32<op, !cast<SOPP_Pseudo>(NAME), real_name>,
+               Select_gfx10<!cast<SOPP_Pseudo>(NAME).Mnemonic>,
+               SOPPRelaxTable<0, !cast<SOPP_Pseudo>(NAME).KeyName, "_gfx10">;
+}
+
+multiclass SOPP_Real_32_gfx8_gfx9_gfx10<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> :
+  SOPP_Real_32_gfx8_gfx9<op, real_name>, SOPP_Real_32_gfx10<op, real_name>;
+
+multiclass SOPP_Real_32_gfx6_gfx7_gfx8_gfx9<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> :
+  SOPP_Real_32_gfx6_gfx7<op, real_name>, SOPP_Real_32_gfx8_gfx9<op, real_name>;
+
+multiclass SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> :
+  SOPP_Real_32_gfx6_gfx7_gfx8_gfx9<op, real_name>, SOPP_Real_32_gfx10<op, real_name>;
+
+//64 bit encodings, for Relaxation
+multiclass SOPP_Real_64_gfx6_gfx7<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> {
+  def _gfx6_gfx7 : SOPP_Real_64<op, !cast<SOPP_Pseudo>(NAME), real_name>,
+                   Select_gfx6_gfx7<!cast<SOPP_Pseudo>(NAME).Mnemonic>,
+                   SOPPRelaxTable<1, !cast<SOPP_Pseudo>(NAME).KeyName, "_gfx6_gfx7">;
+}
+
+multiclass SOPP_Real_64_gfx8_gfx9<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> {
+  def _vi : SOPP_Real_64<op, !cast<SOPP_Pseudo>(NAME), real_name>,
+            Select_vi<!cast<SOPP_Pseudo>(NAME).Mnemonic>,
+            SOPPRelaxTable<1, !cast<SOPP_Pseudo>(NAME).KeyName, "_vi">;
+}
+
+multiclass SOPP_Real_64_gfx10<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> {
+  def _gfx10 : SOPP_Real_64<op, !cast<SOPP_Pseudo>(NAME), real_name>,
+               Select_gfx10<!cast<SOPP_Pseudo>(NAME).Mnemonic>,
+               SOPPRelaxTable<1, !cast<SOPP_Pseudo>(NAME).KeyName, "_gfx10">;
+}
+
+multiclass SOPP_Real_64_gfx8_gfx9_gfx10<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> :
+  SOPP_Real_64_gfx8_gfx9<op, real_name>, SOPP_Real_64_gfx10<op, real_name>;
+
+multiclass SOPP_Real_64_gfx6_gfx7_gfx8_gfx9<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> :
+  SOPP_Real_64_gfx6_gfx7<op, real_name>, SOPP_Real_64_gfx8_gfx9<op, real_name>;
+
+multiclass SOPP_Real_64_gfx6_gfx7_gfx8_gfx9_gfx10<bits<7> op, string real_name = !cast<SOPP_Pseudo>(NAME).Mnemonic # " "> :
+  SOPP_Real_64_gfx6_gfx7_gfx8_gfx9<op, real_name>, SOPP_Real_64_gfx10<op, real_name>;
+
+//relaxation for insts with no operands not implemented
+multiclass SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<bits<7> op> {
+  defm "" : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<op>;
+  defm _pad_s_nop : SOPP_Real_64_gfx6_gfx7_gfx8_gfx9_gfx10<op>;
+}
+
+defm S_NOP                      : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x000>;
+defm S_ENDPGM                   : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x001, "s_endpgm">;
+defm S_BRANCH                   : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x002>;
+defm S_WAKEUP                   : SOPP_Real_32_gfx8_gfx9_gfx10<0x003>;
+defm S_CBRANCH_SCC0             : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x004>;
+defm S_CBRANCH_SCC1             : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x005>;
+defm S_CBRANCH_VCCZ             : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x006>;
+defm S_CBRANCH_VCCNZ            : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x007>;
+defm S_CBRANCH_EXECZ            : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x008>;
+defm S_CBRANCH_EXECNZ           : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x009>;
+defm S_CBRANCH_CDBGSYS          : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x017>;
+defm S_CBRANCH_CDBGUSER         : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x018>;
+defm S_CBRANCH_CDBGSYS_OR_USER  : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x019>;
+defm S_CBRANCH_CDBGSYS_AND_USER : SOPP_Real_With_Relaxation_gfx6_gfx7_gfx8_gfx9_gfx10<0x01A>;
+defm S_BARRIER                  : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x00a>;
+defm S_WAITCNT                  : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x00c>;
+defm S_SETHALT                  : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x00d>;
+defm S_SETKILL                  : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x00b>;
+defm S_SLEEP                    : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x00e>;
+defm S_SETPRIO                  : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x00f>;
+defm S_SENDMSG                  : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x010>;
+defm S_SENDMSGHALT              : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x011>;
+defm S_TRAP                     : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x012>;
+defm S_ICACHE_INV               : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x013>;
+defm S_INCPERFLEVEL             : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x014>;
+defm S_DECPERFLEVEL             : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x015>;
+defm S_TTRACEDATA               : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x016>;
+defm S_ENDPGM_SAVED             : SOPP_Real_32_gfx6_gfx7_gfx8_gfx9_gfx10<0x01B>;
+defm S_SET_GPR_IDX_OFF          : SOPP_Real_32_gfx8_gfx9<0x01c>;
+defm S_SET_GPR_IDX_MODE         : SOPP_Real_32_gfx8_gfx9<0x01d>;
+defm S_ENDPGM_ORDERED_PS_DONE   : SOPP_Real_32_gfx8_gfx9_gfx10<0x01e>;
+defm S_CODE_END                 : SOPP_Real_32_gfx10<0x01f>;
+defm S_INST_PREFETCH            : SOPP_Real_32_gfx10<0x020>;
+defm S_CLAUSE                   : SOPP_Real_32_gfx10<0x021>;
+defm S_WAITCNT_IDLE             : SOPP_Real_32_gfx10<0x022>;
+defm S_WAITCNT_DEPCTR           : SOPP_Real_32_gfx10<0x023>;
+defm S_ROUND_MODE               : SOPP_Real_32_gfx10<0x024>;
+defm S_DENORM_MODE              : SOPP_Real_32_gfx10<0x025>;
+defm S_TTRACEDATA_IMM           : SOPP_Real_32_gfx10<0x028>;
+
+//===----------------------------------------------------------------------===//
+// SOPC - GFX6, GFX7, GFX8, GFX9, GFX10
+//===----------------------------------------------------------------------===//
+
+multiclass SOPC_Real_gfx6_gfx7<bits<7> op> {
+  def _gfx6_gfx7 : SOPC_Real<op, !cast<SOPC_Pseudo>(NAME)>,
+                   Select_gfx6_gfx7<!cast<SOPC_Pseudo>(NAME).Mnemonic>;
+}
+
+multiclass SOPC_Real_gfx8_gfx9<bits<7> op> {
+  def _vi : SOPC_Real<op, !cast<SOPC_Pseudo>(NAME)>,
+            Select_vi<!cast<SOPC_Pseudo>(NAME).Mnemonic>;
 }
 
+multiclass SOPC_Real_gfx10<bits<7> op> {
+  def _gfx10 : SOPC_Real<op, !cast<SOPC_Pseudo>(NAME)>,
+               Select_gfx10<!cast<SOPC_Pseudo>(NAME).Mnemonic>;
+}
+
+multiclass SOPC_Real_gfx8_gfx9_gfx10<bits<7> op> :
+  SOPC_Real_gfx8_gfx9<op>, SOPC_Real_gfx10<op>;
+
+multiclass SOPC_Real_gfx6_gfx7_gfx8_gfx9<bits<7> op> :
+  SOPC_Real_gfx6_gfx7<op>, SOPC_Real_gfx8_gfx9<op>;
+
+multiclass SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<bits<7> op> :
+  SOPC_Real_gfx6_gfx7_gfx8_gfx9<op>, SOPC_Real_gfx10<op>;
+
+defm S_CMP_EQ_I32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x00>;
+defm S_CMP_LG_I32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x01>;
+defm S_CMP_GT_I32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x02>;
+defm S_CMP_GE_I32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x03>;
+defm S_CMP_LT_I32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x04>;
+defm S_CMP_LE_I32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x05>;
+defm S_CMP_EQ_U32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x06>;
+defm S_CMP_LG_U32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x07>;
+defm S_CMP_GT_U32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x08>;
+defm S_CMP_GE_U32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x09>;
+defm S_CMP_LT_U32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x0a>;
+defm S_CMP_LE_U32     : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x0b>;
+defm S_BITCMP0_B32    : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x0c>;
+defm S_BITCMP1_B32    : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x0d>;
+defm S_BITCMP0_B64    : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x0e>;
+defm S_BITCMP1_B64    : SOPC_Real_gfx6_gfx7_gfx8_gfx9_gfx10<0x0f>;
+defm S_SETVSKIP       : SOPC_Real_gfx6_gfx7_gfx8_gfx9<0x10>;
+defm S_SET_GPR_IDX_ON : SOPC_Real_gfx8_gfx9<0x11>;
+defm S_CMP_EQ_U64     : SOPC_Real_gfx8_gfx9_gfx10<0x12>;
+defm S_CMP_LG_U64     : SOPC_Real_gfx8_gfx9_gfx10<0x13>;
+
+//===----------------------------------------------------------------------===//
+// GFX8 (VI), GFX9.
+//===----------------------------------------------------------------------===//
+
 class SOP1_Real_vi<bits<8> op, SOP1_Pseudo ps> :
   SOP1_Real<op, ps>,
   Select_vi<ps.Mnemonic>;
diff --git a/llvm/lib/Target/AMDGPU/VOP3Instructions.td b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
index 8fea9403cc421..2a0b7466c7f2c 100644
--- a/llvm/lib/Target/AMDGPU/VOP3Instructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3Instructions.td
@@ -138,7 +138,7 @@ class VOP3Inst<string OpName, VOPProfile P, SDPatternOperator node = null_frag,
   let AsmMatchConverter =
     !if(P.HasOpSel,
         "cvtVOP3OpSel",
-        !if(!or(P.HasModifiers, !or(P.HasOMod, P.HasIntClamp)),
+        !if(!or(P.HasModifiers, P.HasOMod, P.HasIntClamp),
             "cvtVOP3",
             ""));
 }
@@ -174,7 +174,7 @@ class VOP3_Profile<VOPProfile P, VOP3Features Features = VOP3_REGULAR> : VOPProf
   let IsMAI    = !if(Features.IsMAI,    1, P.IsMAI);
   let IsPacked = !if(Features.IsPacked, 1, P.IsPacked);
 
-  let HasModifiers = !if(Features.IsPacked, !if(Features.IsMAI, 0, 1), P.HasModifiers);
+  let HasModifiers = !if(Features.IsPacked, !not(Features.IsMAI), P.HasModifiers);
 
   // FIXME: Hack to stop printing _e64
   let Outs64 = (outs DstRC.RegClass:$vdst);
@@ -277,7 +277,7 @@ class getInterp16Ins <bit HasSrc2, bit HasOMod,
 
 class VOP3_INTERP16 <list<ValueType> ArgVT> : VOPProfile<ArgVT> {
 
-  let HasOMod = !if(!eq(DstVT.Value, f16.Value), 0, 1);
+  let HasOMod = !ne(DstVT.Value, f16.Value);
   let HasHigh = 1;
 
   let Outs64 = (outs VGPR_32:$vdst);
diff --git a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
index 393fc8b09d446..159731707eabd 100644
--- a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
@@ -226,7 +226,7 @@ foreach Type = ["I", "U"] in
   foreach Index = 0-3 in {
     // Defines patterns that extract each Index'ed 8bit from an unsigned
     // 32bit scalar value;
-    def Type#Index#"_8bit" : Extract<!shl(Index, 3), 255, !if (!eq (Type, "U"), 1, 0)>;
+    def Type#Index#"_8bit" : Extract<!shl(Index, 3), 255, !eq (Type, "U")>;
 
     // Defines multiplication patterns where the multiplication is happening on each
     // Index'ed 8bit of a 32bit scalar value.
@@ -254,7 +254,7 @@ foreach Type = ["I", "U"] in
   foreach Index = 0-7 in {
     // Defines patterns that extract each Index'ed 4bit from an unsigned
     // 32bit scalar value;
-    def Type#Index#"_4bit" : Extract<!shl(Index, 2), 15, !if (!eq (Type, "U"), 1, 0)>;
+    def Type#Index#"_4bit" : Extract<!shl(Index, 2), 15, !eq (Type, "U")>;
 
     // Defines multiplication patterns where the multiplication is happening on each
     // Index'ed 8bit of a 32bit scalar value.
diff --git a/llvm/lib/Target/AMDGPU/VOPInstructions.td b/llvm/lib/Target/AMDGPU/VOPInstructions.td
index b27a1d31863df..8df1886022182 100644
--- a/llvm/lib/Target/AMDGPU/VOPInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOPInstructions.td
@@ -69,7 +69,7 @@ class VOP3Common <dag outs, dag ins, string asm = "",
   let VOP3 = 1;
 
   let AsmVariantName = AMDGPUAsmVariants.VOP3;
-  let AsmMatchConverter = !if(!eq(HasMods,1), "cvtVOP3", "");
+  let AsmMatchConverter = !if(HasMods, "cvtVOP3", "");
 
   let isCodeGenOnly = 0;
 
@@ -129,7 +129,7 @@ class VOP3_Pseudo <string opName, VOPProfile P, list<dag> pattern = [],
   let AsmMatchConverter =
     !if(isVOP3P,
         "cvtVOP3P",
-        !if(!or(P.HasModifiers, !or(P.HasOMod, P.HasIntClamp)),
+        !if(!or(P.HasModifiers, P.HasOMod, P.HasIntClamp),
             "cvtVOP3",
             ""));
 }
@@ -626,7 +626,7 @@ class VOP_DPP_Pseudo <string OpName, VOPProfile P, list<dag> pattern=[]> :
   string Mnemonic = OpName;
   string AsmOperands = P.AsmDPP;
 
-  let AsmMatchConverter = !if(!eq(P.HasModifiers,1), "cvtDPP", "");
+  let AsmMatchConverter = !if(P.HasModifiers, "cvtDPP", "");
   let SubtargetPredicate = HasDPP;
   let AssemblerPredicate = HasDPP;
   let AsmVariantName = !if(P.HasExtDPP, AMDGPUAsmVariants.DPP,
@@ -681,7 +681,7 @@ class VOP_DPP <string OpName, VOPProfile P, bit IsDPP16,
   let DPP = 1;
   let Size = 8;
 
-  let AsmMatchConverter = !if(!eq(P.HasModifiers,1), "cvtDPP", "");
+  let AsmMatchConverter = !if(P.HasModifiers, "cvtDPP", "");
   let SubtargetPredicate = HasDPP;
   let AssemblerPredicate = HasDPP;
   let AsmVariantName = !if(P.HasExtDPP, AMDGPUAsmVariants.DPP,
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
index d81c8efa1597d..817bc3363a115 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.cpp
@@ -589,8 +589,9 @@ bool ARMBaseInstrInfo::SubsumesPredicate(ArrayRef<MachineOperand> Pred1,
   }
 }
 
-bool ARMBaseInstrInfo::DefinesPredicate(
-    MachineInstr &MI, std::vector<MachineOperand> &Pred) const {
+bool ARMBaseInstrInfo::ClobbersPredicate(MachineInstr &MI,
+                                         std::vector<MachineOperand> &Pred,
+                                         bool SkipDead) const {
   bool Found = false;
   for (unsigned i = 0, e = MI.getNumOperands(); i != e; ++i) {
     const MachineOperand &MO = MI.getOperand(i);
@@ -611,12 +612,6 @@ bool ARMBaseInstrInfo::isCPSRDefined(const MachineInstr &MI) {
   return false;
 }
 
-bool ARMBaseInstrInfo::isAddrMode3OpImm(const MachineInstr &MI,
-                                        unsigned Op) const {
-  const MachineOperand &Offset = MI.getOperand(Op + 1);
-  return Offset.getReg() != 0;
-}
-
 // Load with negative register offset requires additional 1cyc and +I unit
 // for Cortex A57
 bool ARMBaseInstrInfo::isAddrMode3OpMinusReg(const MachineInstr &MI,
@@ -631,13 +626,6 @@ bool ARMBaseInstrInfo::isAddrMode3OpMinusReg(const MachineInstr &MI,
   return (isSub && Offset.getReg() != 0);
 }
 
-bool ARMBaseInstrInfo::isLdstScaledReg(const MachineInstr &MI,
-                                       unsigned Op) const {
-  const MachineOperand &Opc = MI.getOperand(Op + 2);
-  unsigned OffImm = Opc.getImm();
-  return ARM_AM::getAM2ShiftOpc(OffImm) != ARM_AM::no_shift;
-}
-
 // Load, scaled register offset, not plus LSL2
 bool ARMBaseInstrInfo::isLdstScaledRegNotPlusLsl2(const MachineInstr &MI,
                                                   unsigned Op) const {
diff --git a/llvm/lib/Target/ARM/ARMBaseInstrInfo.h b/llvm/lib/Target/ARM/ARMBaseInstrInfo.h
index 5bf6e880056de..a03d57c97d5e2 100644
--- a/llvm/lib/Target/ARM/ARMBaseInstrInfo.h
+++ b/llvm/lib/Target/ARM/ARMBaseInstrInfo.h
@@ -171,18 +171,15 @@ class ARMBaseInstrInfo : public ARMGenInstrInfo {
   bool SubsumesPredicate(ArrayRef<MachineOperand> Pred1,
                          ArrayRef<MachineOperand> Pred2) const override;
 
-  bool DefinesPredicate(MachineInstr &MI,
-                        std::vector<MachineOperand> &Pred) const override;
+  bool ClobbersPredicate(MachineInstr &MI, std::vector<MachineOperand> &Pred,
+                         bool SkipDead) const override;
 
   bool isPredicable(const MachineInstr &MI) const override;
 
   // CPSR defined in instruction
   static bool isCPSRDefined(const MachineInstr &MI);
-  bool isAddrMode3OpImm(const MachineInstr &MI, unsigned Op) const;
   bool isAddrMode3OpMinusReg(const MachineInstr &MI, unsigned Op) const;
 
-  // Load, scaled register offset
-  bool isLdstScaledReg(const MachineInstr &MI, unsigned Op) const;
   // Load, scaled register offset, not plus LSL2
   bool isLdstScaledRegNotPlusLsl2(const MachineInstr &MI, unsigned Op) const;
   // Minus reg for ldstso addr mode
diff --git a/llvm/lib/Target/ARM/ARMInstrMVE.td b/llvm/lib/Target/ARM/ARMInstrMVE.td
index f7f403503dc7f..67a3530a7d4e7 100644
--- a/llvm/lib/Target/ARM/ARMInstrMVE.td
+++ b/llvm/lib/Target/ARM/ARMInstrMVE.td
@@ -2474,7 +2474,7 @@ multiclass MVE_VABSNEG_int_m<string iname, bit negate, bit saturate,
 
   let Predicates = [HasMVEInt] in {
     // VQABS and VQNEG have more difficult isel patterns defined elsewhere
-    if !eq(saturate, 0) then {
+    if !not(saturate) then {
       def : Pat<(VTI.Vec (unpred_op (VTI.Vec MQPR:$v))),
                 (VTI.Vec (Inst $v))>;
     }
@@ -4777,7 +4777,7 @@ class MVE_VxMOVxN<string iname, string suffix, bit bit_28, bit bit_17,
   let Inst{16} = 0b1;
   let Inst{12} = T;
   let Inst{8} = 0b0;
-  let Inst{7} = !if(!eq(bit_17, 0), 1, 0);
+  let Inst{7} = !not(bit_17);
   let Inst{0} = 0b1;
   let validForTailPredication = 1;
   let retainsPreviousHalfElement = 1;
@@ -4808,7 +4808,7 @@ multiclass MVE_VMOVN_p<Instruction Inst, bit top,
                                (VTI.Vec MQPR:$Qm), (i32 top))),
             (VTI.Vec (Inst (VTI.Vec MQPR:$Qd_src), (VTI.Vec MQPR:$Qm)))>;
 
-  if !eq(top, 0) then {
+  if !not(top) then {
     // If we see MVEvmovn(a,ARMvrev(b),1), that wants to overwrite the odd
     // lanes of a with the odd lanes of b. In other words, the lanes we're
     // _keeping_ from a are the even ones. So we can flip it round and say that
diff --git a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
index 47f11df7c59e8..d0e362f8adaae 100644
--- a/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
+++ b/llvm/lib/Target/ARM/ARMLowOverheadLoops.cpp
@@ -537,7 +537,8 @@ static bool TryRemove(MachineInstr *MI, ReachingDefAnalysis &RDA,
       for (auto &IT : *MBB) {
         if (IT.getOpcode() != ARM::t2IT)
           continue;
-        RDA.getReachingLocalUses(&IT, ARM::ITSTATE, ITBlocks[&IT]);
+        RDA.getReachingLocalUses(&IT, MCRegister::from(ARM::ITSTATE),
+                                 ITBlocks[&IT]);
       }
     }
 
@@ -620,7 +621,7 @@ bool LowOverheadLoop::ValidateTailPredicate() {
   if (StartInsertPt == Start && Start->getOperand(0).getReg() == ARM::LR) {
     if (auto *IterCount = RDA.getMIOperand(Start, 0)) {
       SmallPtrSet<MachineInstr *, 2> Uses;
-      RDA.getGlobalUses(IterCount, ARM::LR, Uses);
+      RDA.getGlobalUses(IterCount, MCRegister::from(ARM::LR), Uses);
       for (auto *Use : Uses) {
         if (Use != Start && Use != Dec) {
           LLVM_DEBUG(dbgs() << " ARM Loops: Found LR use: " << *Use);
@@ -636,7 +637,7 @@ bool LowOverheadLoop::ValidateTailPredicate() {
   // we can use this register at InsertPt.
   MachineInstr *VCTP = VCTPs.back();
   TPNumElements = VCTP->getOperand(1);
-  Register NumElements = TPNumElements.getReg();
+  MCRegister NumElements = TPNumElements.getReg().asMCReg();
 
   // If the register is defined within loop, then we can't perform TP.
   // TODO: Check whether this is just a mov of a register that would be
@@ -669,8 +670,9 @@ bool LowOverheadLoop::ValidateTailPredicate() {
         // insertion point
         MachineOperand Operand = ElemDef->getOperand(1);
         if (isMovRegOpcode(ElemDef->getOpcode()) &&
-            RDA.getUniqueReachingMIDef(ElemDef, Operand.getReg()) ==
-               RDA.getUniqueReachingMIDef(&*StartInsertPt, Operand.getReg())) {
+            RDA.getUniqueReachingMIDef(ElemDef, Operand.getReg().asMCReg()) ==
+                RDA.getUniqueReachingMIDef(&*StartInsertPt,
+                                           Operand.getReg().asMCReg())) {
           TPNumElements = Operand;
           NumElements = TPNumElements.getReg();
         } else {
@@ -711,7 +713,7 @@ bool LowOverheadLoop::ValidateTailPredicate() {
   // preheader, so we need to check that the register isn't redefined
   // before entering the loop.
   auto CannotProvideElements = [this](MachineBasicBlock *MBB,
-                                      Register NumElements) {
+                                      MCRegister NumElements) {
     if (MBB->empty())
       return false;
     // NumElements is redefined in this block.
@@ -748,8 +750,8 @@ bool LowOverheadLoop::ValidateTailPredicate() {
   // tail predicated loop. Explicitly refer to the vctp operand no matter which
   // register NumElements has been assigned to, since that is what the
   // modifications will be using
-  if (auto *Def = RDA.getUniqueReachingMIDef(&MBB->back(),
-                                             VCTP->getOperand(1).getReg())) {
+  if (auto *Def = RDA.getUniqueReachingMIDef(
+          &MBB->back(), VCTP->getOperand(1).getReg().asMCReg())) {
     SmallPtrSet<MachineInstr*, 2> ElementChain;
     SmallPtrSet<MachineInstr*, 2> Ignore;
     unsigned ExpectedVectorWidth = getTailPredVectorWidth(VCTP->getOpcode());
@@ -942,7 +944,7 @@ bool LowOverheadLoop::ValidateLiveOuts() {
   auto HasPredicatedUsers = [this](MachineInstr *MI, const MachineOperand &MO,
                               SmallPtrSetImpl<MachineInstr *> &Predicated) {
     SmallPtrSet<MachineInstr *, 2> Uses;
-    RDA.getGlobalUses(MI, MO.getReg(), Uses);
+    RDA.getGlobalUses(MI, MO.getReg().asMCReg(), Uses);
     for (auto *Use : Uses) {
       if (Use != MI && !Predicated.count(Use))
         return false;
@@ -1058,7 +1060,7 @@ void LowOverheadLoop::Validate(ARMBasicBlockUtils *BBUtils) {
       return true;
     }
 
-    unsigned CountReg = Start->getOperand(0).getReg();
+    Register CountReg = Start->getOperand(0).getReg();
     auto IsMoveLR = [&CountReg](MachineInstr *MI) {
       return MI->getOpcode() == ARM::tMOVr &&
              MI->getOperand(0).getReg() == ARM::LR &&
@@ -1069,8 +1071,10 @@ void LowOverheadLoop::Validate(ARMBasicBlockUtils *BBUtils) {
     // Find an insertion point:
     // - Is there a (mov lr, Count) before Start? If so, and nothing else
     //   writes to Count before Start, we can insert at start.
-    if (auto *LRDef = RDA.getUniqueReachingMIDef(Start, ARM::LR)) {
-      if (IsMoveLR(LRDef) && RDA.hasSameReachingDef(Start, LRDef, CountReg)) {
+    if (auto *LRDef =
+            RDA.getUniqueReachingMIDef(Start, MCRegister::from(ARM::LR))) {
+      if (IsMoveLR(LRDef) &&
+          RDA.hasSameReachingDef(Start, LRDef, CountReg.asMCReg())) {
         SmallPtrSet<MachineInstr *, 2> Ignore = { Dec };
         if (!TryRemove(LRDef, RDA, ToRemove, Ignore))
           return false;
@@ -1084,7 +1088,8 @@ void LowOverheadLoop::Validate(ARMBasicBlockUtils *BBUtils) {
     //   to Count after Start, we can insert at that mov (which will now be
     //   dead).
     MachineBasicBlock *MBB = Start->getParent();
-    if (auto *LRDef = RDA.getLocalLiveOutMIDef(MBB, ARM::LR)) {
+    if (auto *LRDef =
+            RDA.getLocalLiveOutMIDef(MBB, MCRegister::from(ARM::LR))) {
       if (IsMoveLR(LRDef) && RDA.hasSameReachingDef(Start, LRDef, CountReg)) {
         SmallPtrSet<MachineInstr *, 2> Ignore = { Start, Dec };
         if (!TryRemove(LRDef, RDA, ToRemove, Ignore))
@@ -1097,7 +1102,7 @@ void LowOverheadLoop::Validate(ARMBasicBlockUtils *BBUtils) {
 
     // We've found no suitable LR def and Start doesn't use LR directly. Can we
     // just define LR anyway?
-    if (!RDA.isSafeToDefRegAt(Start, ARM::LR))
+    if (!RDA.isSafeToDefRegAt(Start, MCRegister::from(ARM::LR)))
       return false;
 
     InsertPt = MachineBasicBlock::iterator(Start);
@@ -1133,7 +1138,7 @@ bool LowOverheadLoop::AddVCTP(MachineInstr *MI) {
   // If it does, store it in the VCTPs set, else refuse it.
   MachineInstr *Prev = VCTPs.back();
   if (!Prev->getOperand(1).isIdenticalTo(MI->getOperand(1)) ||
-      !RDA.hasSameReachingDef(Prev, MI, MI->getOperand(1).getReg())) {
+      !RDA.hasSameReachingDef(Prev, MI, MI->getOperand(1).getReg().asMCReg())) {
     LLVM_DEBUG(dbgs() << "ARM Loops: Found VCTP with a different reaching "
                          "definition from the main VCTP");
     return false;
@@ -1324,7 +1329,7 @@ bool ARMLowOverheadLoops::ProcessLoop(MachineLoop *ML) {
   // Check that the only instruction using LoopDec is LoopEnd.
   // TODO: Check for copy chains that really have no effect.
   SmallPtrSet<MachineInstr*, 2> Uses;
-  RDA->getReachingLocalUses(LoLoop.Dec, ARM::LR, Uses);
+  RDA->getReachingLocalUses(LoLoop.Dec, MCRegister::from(ARM::LR), Uses);
   if (Uses.size() > 1 || !Uses.count(LoLoop.End)) {
     LLVM_DEBUG(dbgs() << "ARM Loops: Unable to remove LoopDec.\n");
     LoLoop.Revert = true;
@@ -1371,7 +1376,8 @@ bool ARMLowOverheadLoops::RevertLoopDec(MachineInstr *MI) const {
   }
 
   // If nothing defines CPSR between LoopDec and LoopEnd, use a t2SUBS.
-  bool SetFlags = RDA->isSafeToDefRegAt(MI, ARM::CPSR, Ignore);
+  bool SetFlags =
+      RDA->isSafeToDefRegAt(MI, MCRegister::from(ARM::CPSR), Ignore);
 
   MachineInstrBuilder MIB = BuildMI(*MBB, MI, MI->getDebugLoc(),
                                     TII->get(ARM::t2SUBri));
diff --git a/llvm/lib/Target/ARM/ARMSchedule.td b/llvm/lib/Target/ARM/ARMSchedule.td
index 8cdf12a625ce5..23cbf6b6c0be9 100644
--- a/llvm/lib/Target/ARM/ARMSchedule.td
+++ b/llvm/lib/Target/ARM/ARMSchedule.td
@@ -164,6 +164,10 @@ def IsCPSRDefined : CheckFunctionPredicateWithTII<
 
 def IsCPSRDefinedPred : MCSchedPredicate<IsCPSRDefined>;
 
+let FunctionMapper = "ARM_AM::getAM2ShiftOpc" in {
+  def CheckExtNoShift : CheckImmOperand_s<4, "ARM_AM::no_shift">;
+}
+
 //===----------------------------------------------------------------------===//
 // Instruction Itinerary classes used for ARM
 //
diff --git a/llvm/lib/Target/ARM/ARMScheduleA57.td b/llvm/lib/Target/ARM/ARMScheduleA57.td
index 3663b9d7d653a..99a31055645a5 100644
--- a/llvm/lib/Target/ARM/ARMScheduleA57.td
+++ b/llvm/lib/Target/ARM/ARMScheduleA57.td
@@ -28,14 +28,9 @@ def IsCPSRDefinedAndPredicatedPred :
 // Cortex A57 rev. r1p0 or later (false = r0px)
 def IsR1P0AndLaterPred : SchedPredicate<[{false}]>;
 
-// If Addrmode3 contains register offset (not immediate)
-def IsLdrAm3RegOffPred :
-  SchedPredicate<[{!TII->isAddrMode3OpImm(*MI, 1)}]>;
-// The same predicate with operand offset 2 and 3:
-def IsLdrAm3RegOffPredX2 :
-  SchedPredicate<[{!TII->isAddrMode3OpImm(*MI, 2)}]>;
-def IsLdrAm3RegOffPredX3 :
-  SchedPredicate<[{!TII->isAddrMode3OpImm(*MI, 3)}]>;
+def IsLdrAm3RegOffPred : MCSchedPredicate<CheckInvalidRegOperand<2>>;
+def IsLdrAm3RegOffPredX2 : MCSchedPredicate<CheckInvalidRegOperand<3>>;
+def IsLdrAm3RegOffPredX3 : MCSchedPredicate<CheckInvalidRegOperand<4>>;
 
 // If Addrmode3 contains "minus register"
 def IsLdrAm3NegRegOffPred :
@@ -54,11 +49,7 @@ def IsLdstsoScaledNotOptimalPred :
 def IsLdstsoScaledNotOptimalPredX2 :
   SchedPredicate<[{TII->isLdstScaledRegNotPlusLsl2(*MI, 2)}]>;
 
-// Load, scaled register offset
-def IsLdstsoScaledPred :
-  SchedPredicate<[{TII->isLdstScaledReg(*MI, 1)}]>;
-def IsLdstsoScaledPredX2 :
-  SchedPredicate<[{TII->isLdstScaledReg(*MI, 2)}]>;
+def IsLdstsoScaledPredX2 : MCSchedPredicate<CheckNot<CheckExtNoShift>>;
 
 def IsLdstsoMinusRegPredX0 :
   SchedPredicate<[{TII->isLdstSoMinusReg(*MI, 0)}]>;
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
index 293c2d639f887..774f2507b8d2d 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMMCTargetDesc.cpp
@@ -11,6 +11,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "ARMMCTargetDesc.h"
+#include "ARMAddressingModes.h"
 #include "ARMBaseInfo.h"
 #include "ARMInstPrinter.h"
 #include "ARMMCAsmInfo.h"
diff --git a/llvm/lib/Target/Hexagon/HexagonInstrInfo.cpp b/llvm/lib/Target/Hexagon/HexagonInstrInfo.cpp
index 93215a4b61870..26fc093d15a76 100644
--- a/llvm/lib/Target/Hexagon/HexagonInstrInfo.cpp
+++ b/llvm/lib/Target/Hexagon/HexagonInstrInfo.cpp
@@ -1639,8 +1639,9 @@ bool HexagonInstrInfo::SubsumesPredicate(ArrayRef<MachineOperand> Pred1,
   return false;
 }
 
-bool HexagonInstrInfo::DefinesPredicate(MachineInstr &MI,
-      std::vector<MachineOperand> &Pred) const {
+bool HexagonInstrInfo::ClobbersPredicate(MachineInstr &MI,
+                                         std::vector<MachineOperand> &Pred,
+                                         bool SkipDead) const {
   const HexagonRegisterInfo &HRI = *Subtarget.getRegisterInfo();
 
   for (unsigned oper = 0; oper < MI.getNumOperands(); ++oper) {
diff --git a/llvm/lib/Target/Hexagon/HexagonInstrInfo.h b/llvm/lib/Target/Hexagon/HexagonInstrInfo.h
index 847b9a6728916..11717996935d2 100644
--- a/llvm/lib/Target/Hexagon/HexagonInstrInfo.h
+++ b/llvm/lib/Target/Hexagon/HexagonInstrInfo.h
@@ -238,8 +238,8 @@ class HexagonInstrInfo : public HexagonGenInstrInfo {
   /// If the specified instruction defines any predicate
   /// or condition code register(s) used for predication, returns true as well
   /// as the definition predicate(s) by reference.
-  bool DefinesPredicate(MachineInstr &MI,
-                        std::vector<MachineOperand> &Pred) const override;
+  bool ClobbersPredicate(MachineInstr &MI, std::vector<MachineOperand> &Pred,
+                         bool SkipDead) const override;
 
   /// Return true if the specified instruction can be predicated.
   /// By default, this returns true for every instruction with a
diff --git a/llvm/lib/Target/Lanai/LanaiISelLowering.cpp b/llvm/lib/Target/Lanai/LanaiISelLowering.cpp
index 32ccf71725943..9e633cd91729e 100644
--- a/llvm/lib/Target/Lanai/LanaiISelLowering.cpp
+++ b/llvm/lib/Target/Lanai/LanaiISelLowering.cpp
@@ -1020,7 +1020,7 @@ SDValue LanaiTargetLowering::LowerDYNAMIC_STACKALLOC(SDValue Op,
   SDValue Size = Op.getOperand(1);
   SDLoc DL(Op);
 
-  unsigned SPReg = getStackPointerRegisterToSaveRestore();
+  Register SPReg = getStackPointerRegisterToSaveRestore();
 
   // Get a reference to the stack pointer.
   SDValue StackPointer = DAG.getCopyFromReg(Chain, DL, SPReg, MVT::i32);
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp b/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
index fd83b5b6d4b8c..487bf925ac999 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.cpp
@@ -1802,8 +1802,9 @@ bool PPCInstrInfo::SubsumesPredicate(ArrayRef<MachineOperand> Pred1,
   return false;
 }
 
-bool PPCInstrInfo::DefinesPredicate(MachineInstr &MI,
-                                    std::vector<MachineOperand> &Pred) const {
+bool PPCInstrInfo::ClobbersPredicate(MachineInstr &MI,
+                                     std::vector<MachineOperand> &Pred,
+                                     bool SkipDead) const {
   // Note: At the present time, the contents of Pred from this function is
   // unused by IfConversion. This implementation follows ARM by pushing the
   // CR-defining operand. Because the 'DZ' and 'DNZ' count as types of
diff --git a/llvm/lib/Target/PowerPC/PPCInstrInfo.h b/llvm/lib/Target/PowerPC/PPCInstrInfo.h
index e3e87022e97c3..79b7bd4a1d38b 100644
--- a/llvm/lib/Target/PowerPC/PPCInstrInfo.h
+++ b/llvm/lib/Target/PowerPC/PPCInstrInfo.h
@@ -472,8 +472,8 @@ class PPCInstrInfo : public PPCGenInstrInfo {
   bool SubsumesPredicate(ArrayRef<MachineOperand> Pred1,
                          ArrayRef<MachineOperand> Pred2) const override;
 
-  bool DefinesPredicate(MachineInstr &MI,
-                        std::vector<MachineOperand> &Pred) const override;
+  bool ClobbersPredicate(MachineInstr &MI, std::vector<MachineOperand> &Pred,
+                         bool SkipDead) const override;
 
   // Comparison optimization.
 
diff --git a/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp b/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp
index 985722fdcab4a..3e114e17f85c1 100644
--- a/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZFrameLowering.cpp
@@ -316,6 +316,8 @@ void SystemZFrameLowering::
 processFunctionBeforeFrameFinalized(MachineFunction &MF,
                                     RegScavenger *RS) const {
   MachineFrameInfo &MFFrame = MF.getFrameInfo();
+  SystemZMachineFunctionInfo *ZFI = MF.getInfo<SystemZMachineFunctionInfo>();
+  MachineRegisterInfo *MRI = &MF.getRegInfo();
   bool BackChain = MF.getFunction().hasFnAttribute("backchain");
 
   if (!usePackedStack(MF) || BackChain)
@@ -344,6 +346,14 @@ processFunctionBeforeFrameFinalized(MachineFunction &MF,
     RS->addScavengingFrameIndex(MFFrame.CreateStackObject(8, Align(8), false));
     RS->addScavengingFrameIndex(MFFrame.CreateStackObject(8, Align(8), false));
   }
+
+  // If R6 is used as an argument register it is still callee saved. If it in
+  // this case is not clobbered (and restored) it should never be marked as
+  // killed.
+  if (MF.front().isLiveIn(SystemZ::R6D) &&
+      ZFI->getRestoreGPRRegs().LowGPR != SystemZ::R6D)
+    for (auto &MO : MRI->use_nodbg_operands(SystemZ::R6D))
+      MO.setIsKill(false);
 }
 
 // Emit instructions before MBBI (in MBB) to add NumBytes to Reg.
diff --git a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
index 9deb66bd067f0..62c734ef39cb1 100644
--- a/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
+++ b/llvm/lib/Target/SystemZ/SystemZISelLowering.cpp
@@ -3423,7 +3423,7 @@ lowerDYNAMIC_STACKALLOC(SDValue Op, SelectionDAG &DAG) const {
   uint64_t RequiredAlign = std::max(AlignVal, StackAlign);
   uint64_t ExtraAlignSpace = RequiredAlign - StackAlign;
 
-  unsigned SPReg = getStackPointerRegisterToSaveRestore();
+  Register SPReg = getStackPointerRegisterToSaveRestore();
   SDValue NeededSpace = Size;
 
   // Get a reference to the stack pointer.
diff --git a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
index a077179cfd817..f13f9cb76437a 100644
--- a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -164,7 +164,7 @@ class X86AsmParser : public MCTargetAsmParser {
     SmallVector<InfixCalculatorTok, 4> InfixOperatorStack;
     SmallVector<ICToken, 4> PostfixStack;
 
-    bool isUnaryOperator(const InfixCalculatorTok Op) {
+    bool isUnaryOperator(InfixCalculatorTok Op) const {
       return Op == IC_NEG || Op == IC_NOT;
     }
 
@@ -395,25 +395,25 @@ class X86AsmParser : public MCTargetAsmParser {
           MemExpr(false), OffsetOperator(false) {}
 
     void addImm(int64_t imm) { Imm += imm; }
-    short getBracCount() { return BracCount; }
-    bool isMemExpr() { return MemExpr; }
-    bool isOffsetOperator() { return OffsetOperator; }
-    SMLoc getOffsetLoc() { return OffsetOperatorLoc; }
-    unsigned getBaseReg() { return BaseReg; }
-    unsigned getIndexReg() { return IndexReg; }
-    unsigned getScale() { return Scale; }
-    const MCExpr *getSym() { return Sym; }
-    StringRef getSymName() { return SymName; }
-    StringRef getType() { return CurType.Name; }
-    unsigned getSize() { return CurType.Size; }
-    unsigned getElementSize() { return CurType.ElementSize; }
-    unsigned getLength() { return CurType.Length; }
+    short getBracCount() const { return BracCount; }
+    bool isMemExpr() const { return MemExpr; }
+    bool isOffsetOperator() const { return OffsetOperator; }
+    SMLoc getOffsetLoc() const { return OffsetOperatorLoc; }
+    unsigned getBaseReg() const { return BaseReg; }
+    unsigned getIndexReg() const { return IndexReg; }
+    unsigned getScale() const { return Scale; }
+    const MCExpr *getSym() const { return Sym; }
+    StringRef getSymName() const { return SymName; }
+    StringRef getType() const { return CurType.Name; }
+    unsigned getSize() const { return CurType.Size; }
+    unsigned getElementSize() const { return CurType.ElementSize; }
+    unsigned getLength() const { return CurType.Length; }
     int64_t getImm() { return Imm + IC.execute(); }
-    bool isValidEndState() {
+    bool isValidEndState() const {
       return State == IES_RBRAC || State == IES_INTEGER;
     }
-    bool hadError() { return State == IES_ERROR; }
-    InlineAsmIdentifierInfo &getIdentifierInfo() { return Info; }
+    bool hadError() const { return State == IES_ERROR; }
+    const InlineAsmIdentifierInfo &getIdentifierInfo() const { return Info; }
 
     void onOr() {
       IntelExprState CurrState = State;
@@ -1210,8 +1210,6 @@ bool X86AsmParser::MatchRegisterByName(unsigned &RegNo, StringRef RegName,
     // FIXME: This should be done using Requires<Not64BitMode> and
     // Requires<In64BitMode> so "eiz" usage in 64-bit instructions can be also
     // checked.
-    // FIXME: Check AH, CH, DH, BH cannot be used in an instruction requiring a
-    // REX prefix.
     if (RegNo == X86::RIZ || RegNo == X86::RIP ||
         X86MCRegisterClasses[X86::GR64RegClassID].contains(RegNo) ||
         X86II::isX86_64NonExtLowByteReg(RegNo) ||
@@ -2315,7 +2313,7 @@ bool X86AsmParser::ParseIntelOperand(OperandVector &Operands) {
   // and we are parsing a segment override
   if (!SM.isMemExpr() && !RegNo) {
     if (isParsingMSInlineAsm() && SM.isOffsetOperator()) {
-      const InlineAsmIdentifierInfo Info = SM.getIdentifierInfo();
+      const InlineAsmIdentifierInfo &Info = SM.getIdentifierInfo();
       if (Info.isKind(InlineAsmIdentifierInfo::IK_Var)) {
         // Disp includes the address of a variable; make sure this is recorded
         // for later handling.
@@ -3619,6 +3617,33 @@ bool X86AsmParser::validateInstruction(MCInst &Inst, const OperandVector &Ops) {
   }
   }
 
+  const MCInstrDesc &MCID = MII.get(Inst.getOpcode());
+  // Check that we aren't mixing AH/BH/CH/DH with REX prefix. We only need to
+  // check this with the legacy encoding, VEX/EVEX/XOP don't use REX.
+  if ((MCID.TSFlags & X86II::EncodingMask) == 0) {
+    MCPhysReg HReg = X86::NoRegister;
+    bool UsesRex = MCID.TSFlags & X86II::REX_W;
+    unsigned NumOps = Inst.getNumOperands();
+    for (unsigned i = 0; i != NumOps; ++i) {
+      const MCOperand &MO = Inst.getOperand(i);
+      if (!MO.isReg())
+        continue;
+      unsigned Reg = MO.getReg();
+      if (Reg == X86::AH || Reg == X86::BH || Reg == X86::CH || Reg == X86::DH)
+        HReg = Reg;
+      if (X86II::isX86_64NonExtLowByteReg(Reg) ||
+          X86II::isX86_64ExtendedReg(Reg))
+        UsesRex = true;
+    }
+
+    if (UsesRex && HReg != X86::NoRegister) {
+      StringRef RegName = X86IntelInstPrinter::getRegisterName(HReg);
+      return Error(Ops[0]->getStartLoc(),
+                   "can't encode '" + RegName + "' in an instruction requiring "
+                   "REX prefix");
+    }
+  }
+
   return false;
 }
 
@@ -3989,6 +4014,8 @@ bool X86AsmParser::MatchAndEmitATTInstruction(SMLoc IDLoc, unsigned &Opcode,
   unsigned NumSuccessfulMatches =
       std::count(std::begin(Match), std::end(Match), Match_Success);
   if (NumSuccessfulMatches == 1) {
+    if (!MatchingInlineAsm && validateInstruction(Inst, Operands))
+      return true;
     // Some instructions need post-processing to, for example, tweak which
     // encoding is selected. Loop on it while changes happen so the
     // individual transformations can chain off each other.
diff --git a/llvm/lib/Target/X86/X86.td b/llvm/lib/Target/X86/X86.td
index 428fe78b8213b..3c5dd73005521 100644
--- a/llvm/lib/Target/X86/X86.td
+++ b/llvm/lib/Target/X86/X86.td
@@ -291,6 +291,8 @@ def FeatureSERIALIZE : SubtargetFeature<"serialize", "HasSERIALIZE", "true",
                                         "Has serialize instruction">;
 def FeatureTSXLDTRK : SubtargetFeature<"tsxldtrk", "HasTSXLDTRK", "true",
                                        "Support TSXLDTRK instructions">;
+def FeatureUINTR : SubtargetFeature<"uintr", "HasUINTR", "true",
+                                    "Has UINTR Instructions">;
 // On some processors, instructions that implicitly take two memory operands are
 // slow. In practice, this means that CALL, PUSH, and POP with memory operands
 // should be avoided in favor of a MOV + register CALL/PUSH/POP.
@@ -772,7 +774,8 @@ def ProcessorFeatures {
                                                   FeatureSHSTK,
                                                   FeatureVP2INTERSECT,
                                                   FeatureMOVDIRI,
-                                                  FeatureMOVDIR64B];
+                                                  FeatureMOVDIR64B,
+                                                  FeatureUINTR];
   list<SubtargetFeature> SPRTuning = ICXTuning;
   list<SubtargetFeature> SPRFeatures =
     !listconcat(ICXFeatures, SPRAdditionalFeatures);
diff --git a/llvm/lib/Target/X86/X86FixupLEAs.cpp b/llvm/lib/Target/X86/X86FixupLEAs.cpp
index ca9d8a1a4ffda..0054d5818a964 100644
--- a/llvm/lib/Target/X86/X86FixupLEAs.cpp
+++ b/llvm/lib/Target/X86/X86FixupLEAs.cpp
@@ -450,6 +450,7 @@ bool FixupLEAPass::optTwoAddrLEA(MachineBasicBlock::iterator &I,
   } else
     return false;
 
+  MBB.getParent()->substituteDebugValuesForInst(*I, *NewMI, 1);
   MBB.erase(I);
   I = NewMI;
   return true;
@@ -485,6 +486,7 @@ void FixupLEAPass::seekLEAFixup(MachineOperand &p,
       LLVM_DEBUG(dbgs() << "FixLEA: Candidate to replace:"; MBI->dump(););
       // now to replace with an equivalent LEA...
       LLVM_DEBUG(dbgs() << "FixLEA: Replaced by: "; NewMI->dump(););
+      MBB.getParent()->substituteDebugValuesForInst(*MBI, *NewMI, 1);
       MBB.erase(MBI);
       MachineBasicBlock::iterator J =
           static_cast<MachineBasicBlock::iterator>(NewMI);
@@ -538,6 +540,7 @@ void FixupLEAPass::processInstructionForSlowLEA(MachineBasicBlock::iterator &I,
     LLVM_DEBUG(NewMI->dump(););
   }
   if (NewMI) {
+    MBB.getParent()->substituteDebugValuesForInst(*I, *NewMI, 1);
     MBB.erase(I);
     I = NewMI;
   }
@@ -644,6 +647,7 @@ void FixupLEAPass::processInstrForSlow3OpLEA(MachineBasicBlock::iterator &I,
       }
     }
 
+    MBB.getParent()->substituteDebugValuesForInst(*I, *NewMI, 1);
     MBB.erase(I);
     I = NewMI;
     return;
@@ -669,6 +673,7 @@ void FixupLEAPass::processInstrForSlow3OpLEA(MachineBasicBlock::iterator &I,
                 .add(Index);
     LLVM_DEBUG(NewMI->dump(););
 
+    MBB.getParent()->substituteDebugValuesForInst(*I, *NewMI, 1);
     MBB.erase(I);
     I = NewMI;
     return;
@@ -691,6 +696,7 @@ void FixupLEAPass::processInstrForSlow3OpLEA(MachineBasicBlock::iterator &I,
               .add(Base);
   LLVM_DEBUG(NewMI->dump(););
 
+  MBB.getParent()->substituteDebugValuesForInst(*I, *NewMI, 1);
   MBB.erase(I);
   I = NewMI;
 }
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index f4c4425746d0a..913073e662219 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -2469,13 +2469,23 @@ Value *X86TargetLowering::getIRStackGuard(IRBuilder<> &IRB) const {
       // <zircon/tls.h> defines ZX_TLS_STACK_GUARD_OFFSET with this value.
       return SegmentOffset(IRB, 0x10, getAddressSpace());
     } else {
+      unsigned AddressSpace = getAddressSpace();
+      // Specially, some users may customize the base reg and offset.
+      unsigned Offset = getTargetMachine().Options.StackProtectorGuardOffset;
+      // If we don't set -stack-protector-guard-offset value:
       // %fs:0x28, unless we're using a Kernel code model, in which case
       // it's %gs:0x28.  gs:0x14 on i386.
-      unsigned Offset = (Subtarget.is64Bit()) ? 0x28 : 0x14;
-      return SegmentOffset(IRB, Offset, getAddressSpace());
+      if (Offset == (unsigned)-1)
+        Offset = (Subtarget.is64Bit()) ? 0x28 : 0x14;
+
+      auto GuardReg = getTargetMachine().Options.StackProtectorGuardReg;
+        if (GuardReg == "fs")
+          AddressSpace = X86AS::FS;
+        else if (GuardReg == "gs")
+          AddressSpace = X86AS::GS;
+      return SegmentOffset(IRB, Offset, AddressSpace);
     }
   }
-
   return TargetLowering::getIRStackGuard(IRB);
 }
 
@@ -2497,8 +2507,13 @@ void X86TargetLowering::insertSSPDeclarations(Module &M) const {
     }
     return;
   }
+
+  auto GuardMode = getTargetMachine().Options.StackProtectorGuard;
+
   // glibc, bionic, and Fuchsia have a special slot for the stack guard.
-  if (hasStackGuardSlotTLS(Subtarget.getTargetTriple()))
+  if ((GuardMode == llvm::StackProtectorGuards::TLS ||
+       GuardMode == llvm::StackProtectorGuards::None)
+      && hasStackGuardSlotTLS(Subtarget.getTargetTriple()))
     return;
   TargetLowering::insertSSPDeclarations(M);
 }
@@ -19885,6 +19900,10 @@ static bool shouldUseHorizontalOp(bool IsSingleSource, SelectionDAG &DAG,
 /// 64-bit unsigned integer to double expansion.
 static SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG,
                                    const X86Subtarget &Subtarget) {
+  // We can't use this algorithm for strict fp. It produces -0.0 instead of +0.0
+  // when converting 0 when rounding toward negative infinity. Caller will
+  // fall back to Expand for when i64 or is legal or use FILD in 32-bit mode.
+  assert(!Op->isStrictFPOpcode() && "Expected non-strict uint_to_fp!");
   // This algorithm is not obvious. Here it is what we're trying to output:
   /*
      movq       %rax,  %xmm0
@@ -19898,8 +19917,6 @@ static SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG,
      #endif
   */
 
-  bool IsStrict = Op->isStrictFPOpcode();
-  unsigned OpNo = IsStrict ? 1 : 0;
   SDLoc dl(Op);
   LLVMContext *Context = DAG.getContext();
 
@@ -19921,7 +19938,7 @@ static SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG,
 
   // Load the 64-bit value into an XMM register.
   SDValue XR1 =
-      DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, MVT::v2i64, Op.getOperand(OpNo));
+      DAG.getNode(ISD::SCALAR_TO_VECTOR, dl, MVT::v2i64, Op.getOperand(0));
   SDValue CLod0 = DAG.getLoad(
       MVT::v4i32, dl, DAG.getEntryNode(), CPIdx0,
       MachinePointerInfo::getConstantPool(DAG.getMachineFunction()), Align(16));
@@ -19932,35 +19949,19 @@ static SDValue LowerUINT_TO_FP_i64(SDValue Op, SelectionDAG &DAG,
       MVT::v2f64, dl, CLod0.getValue(1), CPIdx1,
       MachinePointerInfo::getConstantPool(DAG.getMachineFunction()), Align(16));
   SDValue XR2F = DAG.getBitcast(MVT::v2f64, Unpck1);
-  SDValue Sub;
-  SDValue Chain;
   // TODO: Are there any fast-math-flags to propagate here?
-  if (IsStrict) {
-    Sub = DAG.getNode(ISD::STRICT_FSUB, dl, {MVT::v2f64, MVT::Other},
-                      {Op.getOperand(0), XR2F, CLod1});
-    Chain = Sub.getValue(1);
-  } else
-    Sub = DAG.getNode(ISD::FSUB, dl, MVT::v2f64, XR2F, CLod1);
+  SDValue Sub = DAG.getNode(ISD::FSUB, dl, MVT::v2f64, XR2F, CLod1);
   SDValue Result;
 
-  if (!IsStrict && Subtarget.hasSSE3() &&
+  if (Subtarget.hasSSE3() &&
       shouldUseHorizontalOp(true, DAG, Subtarget)) {
-    // FIXME: Do we need a STRICT version of FHADD?
     Result = DAG.getNode(X86ISD::FHADD, dl, MVT::v2f64, Sub, Sub);
   } else {
     SDValue Shuffle = DAG.getVectorShuffle(MVT::v2f64, dl, Sub, Sub, {1,-1});
-    if (IsStrict) {
-      Result = DAG.getNode(ISD::STRICT_FADD, dl, {MVT::v2f64, MVT::Other},
-                           {Chain, Shuffle, Sub});
-      Chain = Result.getValue(1);
-    } else
-      Result = DAG.getNode(ISD::FADD, dl, MVT::v2f64, Shuffle, Sub);
+    Result = DAG.getNode(ISD::FADD, dl, MVT::v2f64, Shuffle, Sub);
   }
   Result = DAG.getNode(ISD::EXTRACT_VECTOR_ELT, dl, MVT::f64, Result,
                        DAG.getIntPtrConstant(0, dl));
-  if (IsStrict)
-    return DAG.getMergeValues({Result, Chain}, dl);
-
   return Result;
 }
 
@@ -20286,11 +20287,14 @@ SDValue X86TargetLowering::LowerUINT_TO_FP(SDValue Op,
   if (SDValue V = LowerI64IntToFP_AVX512DQ(Op, DAG, Subtarget))
     return V;
 
-  if (SrcVT == MVT::i64 && DstVT == MVT::f64 && X86ScalarSSEf64)
+  // The transform for i64->f64 isn't correct for 0 when rounding to negative
+  // infinity. It produces -0.0, so disable under strictfp.
+  if (SrcVT == MVT::i64 && DstVT == MVT::f64 && X86ScalarSSEf64 && !IsStrict)
     return LowerUINT_TO_FP_i64(Op, DAG, Subtarget);
   if (SrcVT == MVT::i32 && X86ScalarSSEf64 && DstVT != MVT::f80)
     return LowerUINT_TO_FP_i32(Op, DAG, Subtarget);
-  if (Subtarget.is64Bit() && SrcVT == MVT::i64 && DstVT == MVT::f32)
+  if (Subtarget.is64Bit() && SrcVT == MVT::i64 &&
+      (DstVT == MVT::f32 || DstVT == MVT::f64))
     return SDValue();
 
   // Make a 64-bit buffer, and use it to build an FILD.
@@ -24152,7 +24156,7 @@ X86TargetLowering::LowerDYNAMIC_STACKALLOC(SDValue Op,
   SDValue Result;
   if (!Lower) {
     const TargetLowering &TLI = DAG.getTargetLoweringInfo();
-    unsigned SPReg = TLI.getStackPointerRegisterToSaveRestore();
+    Register SPReg = TLI.getStackPointerRegisterToSaveRestore();
     assert(SPReg && "Target cannot require DYNAMIC_STACKALLOC expansion and"
                     " not tell us which reg is the stack pointer!");
 
@@ -26084,6 +26088,15 @@ static SDValue LowerINTRINSIC_W_CHAIN(SDValue Op, const X86Subtarget &Subtarget,
                           Operation.getValue(7), Operation.getValue(8),
                           Operation.getValue(9)});
     }
+    case Intrinsic::x86_testui: {
+      SDLoc dl(Op);
+      SDValue Chain = Op.getOperand(0);
+      SDVTList VTs = DAG.getVTList(MVT::i32, MVT::Other);
+      SDValue Operation = DAG.getNode(X86ISD::TESTUI, dl, VTs, Chain);
+      SDValue SetCC = getSETCC(X86::COND_B, Operation.getValue(0), dl, DAG);
+      return DAG.getNode(ISD::MERGE_VALUES, dl, Op->getVTList(), SetCC,
+                         Operation.getValue(1));
+    }
     }
     return SDValue();
   }
@@ -31101,6 +31114,7 @@ const char *X86TargetLowering::getTargetNodeName(unsigned Opcode) const {
   NODE_NAME_CASE(AESDECWIDE128KL)
   NODE_NAME_CASE(AESENCWIDE256KL)
   NODE_NAME_CASE(AESDECWIDE256KL)
+  NODE_NAME_CASE(TESTUI)
   }
   return nullptr;
 #undef NODE_NAME_CASE
@@ -34848,7 +34862,7 @@ static bool matchBinaryShuffle(MVT MaskVT, ArrayRef<int> Mask,
         DAG.computeKnownBits(V1, DemandedZeroV1).isZero() &&
         DAG.computeKnownBits(V2, DemandedZeroV2).isZero()) {
       Shuffle = ISD::OR;
-      SrcVT = DstVT = EVT(MaskVT).changeTypeToInteger().getSimpleVT();
+      SrcVT = DstVT = MaskVT.changeTypeToInteger();
       return true;
     }
   }
diff --git a/llvm/lib/Target/X86/X86ISelLowering.h b/llvm/lib/Target/X86/X86ISelLowering.h
index d16a8f29a225e..118b3add461d6 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.h
+++ b/llvm/lib/Target/X86/X86ISelLowering.h
@@ -711,6 +711,9 @@ namespace llvm {
     // For avx512-vp2intersect
     VP2INTERSECT,
 
+    // User level interrupts - testui
+    TESTUI,
+
     /// X86 strict FP compare instructions.
     STRICT_FCMP = ISD::FIRST_TARGET_STRICTFP_OPCODE,
     STRICT_FCMPS,
diff --git a/llvm/lib/Target/X86/X86InstrInfo.td b/llvm/lib/Target/X86/X86InstrInfo.td
index d07af444337fa..cc51a32eb4980 100644
--- a/llvm/lib/Target/X86/X86InstrInfo.td
+++ b/llvm/lib/Target/X86/X86InstrInfo.td
@@ -316,6 +316,9 @@ def X86enqcmd : SDNode<"X86ISD::ENQCMD", SDT_X86ENQCMD,
                        [SDNPHasChain, SDNPSideEffect]>;
 def X86enqcmds : SDNode<"X86ISD::ENQCMDS", SDT_X86ENQCMD,
                        [SDNPHasChain, SDNPSideEffect]>;
+def X86testui : SDNode<"X86ISD::TESTUI",
+                       SDTypeProfile<1, 0, [SDTCisVT<0, i32>]>,
+                       [SDNPHasChain, SDNPSideEffect]>;
 
 def X86aesenc128kl : SDNode<"X86ISD::AESENC128KL", SDT_X86AESENCDECKL,
                             [SDNPHasChain, SDNPMayLoad, SDNPSideEffect,
@@ -978,6 +981,7 @@ def HasTSXLDTRK  : Predicate<"Subtarget->hasTSXLDTRK()">;
 def HasAMXTILE   : Predicate<"Subtarget->hasAMXTILE()">;
 def HasAMXBF16   : Predicate<"Subtarget->hasAMXBF16()">;
 def HasAMXINT8   : Predicate<"Subtarget->hasAMXINT8()">;
+def HasUINTR     : Predicate<"Subtarget->hasUINTR()">;
 def Not64BitMode : Predicate<"!Subtarget->is64Bit()">,
                              AssemblerPredicate<(all_of (not Mode64Bit)), "Not 64-bit mode">;
 def In64BitMode  : Predicate<"Subtarget->is64Bit()">,
@@ -2938,6 +2942,25 @@ let Predicates = [HasTSXLDTRK] in {
                     [(int_x86_xresldtrk)]>, XD;
 }
 
+//===----------------------------------------------------------------------===//
+// UINTR Instructions
+//
+let Predicates = [HasUINTR, In64BitMode] in {
+  def UIRET : I<0x01, MRM_EC, (outs), (ins), "uiret",
+               []>, XS;
+  def CLUI : I<0x01, MRM_EE, (outs), (ins), "clui",
+               [(int_x86_clui)]>, XS;
+  def STUI : I<0x01, MRM_EF, (outs), (ins), "stui",
+               [(int_x86_stui)]>, XS;
+
+  def SENDUIPI : I<0xC7, MRM6r, (outs), (ins GR64:$arg), "senduipi\t$arg",
+                   [(int_x86_senduipi GR64:$arg)]>, XS;
+
+  let Defs = [EFLAGS] in
+    def TESTUI : I<0x01, MRM_ED, (outs), (ins), "testui",
+                   [(set EFLAGS, (X86testui))]>, XS;
+}
+
 //===----------------------------------------------------------------------===//
 // Pattern fragments to auto generate TBM instructions.
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/X86/X86Subtarget.h b/llvm/lib/Target/X86/X86Subtarget.h
index 7f0c4283f1d9a..a189ea9d2caa9 100644
--- a/llvm/lib/Target/X86/X86Subtarget.h
+++ b/llvm/lib/Target/X86/X86Subtarget.h
@@ -415,6 +415,9 @@ class X86Subtarget final : public X86GenSubtargetInfo {
   bool HasAMXBF16 = false;
   bool HasAMXINT8 = false;
 
+  /// Processor supports User Level Interrupt instructions
+  bool HasUINTR = false;
+
   /// Processor has a single uop BEXTR implementation.
   bool HasFastBEXTR = false;
 
@@ -742,6 +745,7 @@ class X86Subtarget final : public X86GenSubtargetInfo {
   bool hasHRESET() const { return HasHRESET; }
   bool hasSERIALIZE() const { return HasSERIALIZE; }
   bool hasTSXLDTRK() const { return HasTSXLDTRK; }
+  bool hasUINTR() const { return HasUINTR; }
   bool useRetpolineIndirectCalls() const { return UseRetpolineIndirectCalls; }
   bool useRetpolineIndirectBranches() const {
     return UseRetpolineIndirectBranches;
diff --git a/llvm/lib/Transforms/IPO/BlockExtractor.cpp b/llvm/lib/Transforms/IPO/BlockExtractor.cpp
index 1d1300c6cd1d0..c6e222a096eb1 100644
--- a/llvm/lib/Transforms/IPO/BlockExtractor.cpp
+++ b/llvm/lib/Transforms/IPO/BlockExtractor.cpp
@@ -11,10 +11,12 @@
 //
 //===----------------------------------------------------------------------===//
 
+#include "llvm/Transforms/IPO/BlockExtractor.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/Statistic.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/Module.h"
+#include "llvm/IR/PassManager.h"
 #include "llvm/InitializePasses.h"
 #include "llvm/Pass.h"
 #include "llvm/Support/CommandLine.h"
@@ -38,13 +40,10 @@ cl::opt<bool> BlockExtractorEraseFuncs("extract-blocks-erase-funcs",
                                        cl::desc("Erase the existing functions"),
                                        cl::Hidden);
 namespace {
-class BlockExtractor : public ModulePass {
-  SmallVector<SmallVector<BasicBlock *, 16>, 4> GroupsOfBlocks;
-  bool EraseFunctions;
-  /// Map a function name to groups of blocks.
-  SmallVector<std::pair<std::string, SmallVector<std::string, 4>>, 4>
-      BlocksByName;
-
+class BlockExtractor {
+public:
+  BlockExtractor(bool EraseFunctions) : EraseFunctions(EraseFunctions) {}
+  bool runOnModule(Module &M);
   void init(const SmallVectorImpl<SmallVector<BasicBlock *, 16>>
                 &GroupsOfBlocksToExtract) {
     for (const SmallVectorImpl<BasicBlock *> &GroupOfBlocks :
@@ -57,11 +56,26 @@ class BlockExtractor : public ModulePass {
       loadFile();
   }
 
+private:
+  SmallVector<SmallVector<BasicBlock *, 16>, 4> GroupsOfBlocks;
+  bool EraseFunctions;
+  /// Map a function name to groups of blocks.
+  SmallVector<std::pair<std::string, SmallVector<std::string, 4>>, 4>
+      BlocksByName;
+
+  void loadFile();
+  void splitLandingPadPreds(Function &F);
+};
+
+class BlockExtractorLegacyPass : public ModulePass {
+  BlockExtractor BE;
+  bool runOnModule(Module &M) override;
+
 public:
   static char ID;
-  BlockExtractor(const SmallVectorImpl<BasicBlock *> &BlocksToExtract,
-                 bool EraseFunctions)
-      : ModulePass(ID), EraseFunctions(EraseFunctions) {
+  BlockExtractorLegacyPass(const SmallVectorImpl<BasicBlock *> &BlocksToExtract,
+                           bool EraseFunctions)
+      : ModulePass(ID), BE(EraseFunctions) {
     // We want one group per element of the input list.
     SmallVector<SmallVector<BasicBlock *, 16>, 4> MassagedGroupsOfBlocks;
     for (BasicBlock *BB : BlocksToExtract) {
@@ -69,39 +83,38 @@ class BlockExtractor : public ModulePass {
       NewGroup.push_back(BB);
       MassagedGroupsOfBlocks.push_back(NewGroup);
     }
-    init(MassagedGroupsOfBlocks);
+    BE.init(MassagedGroupsOfBlocks);
   }
 
-  BlockExtractor(const SmallVectorImpl<SmallVector<BasicBlock *, 16>>
-                     &GroupsOfBlocksToExtract,
-                 bool EraseFunctions)
-      : ModulePass(ID), EraseFunctions(EraseFunctions) {
-    init(GroupsOfBlocksToExtract);
+  BlockExtractorLegacyPass(const SmallVectorImpl<SmallVector<BasicBlock *, 16>>
+                               &GroupsOfBlocksToExtract,
+                           bool EraseFunctions)
+      : ModulePass(ID), BE(EraseFunctions) {
+    BE.init(GroupsOfBlocksToExtract);
   }
 
-  BlockExtractor() : BlockExtractor(SmallVector<BasicBlock *, 0>(), false) {}
-  bool runOnModule(Module &M) override;
-
-private:
-  void loadFile();
-  void splitLandingPadPreds(Function &F);
+  BlockExtractorLegacyPass()
+      : BlockExtractorLegacyPass(SmallVector<BasicBlock *, 0>(), false) {}
 };
+
 } // end anonymous namespace
 
-char BlockExtractor::ID = 0;
-INITIALIZE_PASS(BlockExtractor, "extract-blocks",
+char BlockExtractorLegacyPass::ID = 0;
+INITIALIZE_PASS(BlockExtractorLegacyPass, "extract-blocks",
                 "Extract basic blocks from module", false, false)
 
-ModulePass *llvm::createBlockExtractorPass() { return new BlockExtractor(); }
+ModulePass *llvm::createBlockExtractorPass() {
+  return new BlockExtractorLegacyPass();
+}
 ModulePass *llvm::createBlockExtractorPass(
     const SmallVectorImpl<BasicBlock *> &BlocksToExtract, bool EraseFunctions) {
-  return new BlockExtractor(BlocksToExtract, EraseFunctions);
+  return new BlockExtractorLegacyPass(BlocksToExtract, EraseFunctions);
 }
 ModulePass *llvm::createBlockExtractorPass(
     const SmallVectorImpl<SmallVector<BasicBlock *, 16>>
         &GroupsOfBlocksToExtract,
     bool EraseFunctions) {
-  return new BlockExtractor(GroupsOfBlocksToExtract, EraseFunctions);
+  return new BlockExtractorLegacyPass(GroupsOfBlocksToExtract, EraseFunctions);
 }
 
 /// Gets all of the blocks specified in the input file.
@@ -233,3 +246,15 @@ bool BlockExtractor::runOnModule(Module &M) {
 
   return Changed;
 }
+
+bool BlockExtractorLegacyPass::runOnModule(Module &M) {
+  return BE.runOnModule(M);
+}
+
+PreservedAnalyses BlockExtractorPass::run(Module &M,
+                                          ModuleAnalysisManager &AM) {
+  BlockExtractor BE(false);
+  BE.init(SmallVector<SmallVector<BasicBlock *, 16>, 0>());
+  return BE.runOnModule(M) ? PreservedAnalyses::none()
+                           : PreservedAnalyses::all();
+}
diff --git a/llvm/lib/Transforms/IPO/IPO.cpp b/llvm/lib/Transforms/IPO/IPO.cpp
index 901ab3fa33861..1d6a9224f7a44 100644
--- a/llvm/lib/Transforms/IPO/IPO.cpp
+++ b/llvm/lib/Transforms/IPO/IPO.cpp
@@ -41,7 +41,7 @@ void llvm::initializeIPO(PassRegistry &Registry) {
   initializeInferFunctionAttrsLegacyPassPass(Registry);
   initializeInternalizeLegacyPassPass(Registry);
   initializeLoopExtractorLegacyPassPass(Registry);
-  initializeBlockExtractorPass(Registry);
+  initializeBlockExtractorLegacyPassPass(Registry);
   initializeSingleLoopExtractorPass(Registry);
   initializeLowerTypeTestsPass(Registry);
   initializeMergeFunctionsLegacyPassPass(Registry);
diff --git a/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp b/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
index 39f0f8db0abaf..ff85164b491f1 100644
--- a/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
+++ b/llvm/lib/Transforms/IPO/PassManagerBuilder.cpp
@@ -491,6 +491,11 @@ void PassManagerBuilder::addFunctionSimplificationPasses(
   if (OptLevel > 1) {
     MPM.add(createJumpThreadingPass());         // Thread jumps
     MPM.add(createCorrelatedValuePropagationPass());
+  }
+  MPM.add(createAggressiveDCEPass()); // Delete dead instructions
+
+  // TODO: Investigate if this is too expensive at O1.
+  if (OptLevel > 1) {
     MPM.add(createDeadStoreEliminationPass());  // Delete dead stores
     MPM.add(createLICMPass(LicmMssaOptCap, LicmMssaNoAccForPromotionCap));
   }
@@ -500,8 +505,6 @@ void PassManagerBuilder::addFunctionSimplificationPasses(
   if (RerollLoops)
     MPM.add(createLoopRerollPass());
 
-  // TODO: Investigate if this is too expensive at O1.
-  MPM.add(createAggressiveDCEPass());         // Delete dead instructions
   MPM.add(createCFGSimplificationPass()); // Merge & remove BBs
   // Clean up after everything.
   MPM.add(createInstructionCombiningPass());
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
index 9c9b3f4dc8992..21a0924497341 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineAndOrXor.cpp
@@ -2283,6 +2283,8 @@ Value *InstCombinerImpl::foldOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,
   ICmpInst::Predicate PredL = LHS->getPredicate(), PredR = RHS->getPredicate();
   Value *LHS0 = LHS->getOperand(0), *RHS0 = RHS->getOperand(0);
   Value *LHS1 = LHS->getOperand(1), *RHS1 = RHS->getOperand(1);
+  auto *LHSC = dyn_cast<ConstantInt>(LHS1);
+  auto *RHSC = dyn_cast<ConstantInt>(RHS1);
 
   // Fold (icmp ult/ule (A + C1), C3) | (icmp ult/ule (A + C2), C3)
   //                   -->  (icmp ult/ule ((A & ~(C1 ^ C2)) + max(C1, C2)), C3)
@@ -2294,43 +2296,42 @@ Value *InstCombinerImpl::foldOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,
   // 3) C1 ^ C2 is one-bit mask.
   // 4) LowRange1 ^ LowRange2 and HighRange1 ^ HighRange2 are one-bit mask.
   // This implies all values in the two ranges differ by exactly one bit.
-  const APInt *LHSVal, *RHSVal;
   if ((PredL == ICmpInst::ICMP_ULT || PredL == ICmpInst::ICMP_ULE) &&
-      PredL == PredR && LHS->getType() == RHS->getType() &&
-      LHS->getType()->isIntOrIntVectorTy() && match(LHS1, m_APInt(LHSVal)) &&
-      match(RHS1, m_APInt(RHSVal)) && *LHSVal == *RHSVal && LHS->hasOneUse() &&
-      RHS->hasOneUse()) {
+      PredL == PredR && LHSC && RHSC && LHS->hasOneUse() && RHS->hasOneUse() &&
+      LHSC->getType() == RHSC->getType() &&
+      LHSC->getValue() == (RHSC->getValue())) {
+
     Value *AddOpnd;
-    const APInt *LAddVal, *RAddVal;
-    if (match(LHS0, m_Add(m_Value(AddOpnd), m_APInt(LAddVal))) &&
-        match(RHS0, m_Add(m_Specific(AddOpnd), m_APInt(RAddVal))) &&
-        LAddVal->ugt(*LHSVal) && RAddVal->ugt(*LHSVal)) {
+    ConstantInt *LAddC, *RAddC;
+    if (match(LHS0, m_Add(m_Value(AddOpnd), m_ConstantInt(LAddC))) &&
+        match(RHS0, m_Add(m_Specific(AddOpnd), m_ConstantInt(RAddC))) &&
+        LAddC->getValue().ugt(LHSC->getValue()) &&
+        RAddC->getValue().ugt(LHSC->getValue())) {
 
-      APInt DiffC = *LAddVal ^ *RAddVal;
+      APInt DiffC = LAddC->getValue() ^ RAddC->getValue();
       if (DiffC.isPowerOf2()) {
-        const APInt *MaxAddC = nullptr;
-        if (LAddVal->ult(*RAddVal))
-          MaxAddC = RAddVal;
+        ConstantInt *MaxAddC = nullptr;
+        if (LAddC->getValue().ult(RAddC->getValue()))
+          MaxAddC = RAddC;
         else
-          MaxAddC = LAddVal;
+          MaxAddC = LAddC;
 
-        APInt RRangeLow = -*RAddVal;
-        APInt RRangeHigh = RRangeLow + *LHSVal;
-        APInt LRangeLow = -*LAddVal;
-        APInt LRangeHigh = LRangeLow + *LHSVal;
+        APInt RRangeLow = -RAddC->getValue();
+        APInt RRangeHigh = RRangeLow + LHSC->getValue();
+        APInt LRangeLow = -LAddC->getValue();
+        APInt LRangeHigh = LRangeLow + LHSC->getValue();
         APInt LowRangeDiff = RRangeLow ^ LRangeLow;
         APInt HighRangeDiff = RRangeHigh ^ LRangeHigh;
         APInt RangeDiff = LRangeLow.sgt(RRangeLow) ? LRangeLow - RRangeLow
                                                    : RRangeLow - LRangeLow;
 
         if (LowRangeDiff.isPowerOf2() && LowRangeDiff == HighRangeDiff &&
-            RangeDiff.ugt(*LHSVal)) {
-          Value *NewAnd = Builder.CreateAnd(
-              AddOpnd, ConstantInt::get(LHS0->getType(), ~DiffC));
-          Value *NewAdd = Builder.CreateAdd(
-              NewAnd, ConstantInt::get(LHS0->getType(), *MaxAddC));
-          return Builder.CreateICmp(LHS->getPredicate(), NewAdd,
-                                    ConstantInt::get(LHS0->getType(), *LHSVal));
+            RangeDiff.ugt(LHSC->getValue())) {
+          Value *MaskC = ConstantInt::get(LAddC->getType(), ~DiffC);
+
+          Value *NewAnd = Builder.CreateAnd(AddOpnd, MaskC);
+          Value *NewAdd = Builder.CreateAdd(NewAnd, MaxAddC);
+          return Builder.CreateICmp(LHS->getPredicate(), NewAdd, LHSC);
         }
       }
     }
@@ -2416,8 +2417,6 @@ Value *InstCombinerImpl::foldOrOfICmps(ICmpInst *LHS, ICmpInst *RHS,
   }
 
   // This only handles icmp of constants: (icmp1 A, C1) | (icmp2 B, C2).
-  auto *LHSC = dyn_cast<ConstantInt>(LHS1);
-  auto *RHSC = dyn_cast<ConstantInt>(RHS1);
   if (!LHSC || !RHSC)
     return nullptr;
 
diff --git a/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp b/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp
index be7527b3efd18..ef5ea30ffeac1 100644
--- a/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp
+++ b/llvm/lib/Transforms/Scalar/InductiveRangeCheckElimination.cpp
@@ -233,6 +233,8 @@ class InductiveRangeCheck {
                                SmallVectorImpl<InductiveRangeCheck> &Checks);
 };
 
+struct LoopStructure;
+
 class InductiveRangeCheckElimination {
   ScalarEvolution &SE;
   BranchProbabilityInfo *BPI;
@@ -243,6 +245,10 @@ class InductiveRangeCheckElimination {
       llvm::Optional<llvm::function_ref<llvm::BlockFrequencyInfo &()> >;
   GetBFIFunc GetBFI;
 
+  // Returns true if it is profitable to do a transform basing on estimation of
+  // number of iterations.
+  bool isProfitableToTransform(const Loop &L, LoopStructure &LS);
+
 public:
   InductiveRangeCheckElimination(ScalarEvolution &SE,
                                  BranchProbabilityInfo *BPI, DominatorTree &DT,
@@ -505,9 +511,8 @@ struct LoopStructure {
     return Result;
   }
 
-  static Optional<LoopStructure> parseLoopStructure(ScalarEvolution &,
-                                                    BranchProbabilityInfo *BPI,
-                                                    Loop &, const char *&);
+  static Optional<LoopStructure> parseLoopStructure(ScalarEvolution &, Loop &,
+                                                    const char *&);
 };
 
 /// This class is used to constrain loops to run within a given iteration space.
@@ -751,8 +756,7 @@ static bool isSafeIncreasingBound(const SCEV *Start,
 }
 
 Optional<LoopStructure>
-LoopStructure::parseLoopStructure(ScalarEvolution &SE,
-                                  BranchProbabilityInfo *BPI, Loop &L,
+LoopStructure::parseLoopStructure(ScalarEvolution &SE, Loop &L,
                                   const char *&FailureReason) {
   if (!L.isLoopSimplifyForm()) {
     FailureReason = "loop not in LoopSimplify form";
@@ -787,16 +791,6 @@ LoopStructure::parseLoopStructure(ScalarEvolution &SE,
 
   unsigned LatchBrExitIdx = LatchBr->getSuccessor(0) == Header ? 1 : 0;
 
-  BranchProbability ExitProbability =
-      BPI ? BPI->getEdgeProbability(LatchBr->getParent(), LatchBrExitIdx)
-          : BranchProbability::getZero();
-
-  if (!SkipProfitabilityChecks &&
-      ExitProbability > BranchProbability(1, MaxExitProbReciprocal)) {
-    FailureReason = "short running loop, not profitable";
-    return None;
-  }
-
   ICmpInst *ICI = dyn_cast<ICmpInst>(LatchBr->getCondition());
   if (!ICI || !isa<IntegerType>(ICI->getOperand(0)->getType())) {
     FailureReason = "latch terminator branch not conditional on integral icmp";
@@ -1855,6 +1849,37 @@ bool IRCELegacyPass::runOnFunction(Function &F) {
   return Changed;
 }
 
+bool
+InductiveRangeCheckElimination::isProfitableToTransform(const Loop &L,
+                                                        LoopStructure &LS) {
+  if (SkipProfitabilityChecks)
+    return true;
+  if (GetBFI.hasValue()) {
+    BlockFrequencyInfo &BFI = (*GetBFI)();
+    uint64_t hFreq = BFI.getBlockFreq(LS.Header).getFrequency();
+    uint64_t phFreq = BFI.getBlockFreq(L.getLoopPreheader()).getFrequency();
+    if (phFreq != 0 && hFreq != 0 && (hFreq / phFreq < MinRuntimeIterations)) {
+      LLVM_DEBUG(dbgs() << "irce: could not prove profitability: "
+                        << "the estimated number of iterations basing on "
+                           "frequency info is " << (hFreq / phFreq) << "\n";);
+      return false;
+    }
+    return true;
+  }
+
+  if (!BPI)
+    return true;
+  BranchProbability ExitProbability =
+      BPI->getEdgeProbability(LS.Latch, LS.LatchBrExitIdx);
+  if (ExitProbability > BranchProbability(1, MaxExitProbReciprocal)) {
+    LLVM_DEBUG(dbgs() << "irce: could not prove profitability: "
+                      << "the exit probability is too big " << ExitProbability
+                      << "\n";);
+    return false;
+  }
+  return true;
+}
+
 bool InductiveRangeCheckElimination::run(
     Loop *L, function_ref<void(Loop *, bool)> LPMAddNewLoop) {
   if (L->getBlocks().size() >= LoopSizeCutoff) {
@@ -1894,25 +1919,15 @@ bool InductiveRangeCheckElimination::run(
 
   const char *FailureReason = nullptr;
   Optional<LoopStructure> MaybeLoopStructure =
-      LoopStructure::parseLoopStructure(SE, BPI, *L, FailureReason);
+      LoopStructure::parseLoopStructure(SE, *L, FailureReason);
   if (!MaybeLoopStructure.hasValue()) {
     LLVM_DEBUG(dbgs() << "irce: could not parse loop structure: "
                       << FailureReason << "\n";);
     return false;
   }
   LoopStructure LS = MaybeLoopStructure.getValue();
-  // Profitability check.
-  if (!SkipProfitabilityChecks && GetBFI.hasValue()) {
-    BlockFrequencyInfo &BFI = (*GetBFI)();
-    uint64_t hFreq = BFI.getBlockFreq(LS.Header).getFrequency();
-    uint64_t phFreq = BFI.getBlockFreq(Preheader).getFrequency();
-    if (phFreq != 0 && hFreq != 0 && (hFreq / phFreq < MinRuntimeIterations)) {
-      LLVM_DEBUG(dbgs() << "irce: could not prove profitability: "
-                        << "the estimated number of iterations basing on "
-                           "frequency info is " << (hFreq / phFreq) << "\n";);
-      return false;
-    }
-  }
+  if (!isProfitableToTransform(*L, LS))
+    return false;
   const SCEVAddRecExpr *IndVar =
       cast<SCEVAddRecExpr>(SE.getMinusSCEV(SE.getSCEV(LS.IndVarBase), SE.getSCEV(LS.IndVarStep)));
 
diff --git a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
index 053d96430b6b3..3b4ee2d1841d0 100644
--- a/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
+++ b/llvm/lib/Transforms/Scalar/LowerMatrixIntrinsics.cpp
@@ -1945,7 +1945,8 @@ PreservedAnalyses LowerMatrixIntrinsicsPass::run(Function &F,
   LowerMatrixIntrinsics LMT(F, TTI, &AA, &DT, &LI, &ORE);
   if (LMT.Visit()) {
     PreservedAnalyses PA;
-    PA.preserveSet<CFGAnalyses>();
+    PA.preserve<LoopAnalysis>();
+    PA.preserve<DominatorTreeAnalysis>();
     return PA;
   }
   return PreservedAnalyses::all();
diff --git a/llvm/lib/Transforms/Scalar/RewriteStatepointsForGC.cpp b/llvm/lib/Transforms/Scalar/RewriteStatepointsForGC.cpp
index c344c6c684776..0769b23f903b2 100644
--- a/llvm/lib/Transforms/Scalar/RewriteStatepointsForGC.cpp
+++ b/llvm/lib/Transforms/Scalar/RewriteStatepointsForGC.cpp
@@ -1487,7 +1487,9 @@ makeStatepointExplicitImpl(CallBase *Call, /* to replace */
   uint32_t NumPatchBytes = 0;
   uint32_t Flags = uint32_t(StatepointFlags::None);
 
-  ArrayRef<Use> CallArgs(Call->arg_begin(), Call->arg_end());
+  SmallVector<Value *, 8> CallArgs;
+  for (Value *Arg : Call->args())
+    CallArgs.push_back(Arg);
   Optional<ArrayRef<Use>> DeoptArgs;
   if (auto Bundle = Call->getOperandBundle(LLVMContext::OB_deopt))
     DeoptArgs = Bundle->Inputs;
@@ -1520,7 +1522,8 @@ makeStatepointExplicitImpl(CallBase *Call, /* to replace */
 
   Value *CallTarget = Call->getCalledOperand();
   if (Function *F = dyn_cast<Function>(CallTarget)) {
-    if (F->getIntrinsicID() == Intrinsic::experimental_deoptimize) {
+    auto IID = F->getIntrinsicID();
+    if (IID == Intrinsic::experimental_deoptimize) {
       // Calls to llvm.experimental.deoptimize are lowered to calls to the
       // __llvm_deoptimize symbol.  We want to resolve this now, since the
       // verifier does not allow taking the address of an intrinsic function.
diff --git a/llvm/lib/Transforms/Utils/UnifyLoopExits.cpp b/llvm/lib/Transforms/Utils/UnifyLoopExits.cpp
index c4868a6a799fe..0b718ed6136ea 100644
--- a/llvm/lib/Transforms/Utils/UnifyLoopExits.cpp
+++ b/llvm/lib/Transforms/Utils/UnifyLoopExits.cpp
@@ -21,7 +21,6 @@
 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/IR/Dominators.h"
 #include "llvm/InitializePasses.h"
-#include "llvm/Transforms/Scalar/LoopPassManager.h"
 #include "llvm/Transforms/Utils.h"
 #include "llvm/Transforms/Utils/BasicBlockUtils.h"
 
diff --git a/llvm/test/Analysis/BasicAA/dereferenceable.ll b/llvm/test/Analysis/BasicAA/dereferenceable.ll
index 5009c1e3deb34..fcceb50b2c588 100644
--- a/llvm/test/Analysis/BasicAA/dereferenceable.ll
+++ b/llvm/test/Analysis/BasicAA/dereferenceable.ll
@@ -1,4 +1,4 @@
-; RUN: opt -basic-aa -print-all-alias-modref-info -aa-eval -analyze < %s 2>&1 | FileCheck %s
+; RUN: opt -basic-aa -print-all-alias-modref-info -aa-eval < %s 2>&1 | FileCheck %s
 
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
 
diff --git a/llvm/test/Analysis/CostModel/AArch64/min-max.ll b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
new file mode 100644
index 0000000000000..231bcc1a38654
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/AArch64/min-max.ll
@@ -0,0 +1,511 @@
+; RUN: opt < %s -mtriple=aarch64--linux-gnu -cost-model -analyze | FileCheck %s --check-prefix=COST
+; RUN: llc < %s -mtriple=aarch64--linux-gnu | FileCheck %s --check-prefix=CODE
+
+; COST-LABEL: umin.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i8> @llvm.umin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: umin.v8i8
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT:   ret
+
+declare <8 x i8> @llvm.umin.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @umin.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+  %res = call <8 x i8> @llvm.umin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+  ret <8 x i8> %res
+}
+
+; COST-LABEL: umin.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction:   %res = call <9 x i8> @llvm.umin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: umin.v9i8
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <9 x i8> @llvm.umin.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @umin.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+  %res = call <9 x i8> @llvm.umin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+  ret <9 x i8> %res
+}
+
+; COST-LABEL: umin.v4i16
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i16> @llvm.umin.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+
+; CODE-LABEL: umin.v4i16
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.4h, v{{.*}}.4h, v{{.*}}.4h
+; CODE-NEXT:   ret
+
+declare <4 x i16> @llvm.umin.v4i16(<4 x i16>, <4 x i16>)
+define <4 x i16> @umin.v4i16(<4 x i16> %v0, <4 x i16> %v1) {
+  %res = call <4 x i16> @llvm.umin.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+  ret <4 x i16> %res
+}
+
+; COST-LABEL: umin.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction:   %res = call <16 x i8> @llvm.umin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: umin.v16i8
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <16 x i8> @llvm.umin.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @umin.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+  %res = call <16 x i8> @llvm.umin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+  ret <16 x i8> %res
+}
+
+; COST-LABEL: umin.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: umin.v8i16
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT:   ret
+
+declare <8 x i16> @llvm.umin.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @umin.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+  %res = call <8 x i16> @llvm.umin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+  ret <8 x i16> %res
+}
+
+; COST-LABEL: umin.v2i32
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i32> @llvm.umin.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+
+; CODE-LABEL: umin.v2i32
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.2s, v{{.*}}.2s, v{{.*}}.2s
+; CODE-NEXT:   ret
+
+declare <2 x i32> @llvm.umin.v2i32(<2 x i32>, <2 x i32>)
+define <2 x i32> @umin.v2i32(<2 x i32> %v0, <2 x i32> %v1) {
+  %res = call <2 x i32> @llvm.umin.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+  ret <2 x i32> %res
+}
+
+; COST-LABEL: umin.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i32> @llvm.umin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: umin.v4i32
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <4 x i32> @llvm.umin.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @umin.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+  %res = call <4 x i32> @llvm.umin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+  ret <4 x i32> %res
+}
+
+; COST-LABEL: umin.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction:   %res = call <8 x i32> @llvm.umin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: umin.v8i32
+; CODE:       bb.0
+; CODE-NEXT:   umin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   umin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <8 x i32> @llvm.umin.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @umin.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+  %res = call <8 x i32> @llvm.umin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+  ret <8 x i32> %res
+}
+
+; COST-LABEL: umin.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: umin.v2i64
+; CODE:       bb.0
+; CODE:        csel
+; CODE:        csel
+
+declare <2 x i64> @llvm.umin.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @umin.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+  %res = call <2 x i64> @llvm.umin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+  ret <2 x i64> %res
+}
+
+; COST-LABEL: smin.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i8> @llvm.smin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: smin.v8i8
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT:   ret
+
+declare <8 x i8> @llvm.smin.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @smin.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+  %res = call <8 x i8> @llvm.smin.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+  ret <8 x i8> %res
+}
+
+; COST-LABEL: smin.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction:   %res = call <9 x i8> @llvm.smin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: smin.v9i8
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <9 x i8> @llvm.smin.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @smin.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+  %res = call <9 x i8> @llvm.smin.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+  ret <9 x i8> %res
+}
+
+; COST-LABEL: smin.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction:   %res = call <16 x i8> @llvm.smin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: smin.v16i8
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <16 x i8> @llvm.smin.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @smin.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+  %res = call <16 x i8> @llvm.smin.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+  ret <16 x i8> %res
+}
+
+; COST-LABEL: smin.v4i16
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i16> @llvm.smin.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+
+; CODE-LABEL: smin.v4i16
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.4h, v{{.*}}.4h, v{{.*}}.4h
+; CODE-NEXT:   ret
+
+declare <4 x i16> @llvm.smin.v4i16(<4 x i16>, <4 x i16>)
+define <4 x i16> @smin.v4i16(<4 x i16> %v0, <4 x i16> %v1) {
+  %res = call <4 x i16> @llvm.smin.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+  ret <4 x i16> %res
+}
+
+; COST-LABEL: smin.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: smin.v8i16
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT:   ret
+
+declare <8 x i16> @llvm.smin.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @smin.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+  %res = call <8 x i16> @llvm.smin.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+  ret <8 x i16> %res
+}
+
+; COST-LABEL: smin.v2i32
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i32> @llvm.smin.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+
+; CODE-LABEL: smin.v2i32
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.2s, v{{.*}}.2s, v{{.*}}.2s
+; CODE-NEXT:   ret
+
+declare <2 x i32> @llvm.smin.v2i32(<2 x i32>, <2 x i32>)
+define <2 x i32> @smin.v2i32(<2 x i32> %v0, <2 x i32> %v1) {
+  %res = call <2 x i32> @llvm.smin.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+  ret <2 x i32> %res
+}
+
+; COST-LABEL: smin.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: smin.v4i32
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <4 x i32> @llvm.smin.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @smin.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+  %res = call <4 x i32> @llvm.smin.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+  ret <4 x i32> %res
+}
+
+; COST-LABEL: smin.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction:   %res = call <8 x i32> @llvm.smin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: smin.v8i32
+; CODE:       bb.0
+; CODE-NEXT:   smin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   smin v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <8 x i32> @llvm.smin.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @smin.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+  %res = call <8 x i32> @llvm.smin.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+  ret <8 x i32> %res
+}
+
+; COST-LABEL: smin.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: smin.v2i64
+; CODE:       bb.0
+; CODE:        csel
+; CODE:        csel
+
+declare <2 x i64> @llvm.smin.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @smin.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+  %res = call <2 x i64> @llvm.smin.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+  ret <2 x i64> %res
+}
+
+; COST-LABEL: umax.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i8> @llvm.umax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: umax.v8i8
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT:   ret
+
+declare <8 x i8> @llvm.umax.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @umax.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+  %res = call <8 x i8> @llvm.umax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+  ret <8 x i8> %res
+}
+
+; COST-LABEL: umax.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction:   %res = call <9 x i8> @llvm.umax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: umax.v9i8
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <9 x i8> @llvm.umax.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @umax.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+  %res = call <9 x i8> @llvm.umax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+  ret <9 x i8> %res
+}
+
+
+; COST-LABEL: umax.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction:   %res = call <16 x i8> @llvm.umax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: umax.v16i8
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <16 x i8> @llvm.umax.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @umax.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+  %res = call <16 x i8> @llvm.umax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+  ret <16 x i8> %res
+}
+
+; COST-LABEL: umax.v4i16
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i16> @llvm.umax.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+
+; CODE-LABEL: umax.v4i16
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.4h, v{{.*}}.4h, v{{.*}}.4h
+; CODE-NEXT:   ret
+
+declare <4 x i16> @llvm.umax.v4i16(<4 x i16>, <4 x i16>)
+define <4 x i16> @umax.v4i16(<4 x i16> %v0, <4 x i16> %v1) {
+  %res = call <4 x i16> @llvm.umax.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+  ret <4 x i16> %res
+}
+
+; COST-LABEL: umax.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i16> @llvm.umax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: umax.v8i16
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT:   ret
+
+declare <8 x i16> @llvm.umax.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @umax.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+  %res = call <8 x i16> @llvm.umax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+  ret <8 x i16> %res
+}
+
+; COST-LABEL: umax.v2i32
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i32> @llvm.umax.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+
+; CODE-LABEL: umax.v2i32
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.2s, v{{.*}}.2s, v{{.*}}.2s
+; CODE-NEXT:   ret
+
+declare <2 x i32> @llvm.umax.v2i32(<2 x i32>, <2 x i32>)
+define <2 x i32> @umax.v2i32(<2 x i32> %v0, <2 x i32> %v1) {
+  %res = call <2 x i32> @llvm.umax.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+  ret <2 x i32> %res
+}
+
+; COST-LABEL: umax.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i32> @llvm.umax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: umax.v4i32
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <4 x i32> @llvm.umax.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @umax.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+  %res = call <4 x i32> @llvm.umax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+  ret <4 x i32> %res
+}
+
+; COST-LABEL: umax.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction:   %res = call <8 x i32> @llvm.umax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: umax.v8i32
+; CODE:       bb.0
+; CODE-NEXT:   umax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   umax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <8 x i32> @llvm.umax.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @umax.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+  %res = call <8 x i32> @llvm.umax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+  ret <8 x i32> %res
+}
+
+; COST-LABEL: umax.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i64> @llvm.umax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: umax.v2i64
+; CODE:       bb.0
+; CODE:        csel
+; CODE:        csel
+
+declare <2 x i64> @llvm.umax.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @umax.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+  %res = call <2 x i64> @llvm.umax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+  ret <2 x i64> %res
+}
+
+; COST-LABEL: smax.v8i8
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i8> @llvm.smax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+
+; CODE-LABEL: smax.v8i8
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.8b, v{{.*}}.8b, v{{.*}}.8b
+; CODE-NEXT:   ret
+
+declare <8 x i8> @llvm.smax.v8i8(<8 x i8>, <8 x i8>)
+define <8 x i8> @smax.v8i8(<8 x i8> %v0, <8 x i8> %v1) {
+  %res = call <8 x i8> @llvm.smax.v8i8(<8 x i8> %v0, <8 x i8> %v1)
+  ret <8 x i8> %res
+}
+
+; COST-LABEL: smax.v9i8
+; COST-NEXT: Cost Model: Found an estimated cost of 34 for instruction:   %res = call <9 x i8> @llvm.smax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+
+; CODE-LABEL: smax.v9i8
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <9 x i8> @llvm.smax.v9i8(<9 x i8>, <9 x i8>)
+define <9 x i8> @smax.v9i8(<9 x i8> %v0, <9 x i8> %v1) {
+  %res = call <9 x i8> @llvm.smax.v9i8(<9 x i8> %v0, <9 x i8> %v1)
+  ret <9 x i8> %res
+}
+
+; COST-LABEL: smax.v16i8
+; COST-NEXT: Cost Model: Found an estimated cost of 62 for instruction:   %res = call <16 x i8> @llvm.smax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+
+; CODE-LABEL: smax.v16i8
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.16b, v{{.*}}.16b, v{{.*}}.16b
+; CODE-NEXT:   ret
+
+declare <16 x i8> @llvm.smax.v16i8(<16 x i8>, <16 x i8>)
+define <16 x i8> @smax.v16i8(<16 x i8> %v0, <16 x i8> %v1) {
+  %res = call <16 x i8> @llvm.smax.v16i8(<16 x i8> %v0, <16 x i8> %v1)
+  ret <16 x i8> %res
+}
+
+; COST-LABEL: smax.v4i16
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i16> @llvm.smax.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+
+; CODE-LABEL: smax.v4i16
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.4h, v{{.*}}.4h, v{{.*}}.4h
+; CODE-NEXT:   ret
+
+declare <4 x i16> @llvm.smax.v4i16(<4 x i16>, <4 x i16>)
+define <4 x i16> @smax.v4i16(<4 x i16> %v0, <4 x i16> %v1) {
+  %res = call <4 x i16> @llvm.smax.v4i16(<4 x i16> %v0, <4 x i16> %v1)
+  ret <4 x i16> %res
+}
+
+; COST-LABEL: smax.v8i16
+; COST-NEXT: Cost Model: Found an estimated cost of 30 for instruction:   %res = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+
+; CODE-LABEL: smax.v8i16
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.8h, v{{.*}}.8h, v{{.*}}.8h
+; CODE-NEXT:   ret
+
+declare <8 x i16> @llvm.smax.v8i16(<8 x i16>, <8 x i16>)
+define <8 x i16> @smax.v8i16(<8 x i16> %v0, <8 x i16> %v1) {
+  %res = call <8 x i16> @llvm.smax.v8i16(<8 x i16> %v0, <8 x i16> %v1)
+  ret <8 x i16> %res
+}
+
+; COST-LABEL: smax.v2i32
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i32> @llvm.smax.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+
+; CODE-LABEL: smax.v2i32
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.2s, v{{.*}}.2s, v{{.*}}.2s
+; CODE-NEXT:   ret
+
+declare <2 x i32> @llvm.smax.v2i32(<2 x i32>, <2 x i32>)
+define <2 x i32> @smax.v2i32(<2 x i32> %v0, <2 x i32> %v1) {
+  %res = call <2 x i32> @llvm.smax.v2i32(<2 x i32> %v0, <2 x i32> %v1)
+  ret <2 x i32> %res
+}
+
+; COST-LABEL: smax.v4i32
+; COST-NEXT: Cost Model: Found an estimated cost of 14 for instruction:   %res = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+
+; CODE-LABEL: smax.v4i32
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <4 x i32> @llvm.smax.v4i32(<4 x i32>, <4 x i32>)
+define <4 x i32> @smax.v4i32(<4 x i32> %v0, <4 x i32> %v1) {
+  %res = call <4 x i32> @llvm.smax.v4i32(<4 x i32> %v0, <4 x i32> %v1)
+  ret <4 x i32> %res
+}
+
+; COST-LABEL: smax.v8i32
+; COST-NEXT: Cost Model: Found an estimated cost of 10 for instruction:   %res = call <8 x i32> @llvm.smax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+
+; CODE-LABEL: smax.v8i32
+; CODE:       bb.0
+; CODE-NEXT:   smax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   smax v{{.*}}.4s, v{{.*}}.4s, v{{.*}}.4s
+; CODE-NEXT:   ret
+
+declare <8 x i32> @llvm.smax.v8i32(<8 x i32>, <8 x i32>)
+define <8 x i32> @smax.v8i32(<8 x i32> %v0, <8 x i32> %v1) {
+  %res = call <8 x i32> @llvm.smax.v8i32(<8 x i32> %v0, <8 x i32> %v1)
+  ret <8 x i32> %res
+}
+
+; COST-LABEL: smax.v2i64
+; COST-NEXT: Cost Model: Found an estimated cost of 6 for instruction:   %res = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+
+; CODE-LABEL: smax.v2i64
+; CODE:       bb.0
+; CODE:        csel
+; CODE:        csel
+
+declare <2 x i64> @llvm.smax.v2i64(<2 x i64>, <2 x i64>)
+define <2 x i64> @smax.v2i64(<2 x i64> %v0, <2 x i64> %v1) {
+  %res = call <2 x i64> @llvm.smax.v2i64(<2 x i64> %v0, <2 x i64> %v1)
+  ret <2 x i64> %res
+}
diff --git a/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll b/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll
new file mode 100644
index 0000000000000..bffaa98c82aa8
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/ARM/intrinsic-cost-kinds.ll
@@ -0,0 +1,246 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
+; RUN: opt -mtriple=armv8.1m.main -mattr=+mve.fp -cost-model -analyze -cost-kind=throughput   < %s | FileCheck %s --check-prefix=THRU
+; RUN: opt -mtriple=armv8.1m.main -mattr=+mve.fp -cost-model -analyze -cost-kind=latency      < %s | FileCheck %s --check-prefix=LATE
+; RUN: opt -mtriple=armv8.1m.main -mattr=+mve.fp -cost-model -analyze -cost-kind=code-size    < %s | FileCheck %s --check-prefix=SIZE
+; RUN: opt -mtriple=armv8.1m.main -mattr=+mve.fp -cost-model -analyze -cost-kind=size-latency < %s | FileCheck %s --check-prefix=SIZE_LATE
+
+; Test a cross-section of intrinsics for various cost-kinds.
+; Other test files may check for accuracy of a particular intrinsic
+; across subtargets or types. This is just a sanity check using an
+; ARM target and a legal scalar type (i32/float) and/or an
+; illegal vector type (16 x i32/float).
+
+declare i32 @llvm.smax.i32(i32, i32)
+declare <16 x i32> @llvm.smax.v16i32(<16 x i32>, <16 x i32>)
+
+declare float @llvm.fmuladd.f32(float, float, float)
+declare <16 x float> @llvm.fmuladd.v16f32(<16 x float>, <16 x float>, <16 x float>)
+
+declare i32 @llvm.cttz.i32(i32, i1)
+declare <16 x i32> @llvm.cttz.v16i32(<16 x i32>, i1)
+
+declare i32 @llvm.ctlz.i32(i32, i1)
+declare <16 x i32> @llvm.ctlz.v16i32(<16 x i32>, i1)
+
+declare i32 @llvm.fshl.i32(i32, i32, i32)
+declare <16 x i32> @llvm.fshl.v16i32(<16 x i32>, <16 x i32>, <16 x i32>)
+
+declare <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*>, i32, <16 x i1>, <16 x float>)
+declare void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float>, <16 x float*>, i32, <16 x i1>)
+declare float @llvm.vector.reduce.fmax.v16f32(<16 x float>)
+
+declare void @llvm.memcpy.p0i8.p0i8.i32(i8*, i8*, i32, i1)
+
+define void @smax(i32 %a, i32 %b, <16 x i32> %va, <16 x i32> %vb) {
+; THRU-LABEL: 'smax'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %s = call i32 @llvm.smax.i32(i32 %a, i32 %b)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v = call <16 x i32> @llvm.smax.v16i32(<16 x i32> %va, <16 x i32> %vb)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'smax'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.smax.i32(i32 %a, i32 %b)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.smax.v16i32(<16 x i32> %va, <16 x i32> %vb)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'smax'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.smax.i32(i32 %a, i32 %b)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.smax.v16i32(<16 x i32> %va, <16 x i32> %vb)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'smax'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.smax.i32(i32 %a, i32 %b)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.smax.v16i32(<16 x i32> %va, <16 x i32> %vb)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %s = call i32 @llvm.smax.i32(i32 %a, i32 %b)
+  %v = call <16 x i32> @llvm.smax.v16i32(<16 x i32> %va, <16 x i32> %vb)
+  ret void
+}
+
+define void @fmuladd(float %a, float %b, float %c, <16 x float> %va, <16 x float> %vb, <16 x float> %vc) {
+; THRU-LABEL: 'fmuladd'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %vc)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'fmuladd'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %s = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %vc)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'fmuladd'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %vc)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'fmuladd'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %vc)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %s = call float @llvm.fmuladd.f32(float %a, float %b, float %c)
+  %v = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> %va, <16 x float> %vb, <16 x float> %vc)
+  ret void
+}
+
+define void @cttz(i32 %a, <16 x i32> %va) {
+; THRU-LABEL: 'cttz'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.cttz.i32(i32 %a, i1 false)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %va, i1 false)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'cttz'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.cttz.i32(i32 %a, i1 false)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %va, i1 false)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'cttz'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.cttz.i32(i32 %a, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %va, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'cttz'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.cttz.i32(i32 %a, i1 false)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %va, i1 false)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %s = call i32 @llvm.cttz.i32(i32 %a, i1 false)
+  %v = call <16 x i32> @llvm.cttz.v16i32(<16 x i32> %va, i1 false)
+  ret void
+}
+
+define void @ctlz(i32 %a, <16 x i32> %va) {
+; THRU-LABEL: 'ctlz'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.ctlz.i32(i32 %a, i1 true)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %va, i1 true)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'ctlz'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.ctlz.i32(i32 %a, i1 true)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %va, i1 true)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'ctlz'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.ctlz.i32(i32 %a, i1 true)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %va, i1 true)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'ctlz'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.ctlz.i32(i32 %a, i1 true)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 528 for instruction: %v = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %va, i1 true)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %s = call i32 @llvm.ctlz.i32(i32 %a, i1 true)
+  %v = call <16 x i32> @llvm.ctlz.v16i32(<16 x i32> %va, i1 true)
+  ret void
+}
+
+define void @fshl(i32 %a, i32 %b, i32 %c, <16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc) {
+; THRU-LABEL: 'fshl'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'fshl'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'fshl'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'fshl'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %s = call i32 @llvm.fshl.i32(i32 %a, i32 %b, i32 %c)
+  %v = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %va, <16 x i32> %vb, <16 x i32> %vc)
+  ret void
+}
+
+define void @maskedgather(<16 x float*> %va, <16 x i1> %vb, <16 x float> %vc) {
+; THRU-LABEL: 'maskedgather'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: %v = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %va, i32 1, <16 x i1> %vb, <16 x float> %vc)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'maskedgather'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %va, i32 1, <16 x i1> %vb, <16 x float> %vc)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'maskedgather'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: %v = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %va, i32 1, <16 x i1> %vb, <16 x float> %vc)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'maskedgather'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: %v = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %va, i32 1, <16 x i1> %vb, <16 x float> %vc)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v = call <16 x float> @llvm.masked.gather.v16f32.v16p0f32(<16 x float*> %va, i32 1, <16 x i1> %vb, <16 x float> %vc)
+  ret void
+}
+
+define void @maskedscatter(<16 x float> %va, <16 x float*> %vb, <16 x i1> %vc) {
+; THRU-LABEL: 'maskedscatter'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> %va, <16 x float*> %vb, i32 1, <16 x i1> %vc)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'maskedscatter'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> %va, <16 x float*> %vb, i32 1, <16 x i1> %vc)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'maskedscatter'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> %va, <16 x float*> %vb, i32 1, <16 x i1> %vc)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'maskedscatter'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 576 for instruction: call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> %va, <16 x float*> %vb, i32 1, <16 x i1> %vc)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  call void @llvm.masked.scatter.v16f32.v16p0f32(<16 x float> %va, <16 x float*> %vb, i32 1, <16 x i1> %vc)
+  ret void
+}
+
+define void @reduce_fmax(<16 x float> %va) {
+; THRU-LABEL: 'reduce_fmax'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 632 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'reduce_fmax'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'reduce_fmax'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'reduce_fmax'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  %v = call float @llvm.vector.reduce.fmax.v16f32(<16 x float> %va)
+  ret void
+}
+
+define void @memcpy(i8* %a, i8* %b, i32 %c) {
+; THRU-LABEL: 'memcpy'
+; THRU-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; LATE-LABEL: 'memcpy'
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
+; LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE-LABEL: 'memcpy'
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
+; SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+; SIZE_LATE-LABEL: 'memcpy'
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
+; SIZE_LATE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+  call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
+  ret void
+}
diff --git a/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll b/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
index c76585a51ef36..d3bf703513ebd 100644
--- a/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
+++ b/llvm/test/Analysis/CostModel/X86/intrinsic-cost-kinds.ll
@@ -226,7 +226,7 @@ define void @reduce_fmax(<16 x float> %va) {
 
 define void @memcpy(i8* %a, i8* %b, i32 %c) {
 ; THRU-LABEL: 'memcpy'
-; THRU-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
+; THRU-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 1 %a, i8* align 1 %b, i32 32, i1 false)
 ; THRU-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; LATE-LABEL: 'memcpy'
diff --git a/llvm/test/Analysis/IVUsers/deep_recursion_in_scev.ll b/llvm/test/Analysis/IVUsers/deep_recursion_in_scev.ll
index 3f47516eecf2e..c25294b0a7eaf 100644
--- a/llvm/test/Analysis/IVUsers/deep_recursion_in_scev.ll
+++ b/llvm/test/Analysis/IVUsers/deep_recursion_in_scev.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s -analyze -iv-users
+; RUN: opt < %s -iv-users
 ; This is a regression test against very slow execution...
 ; In bad case it should fail by timeout.
 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128-ni:1"
diff --git a/llvm/test/Analysis/IVUsers/quadradic-exit-value.ll b/llvm/test/Analysis/IVUsers/quadradic-exit-value.ll
index e01c066cbd97a..9135b669d18d0 100644
--- a/llvm/test/Analysis/IVUsers/quadradic-exit-value.ll
+++ b/llvm/test/Analysis/IVUsers/quadradic-exit-value.ll
@@ -5,7 +5,7 @@
 ; passes *always* work on LCSSA. This should stop using a different set of
 ; checks at that point.
 
-; RUN: opt < %s -analyze -iv-users | FileCheck %s --check-prefixes=CHECK,CHECK-NO-LCSSA
+; RUN: opt < %s -analyze -iv-users -enable-new-pm=0 | FileCheck %s --check-prefixes=CHECK,CHECK-NO-LCSSA
 ; RUN: opt < %s -disable-output -passes='print<iv-users>' 2>&1 | FileCheck %s
 
 ; Provide legal integer types.
diff --git a/llvm/test/Analysis/ScalarEvolution/no-wrap-symbolic-becount.ll b/llvm/test/Analysis/ScalarEvolution/no-wrap-symbolic-becount.ll
index b49ec80c6d289..af3092e570529 100644
--- a/llvm/test/Analysis/ScalarEvolution/no-wrap-symbolic-becount.ll
+++ b/llvm/test/Analysis/ScalarEvolution/no-wrap-symbolic-becount.ll
@@ -1,5 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
-; RUN: opt < %s -S -analyze -scalar-evolution | FileCheck %s
+; RUN: opt < %s -S -analyze -scalar-evolution -enable-new-pm=0 | FileCheck %s
+; RUN: opt < %s -S -passes='print<scalar-evolution>' 2>&1 | FileCheck %s
 
 define i32 @test_01(i32 %start, i32* %p, i32* %q) {
 ; CHECK-LABEL: 'test_01'
@@ -7,7 +8,7 @@ define i32 @test_01(i32 %start, i32* %p, i32* %q) {
 ; CHECK-NEXT:    %0 = zext i32 %start to i64
 ; CHECK-NEXT:    --> (zext i32 %start to i64) U: [0,4294967296) S: [0,4294967296)
 ; CHECK-NEXT:    %indvars.iv = phi i64 [ %indvars.iv.next, %backedge ], [ %0, %entry ]
-; CHECK-NEXT:    --> {(zext i32 %start to i64),+,-1}<nsw><%loop> U: [0,4294967296) S: [0,4294967296) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {(zext i32 %start to i64),+,-1}<nsw><%loop> U: [-4294967295,4294967296) S: [-4294967295,4294967296) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %iv = phi i32 [ %start, %entry ], [ %iv.next, %backedge ]
 ; CHECK-NEXT:    --> {%start,+,-1}<%loop> U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:    %iv.next = add i32 %iv, -1
@@ -21,7 +22,7 @@ define i32 @test_01(i32 %start, i32* %p, i32* %q) {
 ; CHECK-NEXT:    %stop = load i32, i32* %load.addr, align 4
 ; CHECK-NEXT:    --> %stop U: full-set S: full-set Exits: <<Unknown>> LoopDispositions: { %loop: Variant }
 ; CHECK-NEXT:    %indvars.iv.next = add nsw i64 %indvars.iv, -1
-; CHECK-NEXT:    --> {(-1 + (zext i32 %start to i64))<nsw>,+,-1}<nsw><%loop> U: [-4294967296,4294967295) S: [-1,4294967295) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
+; CHECK-NEXT:    --> {(-1 + (zext i32 %start to i64))<nsw>,+,-1}<nsw><%loop> U: [-4294967296,4294967295) S: [-4294967296,4294967295) Exits: <<Unknown>> LoopDispositions: { %loop: Computable }
 ; CHECK-NEXT:  Determining loop execution counts for: @test_01
 ; CHECK-NEXT:  Loop %loop: <multiple exits> Unpredictable backedge-taken count.
 ; CHECK-NEXT:    exit count for loop: (zext i32 %start to i64)
diff --git a/llvm/test/CodeGen/AMDGPU/dbg-value-ends-sched-region.mir b/llvm/test/CodeGen/AMDGPU/dbg-value-ends-sched-region.mir
new file mode 100644
index 0000000000000..9c0f34ec4c53e
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/dbg-value-ends-sched-region.mir
@@ -0,0 +1,132 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx906 -verify-machineinstrs -run-pass=machine-scheduler -o - %s | FileCheck %s
+# The DBG_VALUE in bb.5 ends a scheduling region, and its uses should
+# not be tracked like a normal instruction.
+
+--- |
+  define void @dbg_value_ends_sched_region() {
+    unreachable
+  }
+
+  !llvm.dbg.cu = !{!0}
+  !llvm.module.flags = !{!3, !4}
+
+  !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 12.0.0 (/src/external/llvm-project/clang ff12a3768d17240b01cbe9957e3c9e8f8ffe9854)", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !2, globals: !2, splitDebugInlining: false, nameTableKind: None)
+  !1 = !DIFile(filename: "/tmp/comgr-054202/input/CompileSource", directory: "/home/matt/bugs/SWDEV-253507")
+  !2 = !{}
+  !3 = !{i32 2, !"Debug Info Version", i32 3}
+  !4 = !{i32 7, !"PIC Level", i32 1}
+  !5 = !{!6, !9, i64 68}
+  !6 = !{!"ShaderData", !7, i64 0, !7, i64 16, !7, i64 32, !7, i64 48, !9, i64 64, !9, i64 68, !9, i64 72, !9, i64 76, !9, i64 80, !10, i64 84, !10, i64 88, !9, i64 92, !9, i64 96, !10, i64 100, !10, i64 104, !11, i64 112, !11, i64 144, !12, i64 176, !12, i64 184, !7, i64 192, !7, i64 208, !7, i64 224, !11, i64 240, !9, i64 272, !9, i64 276, !9, i64 280, !10, i64 284, !7, i64 288, !7, i64 304, !7, i64 320, !7, i64 336}
+  !7 = !{!"omnipotent char", !8, i64 0}
+  !8 = !{!"Simple C/C++ TBAA"}
+  !9 = !{!"int", !7, i64 0}
+  !10 = !{!"float", !7, i64 0}
+  !11 = !{!"differential3", !7, i64 0, !7, i64 16}
+  !12 = !{!"differential", !10, i64 0, !10, i64 4}
+  !13 = !{!6, !9, i64 280}
+  !16 = !DILocalVariable(name: "data", scope: !17, file: !18, line: 24)
+  !17 = distinct !DISubprogram(name: "svm_node_geometry", scope: !18, file: !18, line: 21, type: !19, scopeLine: 23, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !2)
+  !18 = !DIFile(filename: "source/kernel/svm/svm_geometry.h", directory: "/home/matt/bugs/SWDEV-253507")
+  !19 = !DISubroutineType(types: !20)
+  !20 = !{null}
+  !21 = !DILocation(line: 0, scope: !17)
+
+...
+---
+name:            dbg_value_ends_sched_region
+tracksRegLiveness: true
+machineFunctionInfo:
+  scratchRSrcReg:  '$sgpr0_sgpr1_sgpr2_sgpr3'
+  frameOffsetReg:  '$sgpr33'
+  stackPtrOffsetReg: '$sgpr32'
+body:             |
+  ; CHECK-LABEL: name: dbg_value_ends_sched_region
+  ; CHECK: bb.0:
+  ; CHECK:   successors: %bb.1(0x80000000)
+  ; CHECK:   liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+  ; CHECK:   [[COPY:%[0-9]+]]:vreg_64 = COPY $vgpr0_vgpr1
+  ; CHECK:   [[COPY1:%[0-9]+]]:vreg_64 = COPY $vgpr2_vgpr3
+  ; CHECK:   [[DEF:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+  ; CHECK:   [[DEF1:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+  ; CHECK:   [[DEF2:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+  ; CHECK:   [[DEF3:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+  ; CHECK:   [[DEF4:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+  ; CHECK:   [[DEF5:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+  ; CHECK:   [[DEF6:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+  ; CHECK:   [[DEF7:%[0-9]+]]:vgpr_32 = IMPLICIT_DEF
+  ; CHECK:   [[DEF8:%[0-9]+]]:vreg_64 = IMPLICIT_DEF
+  ; CHECK:   undef %11.sub0:vreg_64 = V_MOV_B32_e32 0, implicit $exec
+  ; CHECK: bb.1:
+  ; CHECK:   successors: %bb.2(0x80000000)
+  ; CHECK:   dead %12:vreg_64 = COPY [[DEF]]
+  ; CHECK:   [[COPY2:%[0-9]+]]:vgpr_32 = COPY [[DEF3]]
+  ; CHECK:   dead %14:vgpr_32 = COPY [[DEF2]]
+  ; CHECK:   [[COPY3:%[0-9]+]]:vgpr_32 = COPY [[DEF5]].sub1
+  ; CHECK:   dead %16:vgpr_32 = COPY %11.sub0
+  ; CHECK:   undef %17.sub0:vreg_64, %18:sreg_64_xexec = V_ADD_CO_U32_e64 [[DEF4]].sub0, [[DEF6]].sub0, 0, implicit $exec
+  ; CHECK:   dead undef %17.sub1:vreg_64, dead %19:sreg_64_xexec = V_ADDC_U32_e64 [[DEF4]].sub1, [[DEF6]].sub1, %18, 0, implicit $exec
+  ; CHECK:   [[GLOBAL_LOAD_DWORDX2_:%[0-9]+]]:vreg_64 = GLOBAL_LOAD_DWORDX2 [[DEF1]], 0, 0, 0, 0, implicit $exec :: (load 8, addrspace 1)
+  ; CHECK:   dead %20:sreg_64 = V_CMP_GT_I32_e64 4, [[DEF7]], implicit $exec
+  ; CHECK:   GLOBAL_STORE_DWORDX2 [[COPY]], [[DEF8]], 288, 0, 0, 0, implicit $exec :: (store 8, addrspace 1)
+  ; CHECK: bb.2:
+  ; CHECK:   successors: %bb.3(0x80000000)
+  ; CHECK:   [[COPY4:%[0-9]+]]:vgpr_32 = COPY [[COPY2]]
+  ; CHECK: bb.3:
+  ; CHECK:   successors: %bb.2(0x40000000), %bb.4(0x40000000)
+  ; CHECK:   undef [[DEF5]].sub1:vreg_64 = COPY [[COPY3]]
+  ; CHECK:   S_CBRANCH_EXECZ %bb.2, implicit $exec
+  ; CHECK: bb.4:
+  ; CHECK:   successors: %bb.5(0x80000000)
+  ; CHECK:   dead %21:sreg_64 = COPY $exec
+  ; CHECK:   dead %22:vreg_128 = GLOBAL_LOAD_DWORDX4 [[COPY1]], 0, 0, 0, 0, implicit $exec :: (load 16, addrspace 1)
+  ; CHECK:   DBG_VALUE %22,
+  ; CHECK: bb.5:
+  ; CHECK:   successors: %bb.3(0x40000000), %bb.1(0x40000000)
+  ; CHECK:   S_CBRANCH_EXECZ %bb.3, implicit $exec
+  ; CHECK:   S_BRANCH %bb.1
+  bb.0:
+    liveins: $vgpr0_vgpr1, $vgpr2_vgpr3
+
+    %0:vreg_64 = COPY $vgpr0_vgpr1
+    %1:vreg_64 = COPY $vgpr2_vgpr3
+    %2:vreg_64 = IMPLICIT_DEF
+    %3:vreg_64 = IMPLICIT_DEF
+    %4:vgpr_32 = IMPLICIT_DEF
+    %5:vgpr_32 = IMPLICIT_DEF
+    %6:vreg_64 = IMPLICIT_DEF
+    %7:vreg_64 = IMPLICIT_DEF
+    %8:vreg_64 = IMPLICIT_DEF
+    %9:vgpr_32 = IMPLICIT_DEF
+    %10:vreg_64 = IMPLICIT_DEF
+    undef %11.sub0:vreg_64 = V_MOV_B32_e32 0, implicit $exec
+
+  bb.1:
+    %12:vreg_64 = COPY %2
+    %13:vgpr_32 = COPY %5
+    dead %14:vgpr_32 = COPY %4
+    %15:vgpr_32 = COPY %7.sub1
+    dead %16:vgpr_32 = COPY %11.sub0
+    undef %17.sub0:vreg_64, %18:sreg_64_xexec = V_ADD_CO_U32_e64 %6.sub0, %8.sub0, 0, implicit $exec
+    dead %17.sub1:vreg_64, dead %19:sreg_64_xexec = V_ADDC_U32_e64 %6.sub1, %8.sub1, %18, 0, implicit $exec
+    %6:vreg_64 = GLOBAL_LOAD_DWORDX2 %3, 0, 0, 0, 0, implicit $exec :: (load 8, addrspace 1)
+    dead %20:sreg_64 = V_CMP_GT_I32_e64 4, %9, implicit $exec
+    GLOBAL_STORE_DWORDX2 %0, %10, 288, 0, 0, 0, implicit $exec :: (store 8, addrspace 1)
+
+  bb.2:
+    %5:vgpr_32 = COPY %13
+
+  bb.3:
+    undef %7.sub1:vreg_64 = COPY %15
+    S_CBRANCH_EXECZ %bb.2, implicit $exec
+
+  bb.4:
+    dead %21:sreg_64 = COPY $exec
+    %22:vreg_128 = GLOBAL_LOAD_DWORDX4 %1, 0, 0, 0, 0, implicit $exec :: (load 16, addrspace 1)
+    DBG_VALUE %22, $noreg, !16, !DIExpression(DW_OP_constu, 1, DW_OP_swap, DW_OP_xderef), debug-location !21
+
+  bb.5:
+    S_CBRANCH_EXECZ %bb.3, implicit $exec
+    S_BRANCH %bb.1
+
+...
diff --git a/llvm/test/CodeGen/AMDGPU/opt-pipeline.ll b/llvm/test/CodeGen/AMDGPU/opt-pipeline.ll
index 2c157cb801863..e99428689f501 100644
--- a/llvm/test/CodeGen/AMDGPU/opt-pipeline.ll
+++ b/llvm/test/CodeGen/AMDGPU/opt-pipeline.ll
@@ -513,9 +513,10 @@
 ; GCN-O2-NEXT:       Lazy Value Information Analysis
 ; GCN-O2-NEXT:       Jump Threading
 ; GCN-O2-NEXT:       Value Propagation
+; GCN-O2-NEXT:       Post-Dominator Tree Construction
+; GCN-O2-NEXT:       Aggressive Dead Code Elimination
 ; GCN-O2-NEXT:       Basic Alias Analysis (stateless AA impl)
 ; GCN-O2-NEXT:       Function Alias Analysis Results
-; GCN-O2-NEXT:       Post-Dominator Tree Construction
 ; GCN-O2-NEXT:       Memory SSA
 ; GCN-O2-NEXT:       Dead Store Elimination
 ; GCN-O2-NEXT:       Natural Loop Information
@@ -528,8 +529,6 @@
 ; GCN-O2-NEXT:       Lazy Block Frequency Analysis
 ; GCN-O2-NEXT:       Loop Pass Manager
 ; GCN-O2-NEXT:         Loop Invariant Code Motion
-; GCN-O2-NEXT:       Post-Dominator Tree Construction
-; GCN-O2-NEXT:       Aggressive Dead Code Elimination
 ; GCN-O2-NEXT:       Simplify the CFG
 ; GCN-O2-NEXT:       Dominator Tree Construction
 ; GCN-O2-NEXT:       Basic Alias Analysis (stateless AA impl)
@@ -876,9 +875,10 @@
 ; GCN-O3-NEXT:       Lazy Value Information Analysis
 ; GCN-O3-NEXT:       Jump Threading
 ; GCN-O3-NEXT:       Value Propagation
+; GCN-O3-NEXT:       Post-Dominator Tree Construction
+; GCN-O3-NEXT:       Aggressive Dead Code Elimination
 ; GCN-O3-NEXT:       Basic Alias Analysis (stateless AA impl)
 ; GCN-O3-NEXT:       Function Alias Analysis Results
-; GCN-O3-NEXT:       Post-Dominator Tree Construction
 ; GCN-O3-NEXT:       Memory SSA
 ; GCN-O3-NEXT:       Dead Store Elimination
 ; GCN-O3-NEXT:       Natural Loop Information
@@ -891,8 +891,6 @@
 ; GCN-O3-NEXT:       Lazy Block Frequency Analysis
 ; GCN-O3-NEXT:       Loop Pass Manager
 ; GCN-O3-NEXT:         Loop Invariant Code Motion
-; GCN-O3-NEXT:       Post-Dominator Tree Construction
-; GCN-O3-NEXT:       Aggressive Dead Code Elimination
 ; GCN-O3-NEXT:       Simplify the CFG
 ; GCN-O3-NEXT:       Dominator Tree Construction
 ; GCN-O3-NEXT:       Basic Alias Analysis (stateless AA impl)
diff --git a/llvm/test/CodeGen/AMDGPU/propagate-attributes-flat-work-group-size.ll b/llvm/test/CodeGen/AMDGPU/propagate-attributes-flat-work-group-size.ll
new file mode 100644
index 0000000000000..30c6eded2397a
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/propagate-attributes-flat-work-group-size.ll
@@ -0,0 +1,48 @@
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -amdgpu-propagate-attributes-late %s | FileCheck %s
+
+; CHECK: define internal void @max_flat_1_1024() #0 {
+define internal void @max_flat_1_1024() #0 {
+  ret void
+}
+
+; CHECK: define internal void @max_flat_1_256() #1 {
+define internal void @max_flat_1_256() #1 {
+  ret void
+}
+
+; CHECK: define amdgpu_kernel void @kernel_1_256_call_default() #1 {
+define amdgpu_kernel void @kernel_1_256_call_default() #1 {
+  call void @default()
+  ret void
+}
+
+; CHECK: define amdgpu_kernel void @kernel_1_256_call_1_256() #1 {
+define amdgpu_kernel void @kernel_1_256_call_1_256() #1 {
+  call void @max_flat_1_256()
+  ret void
+}
+
+; CHECK: define amdgpu_kernel void @kernel_1_256_call_64_64() #1 {
+define amdgpu_kernel void @kernel_1_256_call_64_64() #1 {
+  call void @max_flat_64_64()
+  ret void
+}
+
+; CHECK: define internal void @max_flat_64_64() #2 {
+define internal void @max_flat_64_64() #2 {
+  ret void
+}
+
+; CHECK: define internal void @default() #2 {
+define internal void @default() #3 {
+  ret void
+}
+
+attributes #0 = { noinline "amdgpu-flat-work-group-size"="1,1024" }
+attributes #1 = { noinline "amdgpu-flat-work-group-size"="1,256" }
+attributes #2 = { noinline "amdgpu-flat-work-group-size"="64,64" }
+attributes #3 = { noinline }
+
+; CHECK: attributes #0 = { noinline "amdgpu-flat-work-group-size"="1,1024"
+; CHECK-NEXT: attributes #1 = { noinline "amdgpu-flat-work-group-size"="1,256"
+; CHECK-NEXT: attributes #2 = { noinline "amdgpu-flat-work-group-size"="1,256"
diff --git a/llvm/test/CodeGen/AMDGPU/return-with-successors.mir b/llvm/test/CodeGen/AMDGPU/return-with-successors.mir
new file mode 100644
index 0000000000000..00b1faca0e902
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/return-with-successors.mir
@@ -0,0 +1,44 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx900 -verify-machineinstrs -run-pass=liveintervals -o - %s | FileCheck %s
+# Test that getNoPreservedMask is implemented, which is called when
+# return blocks have successors.
+
+---
+name: endpgm_with_successors
+tracksRegLiveness: true
+body:             |
+  ; CHECK-LABEL: name: endpgm_with_successors
+  ; CHECK: bb.0:
+  ; CHECK:   successors: %bb.1(0x80000000)
+  ; CHECK:   S_ENDPGM 0
+  ; CHECK: bb.1:
+  ; CHECK:   S_NOP 0
+  bb.0:
+    successors: %bb.1
+    S_ENDPGM 0
+
+  bb.1:
+    S_NOP 0
+
+...
+
+---
+name: setpc_with_successors
+tracksRegLiveness: true
+body:             |
+  ; CHECK-LABEL: name: setpc_with_successors
+  ; CHECK: bb.0:
+  ; CHECK:   successors: %bb.1(0x80000000)
+  ; CHECK:   liveins: $sgpr30_sgpr31
+  ; CHECK:   S_SETPC_B64 $sgpr30_sgpr31
+  ; CHECK: bb.1:
+  ; CHECK:   S_NOP 0
+  bb.0:
+    liveins: $sgpr30_sgpr31
+    successors: %bb.1
+    S_SETPC_B64 $sgpr30_sgpr31
+
+  bb.1:
+    S_NOP 0
+
+...
diff --git a/llvm/test/CodeGen/AMDGPU/srem.ll b/llvm/test/CodeGen/AMDGPU/srem.ll
index 34bf6be15354d..e4dd327a0379a 100644
--- a/llvm/test/CodeGen/AMDGPU/srem.ll
+++ b/llvm/test/CodeGen/AMDGPU/srem.ll
@@ -1,6 +1,17 @@
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
-; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood < %s
+; RUN: llc -amdgpu-scalarize-global-loads=false --verify-machineinstrs -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=SI,GFX9 -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tahiti -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=amdgcn -mcpu=tonga -verify-machineinstrs < %s | FileCheck -check-prefix=SI -check-prefix=FUNC %s
+; RUN: llc -amdgpu-scalarize-global-loads=false -march=r600 -mcpu=redwood -verify-machineinstrs < %s
+
+; FUNC-LABEL: {{^}}srem_i16_7:
+; GFX9: s_movk_i32 {{s[0-9]+}}, 0x4925
+; GFX9: v_mul_i32_i24_sdwa
+define amdgpu_kernel void @srem_i16_7(i16 addrspace(1)* %out, i16 addrspace(1)* %in) {
+  %num = load i16, i16 addrspace(1) * %in
+  %result = srem i16 %num, 7
+  store i16 %result, i16 addrspace(1)* %out
+  ret void
+}
 
 define amdgpu_kernel void @srem_i32(i32 addrspace(1)* %out, i32 addrspace(1)* %in) {
   %den_ptr = getelementptr i32, i32 addrspace(1)* %in, i32 1
diff --git a/llvm/test/CodeGen/AMDGPU/stack-size-overflow.ll b/llvm/test/CodeGen/AMDGPU/stack-size-overflow.ll
index ed7e2092e8867..8550b7b1ed484 100644
--- a/llvm/test/CodeGen/AMDGPU/stack-size-overflow.ll
+++ b/llvm/test/CodeGen/AMDGPU/stack-size-overflow.ll
@@ -3,12 +3,45 @@
 
 declare void @llvm.memset.p5i8.i32(i8 addrspace(5)* nocapture, i8, i32, i32, i1) #1
 
-; ERROR: error: stack size limit exceeded (4294967296) in stack_size_limit
-; GCN: ; ScratchSize: 4294967296
-define amdgpu_kernel void @stack_size_limit() #0 {
+; ERROR: error: stack size limit exceeded (131061) in stack_size_limit_wave64
+; GCN: ; ScratchSize: 131061
+define amdgpu_kernel void @stack_size_limit_wave64() #0 {
 entry:
-  %alloca = alloca [1073741823 x i32], align 4, addrspace(5)
-  %bc = bitcast [1073741823 x i32] addrspace(5)* %alloca to i8 addrspace(5)*
-  call void @llvm.memset.p5i8.i32(i8 addrspace(5)* %bc, i8 9, i32 1073741823, i32 1, i1 true)
+  %alloca = alloca [131057 x i8], align 1, addrspace(5)
+  %alloca.bc = bitcast [131057 x i8] addrspace(5)* %alloca to i8 addrspace(5)*
+  call void @llvm.memset.p5i8.i32(i8 addrspace(5)* %alloca.bc, i8 9, i32 131057, i32 1, i1 true)
   ret void
 }
+
+; ERROR: error: stack size limit exceeded (262117) in stack_size_limit_wave32
+; GCN: ; ScratchSize: 262117
+define amdgpu_kernel void @stack_size_limit_wave32() #1 {
+entry:
+  %alloca = alloca [262113 x i8], align 1, addrspace(5)
+  %alloca.bc = bitcast [262113 x i8] addrspace(5)* %alloca to i8 addrspace(5)*
+  call void @llvm.memset.p5i8.i32(i8 addrspace(5)* %alloca.bc, i8 9, i32 262113, i32 1, i1 true)
+  ret void
+}
+
+; ERROR-NOT: error:
+; GCN: ; ScratchSize: 131056
+define amdgpu_kernel void @max_stack_size_wave64() #0 {
+entry:
+  %alloca = alloca [131052 x i8], align 1, addrspace(5)
+  %alloca.bc = bitcast [131052 x i8] addrspace(5)* %alloca to i8 addrspace(5)*
+  call void @llvm.memset.p5i8.i32(i8 addrspace(5)* %alloca.bc, i8 9, i32 131052, i32 1, i1 true)
+  ret void
+}
+
+; ERROR-NOT: error:
+; GCN: ; ScratchSize: 262112
+define amdgpu_kernel void @max_stack_size_wave32() #1 {
+entry:
+  %alloca = alloca [262108 x i8], align 1, addrspace(5)
+  %alloca.bc = bitcast [262108 x i8] addrspace(5)* %alloca to i8 addrspace(5)*
+  call void @llvm.memset.p5i8.i32(i8 addrspace(5)* %alloca.bc, i8 9, i32 262108, i32 1, i1 true)
+  ret void
+}
+
+attributes #0 = { "target-cpu" = "gfx900" }
+attributes #1 = { "target-cpu" = "gfx1010" }
diff --git a/llvm/test/CodeGen/AMDGPU/v_swap_b32.mir b/llvm/test/CodeGen/AMDGPU/v_swap_b32.mir
index 3190641ae6910..d557060207df5 100644
--- a/llvm/test/CodeGen/AMDGPU/v_swap_b32.mir
+++ b/llvm/test/CodeGen/AMDGPU/v_swap_b32.mir
@@ -517,7 +517,9 @@ body:             |
 ...
 
 # GCN-LABEL: name: swap_virt_copy_subreg_impdef_super
-# GCN: %0.sub0:vreg_64, %1.sub0:vreg_64 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
+# GCN:      %2:vreg_64 = IMPLICIT_DEF
+# GCN-NEXT: %2.sub1:vreg_64 = COPY %0.sub1
+# GCN-NEXT: %0.sub0:vreg_64, %1.sub0:vreg_64 = V_SWAP_B32 %1.sub0, %0.sub0, implicit $exec
 ---
 name:            swap_virt_copy_subreg_impdef_super
 registers:
@@ -672,3 +674,222 @@ body:             |
     %1 = COPY %2
     S_ENDPGM 0
 ...
+
+# GCN-LABEL: name: swap_liveness_error_mov
+# GCN:      $vgpr6 = V_MOV_B32_e32 $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr1, $vgpr5 = V_SWAP_B32 $vgpr5, $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr5_vgpr6 = IMPLICIT_DEF
+# GCN-NEXT: $vgpr6 = V_MOV_B32_e32 $vgpr7, implicit $exec, implicit $vgpr6_vgpr7
+# GCN-NEXT: $vgpr5 = V_MOV_B32_e32 $vgpr6, implicit $exec
+
+---
+name: swap_liveness_error_mov
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr5, $vgpr1_vgpr2
+
+    $vgpr6 = V_MOV_B32_e32 $vgpr1, implicit $exec, implicit killed $vgpr1_vgpr2
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr5, implicit $exec
+    $vgpr5 = V_MOV_B32_e32 $vgpr6, implicit $exec, implicit-def $vgpr5_vgpr6, implicit $vgpr6_vgpr7
+    $vgpr6 = V_MOV_B32_e32 $vgpr7, implicit $exec, implicit $vgpr6_vgpr7
+    $vgpr5 = V_MOV_B32_e32 $vgpr6, implicit $exec
+    S_ENDPGM 0
+...
+
+# GCN-LABEL: name: swap_liveness_error_copy
+# GCN:      $vgpr6 = COPY $vgpr1
+# GCN-NEXT: $vgpr1, $vgpr5 = V_SWAP_B32 $vgpr5, $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr5_vgpr6 = IMPLICIT_DEF
+# GCN-NEXT: $vgpr6 = COPY $vgpr7, implicit $vgpr6_vgpr7
+# GCN-NEXT: $vgpr5 = COPY $vgpr6
+
+---
+name: swap_liveness_error_copy
+tracksRegLiveness: true
+body:             |
+  bb.0:
+    liveins: $vgpr5, $vgpr1_vgpr2
+
+    $vgpr6 = COPY $vgpr1, implicit killed $vgpr1_vgpr2
+    $vgpr1 = COPY killed $vgpr5
+    $vgpr5 = COPY $vgpr6, implicit-def $vgpr5_vgpr6, implicit $vgpr6_vgpr7
+    $vgpr6 = COPY $vgpr7, implicit $vgpr6_vgpr7
+    $vgpr5 = COPY $vgpr6
+    S_ENDPGM 0
+...
+
+# GCN-LABEL: name: swap_killed_t_early
+# GCN:      $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+# GCN-NEXT: $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec, implicit killed $vgpr2
+# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec
+# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 undef $vgpr2, implicit $exec
+
+---
+name:            swap_killed_t_early
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec, implicit killed $vgpr2
+    $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+    $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec
+    $vgpr1 = V_MOV_B32_e32 undef $vgpr2, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: swap_killed_t_late
+# GCN:      $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+# GCN-NEXT: $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec
+# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec, implicit killed $vgpr2
+# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 undef $vgpr2, implicit $exec
+
+---
+name:            swap_killed_t_late
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec
+    $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+    $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec, implicit killed $vgpr2
+    $vgpr1 = V_MOV_B32_e32 undef $vgpr2, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: swap_killed_x
+# GCN:      $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+# GCN-NEXT: $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec
+# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec, implicit killed $vgpr0
+# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
+
+---
+name:            swap_killed_x
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr3 = V_MOV_B32_e32 killed $vgpr4, implicit $exec
+    $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+    $vgpr5 = V_MOV_B32_e32 killed $vgpr6, implicit $exec, implicit killed $vgpr0
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: indirect_mov_t
+# GCN:      $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec, implicit $m0
+# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
+# GCN-NEXT: S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+
+---
+name:            indirect_mov_t
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec, implicit $m0
+    $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: indirect_mov_x
+# GCN:      $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec, implicit $m0
+# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
+# GCN-NEXT: S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+
+---
+name:            indirect_mov_x
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec, implicit $m0
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: indirect_mov_y
+# GCN:      $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+# GCN-NEXT: $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+# GCN-NEXT: $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec, implicit $m0
+# GCN-NEXT: S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+
+---
+name:            indirect_mov_y
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr2 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr0 = V_MOV_B32_e32 killed $vgpr1, implicit $exec
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr2, implicit $exec, implicit $m0
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: implicit_ops_mov_x_swap_b32
+# GCN: $vgpr0, $vgpr1 = V_SWAP_B32 $vgpr1, $vgpr0, implicit $exec, implicit $vgpr2, implicit killed $vgpr1_vgpr2
+
+---
+name:            implicit_ops_mov_x_swap_b32
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr3 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr0 = V_MOV_B32_e32 $vgpr1, implicit $exec, implicit $vgpr2, implicit killed $vgpr1_vgpr2
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr3, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: name: implict_ops_mov_x_swap_b64
+# GCN:      %2:vreg_64 = COPY %0
+# GCN-NEXT: %0:vreg_64 = COPY %1, implicit $vgpr0
+# GCN-NEXT: %1:vreg_64 = COPY %2
+
+---
+name:            implict_ops_mov_x_swap_b64
+registers:
+  - { id: 0, class: vreg_64 }
+  - { id: 1, class: vreg_64 }
+  - { id: 2, class: vreg_64 }
+body:             |
+  bb.0:
+    %0 = IMPLICIT_DEF
+    %1 = IMPLICIT_DEF
+    %2 = COPY %0
+    %0 = COPY %1, implicit $vgpr0
+    %1 = COPY %2
+...
+
+# GCN-LABEL: implicit_ops_mov_t_swap_b32
+# GCN:      $vgpr1 = IMPLICIT_DEF
+# GCN-NEXT: $vgpr0, $vgpr1 = V_SWAP_B32 $vgpr1, $vgpr0, implicit $exec
+
+---
+name:            implicit_ops_mov_t_swap_b32
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr3 = V_MOV_B32_e32 killed $vgpr0, implicit $exec, implicit $vgpr2, implicit killed $vgpr1_vgpr2, implicit-def $vgpr1
+    $vgpr0 = V_MOV_B32_e32 $vgpr1, implicit $exec
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr3, implicit $exec
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
+
+# GCN-LABEL: implicit_ops_mov_y_swap_b32
+# GCN:      $vgpr0, $vgpr1 = V_SWAP_B32 $vgpr1, $vgpr0, implicit $exec
+# GCN-NEXT: $vgpr0_vgpr1 = IMPLICIT_DEF
+
+---
+name:            implicit_ops_mov_y_swap_b32
+body:             |
+  bb.0:
+    liveins: $vgpr0, $vgpr1, $sgpr30_sgpr31
+    $vgpr3 = V_MOV_B32_e32 killed $vgpr0, implicit $exec
+    $vgpr0 = V_MOV_B32_e32 $vgpr1, implicit $exec
+    $vgpr1 = V_MOV_B32_e32 killed $vgpr3, implicit $exec, implicit $vgpr2, implicit-def $vgpr0_vgpr1, implicit killed $vgpr3
+    S_SETPC_B64_return $sgpr30_sgpr31, implicit $vgpr0, implicit $vgpr1
+...
diff --git a/llvm/test/CodeGen/NVPTX/pow2_mask_cmp.ll b/llvm/test/CodeGen/NVPTX/pow2_mask_cmp.ll
new file mode 100644
index 0000000000000..45701df28a30f
--- /dev/null
+++ b/llvm/test/CodeGen/NVPTX/pow2_mask_cmp.ll
@@ -0,0 +1,19 @@
+; RUN: llc -march=nvptx -verify-machineinstrs < %s | FileCheck %s
+
+; Tests the following pattern:
+; (X & 8) != 0 --> (X & 8) >> 3
+
+; This produces incorrect code when boolean false is represented
+; as a negative one, and this test checks that the transform is
+; not triggered.
+
+; CHECK-LABEL: @pow2_mask_cmp
+; CHECK: and.b32 [[AND:%r[0-9]+]], %r{{[0-9]+}}, 8
+; CHECK: setp.ne.s32 [[SETP:%p[0-9+]]], [[AND]], 0
+; CHECK: selp.u32 %r{{[0-9]+}}, 1, 0, [[SETP]]
+define i32 @pow2_mask_cmp(i32 %x) {
+  %a = and i32 %x, 8
+  %cmp = icmp ne i32 %a, 0
+  %r = zext i1 %cmp to i32
+  ret i32 %r
+}
diff --git a/llvm/test/CodeGen/PowerPC/fold-rlwinm.mir b/llvm/test/CodeGen/PowerPC/fold-rlwinm.mir
index ded5329a46cc9..7a220964b0fd1 100644
--- a/llvm/test/CodeGen/PowerPC/fold-rlwinm.mir
+++ b/llvm/test/CodeGen/PowerPC/fold-rlwinm.mir
@@ -182,3 +182,22 @@ body: |
     STW %2:gprc, %0:gprc, 100
     BLR8 implicit $lr8, implicit $rm
 ...
+---
+name: testFoldRLWINMAndANDI
+tracksRegLiveness: true
+body: |
+  bb.0.entry:
+    liveins: $x3
+    ; CHECK-LABEL: name: testFoldRLWINMAndANDI
+    ; CHECK: liveins: $x3
+    ; CHECK: [[COPY:%[0-9]+]]:g8rc = COPY $x3
+    ; CHECK: [[COPY1:%[0-9]+]]:gprc = COPY [[COPY]].sub_32
+    ; CHECK: [[RLWINM:%[0-9]+]]:gprc = RLWINM [[COPY1]], 4, 28, 31
+    ; CHECK: [[ANDI_rec:%[0-9]+]]:gprc = ANDI_rec [[RLWINM]], 4, implicit-def $cr0
+    ; CHECK: BLR8 implicit $lr8, implicit $rm
+    %0:g8rc = COPY $x3
+    %1:gprc = COPY %0.sub_32:g8rc
+    %2:gprc = RLWINM %1:gprc, 4, 28, 31
+    %3:gprc = ANDI_rec %2:gprc, 4, implicit-def $cr0
+    BLR8 implicit $lr8, implicit $rm
+...
diff --git a/llvm/test/CodeGen/PowerPC/lsr-profitable-chain.ll b/llvm/test/CodeGen/PowerPC/lsr-profitable-chain.ll
new file mode 100644
index 0000000000000..f23c6ba9ca2e0
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/lsr-profitable-chain.ll
@@ -0,0 +1,211 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -verify-machineinstrs -mtriple=powerpc64le-unknown-linux-gnu \
+; RUN:   -mcpu=pwr9 < %s | FileCheck %s
+
+define void @foo(double* readonly %0, double* %1, i64 %2, i64 %3, i64 %4, i64 %5, i64 %6, i64 %7) {
+; CHECK-LABEL: foo:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    cmpd 5, 7
+; CHECK-NEXT:    std 23, -72(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 24, -64(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 25, -56(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 26, -48(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 27, -40(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 28, -32(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 29, -24(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 30, -16(1) # 8-byte Folded Spill
+; CHECK-NEXT:    bge 0, .LBB0_6
+; CHECK-NEXT:  # %bb.1: # %.preheader
+; CHECK-NEXT:    mulld 11, 8, 5
+; CHECK-NEXT:    mulld 12, 9, 8
+; CHECK-NEXT:    sldi 0, 10, 3
+; CHECK-NEXT:    sldi 11, 11, 3
+; CHECK-NEXT:    sldi 12, 12, 3
+; CHECK-NEXT:    add 11, 3, 11
+; CHECK-NEXT:    b .LBB0_3
+; CHECK-NEXT:    .p2align 4
+; CHECK-NEXT:  .LBB0_2:
+; CHECK-NEXT:    add 5, 5, 9
+; CHECK-NEXT:    add 11, 11, 12
+; CHECK-NEXT:    cmpd 5, 7
+; CHECK-NEXT:    bge 0, .LBB0_6
+; CHECK-NEXT:  .LBB0_3: # =>This Loop Header: Depth=1
+; CHECK-NEXT:    # Child Loop BB0_5 Depth 2
+; CHECK-NEXT:    sub 30, 5, 10
+; CHECK-NEXT:    cmpd 6, 30
+; CHECK-NEXT:    bge 0, .LBB0_2
+; CHECK-NEXT:  # %bb.4:
+; CHECK-NEXT:    addi 28, 5, 1
+; CHECK-NEXT:    addi 27, 5, 2
+; CHECK-NEXT:    addi 26, 5, 3
+; CHECK-NEXT:    mulld 29, 5, 8
+; CHECK-NEXT:    sldi 25, 6, 3
+; CHECK-NEXT:    mulld 28, 28, 8
+; CHECK-NEXT:    mulld 27, 27, 8
+; CHECK-NEXT:    mulld 26, 26, 8
+; CHECK-NEXT:    .p2align 5
+; CHECK-NEXT:  .LBB0_5: # Parent Loop BB0_3 Depth=1
+; CHECK-NEXT:    # => This Inner Loop Header: Depth=2
+; CHECK-NEXT:    lfdx 0, 4, 25
+; CHECK-NEXT:    lfdx 1, 11, 25
+; CHECK-NEXT:    add 24, 6, 29
+; CHECK-NEXT:    mr 23, 3
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    sldi 24, 24, 3
+; CHECK-NEXT:    add 24, 3, 24
+; CHECK-NEXT:    lfd 1, 8(24)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 16(24)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 24(24)
+; CHECK-NEXT:    add 24, 6, 28
+; CHECK-NEXT:    sldi 24, 24, 3
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfdux 1, 23, 24
+; CHECK-NEXT:    add 24, 6, 27
+; CHECK-NEXT:    sldi 24, 24, 3
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 8(23)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 16(23)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 24(23)
+; CHECK-NEXT:    mr 23, 3
+; CHECK-NEXT:    lfdux 2, 23, 24
+; CHECK-NEXT:    add 24, 6, 26
+; CHECK-NEXT:    add 6, 6, 10
+; CHECK-NEXT:    sldi 24, 24, 3
+; CHECK-NEXT:    cmpd 6, 30
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 8(23)
+; CHECK-NEXT:    xsadddp 0, 0, 2
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 16(23)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 24(23)
+; CHECK-NEXT:    mr 23, 3
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfdux 1, 23, 24
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 8(23)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 16(23)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    lfd 1, 24(23)
+; CHECK-NEXT:    xsadddp 0, 0, 1
+; CHECK-NEXT:    stfdx 0, 4, 25
+; CHECK-NEXT:    add 25, 25, 0
+; CHECK-NEXT:    blt 0, .LBB0_5
+; CHECK-NEXT:    b .LBB0_2
+; CHECK-NEXT:  .LBB0_6:
+; CHECK-NEXT:    ld 30, -16(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 29, -24(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 28, -32(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 27, -40(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 26, -48(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 25, -56(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 24, -64(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 23, -72(1) # 8-byte Folded Reload
+; CHECK-NEXT:    blr
+  %9 = icmp slt i64 %2, %4
+  br i1 %9, label %10, label %97
+
+10:                                               ; preds = %8, %93
+  %11 = phi i64 [ %95, %93 ], [ %2, %8 ]
+  %12 = phi i64 [ %94, %93 ], [ %3, %8 ]
+  %13 = sub nsw i64 %11, %7
+  %14 = icmp slt i64 %12, %13
+  br i1 %14, label %15, label %93
+
+15:                                               ; preds = %10
+  %16 = mul nsw i64 %11, %5
+  %17 = add nsw i64 %11, 1
+  %18 = mul nsw i64 %17, %5
+  %19 = add nsw i64 %11, 2
+  %20 = mul nsw i64 %19, %5
+  %21 = add nsw i64 %11, 3
+  %22 = mul nsw i64 %21, %5
+  br label %23
+
+23:                                               ; preds = %15, %23
+  %24 = phi i64 [ %12, %15 ], [ %91, %23 ]
+  %25 = getelementptr inbounds double, double* %1, i64 %24
+  %26 = load double, double* %25, align 8
+  %27 = add nsw i64 %24, %16
+  %28 = getelementptr inbounds double, double* %0, i64 %27
+  %29 = load double, double* %28, align 8
+  %30 = fadd double %26, %29
+  %31 = add nsw i64 %27, 1
+  %32 = getelementptr inbounds double, double* %0, i64 %31
+  %33 = load double, double* %32, align 8
+  %34 = fadd double %30, %33
+  %35 = add nsw i64 %27, 2
+  %36 = getelementptr inbounds double, double* %0, i64 %35
+  %37 = load double, double* %36, align 8
+  %38 = fadd double %34, %37
+  %39 = add nsw i64 %27, 3
+  %40 = getelementptr inbounds double, double* %0, i64 %39
+  %41 = load double, double* %40, align 8
+  %42 = fadd double %38, %41
+  %43 = add nsw i64 %24, %18
+  %44 = getelementptr inbounds double, double* %0, i64 %43
+  %45 = load double, double* %44, align 8
+  %46 = fadd double %42, %45
+  %47 = add nsw i64 %43, 1
+  %48 = getelementptr inbounds double, double* %0, i64 %47
+  %49 = load double, double* %48, align 8
+  %50 = fadd double %46, %49
+  %51 = add nsw i64 %43, 2
+  %52 = getelementptr inbounds double, double* %0, i64 %51
+  %53 = load double, double* %52, align 8
+  %54 = fadd double %50, %53
+  %55 = add nsw i64 %43, 3
+  %56 = getelementptr inbounds double, double* %0, i64 %55
+  %57 = load double, double* %56, align 8
+  %58 = fadd double %54, %57
+  %59 = add nsw i64 %24, %20
+  %60 = getelementptr inbounds double, double* %0, i64 %59
+  %61 = load double, double* %60, align 8
+  %62 = fadd double %58, %61
+  %63 = add nsw i64 %59, 1
+  %64 = getelementptr inbounds double, double* %0, i64 %63
+  %65 = load double, double* %64, align 8
+  %66 = fadd double %62, %65
+  %67 = add nsw i64 %59, 2
+  %68 = getelementptr inbounds double, double* %0, i64 %67
+  %69 = load double, double* %68, align 8
+  %70 = fadd double %66, %69
+  %71 = add nsw i64 %59, 3
+  %72 = getelementptr inbounds double, double* %0, i64 %71
+  %73 = load double, double* %72, align 8
+  %74 = fadd double %70, %73
+  %75 = add nsw i64 %24, %22
+  %76 = getelementptr inbounds double, double* %0, i64 %75
+  %77 = load double, double* %76, align 8
+  %78 = fadd double %74, %77
+  %79 = add nsw i64 %75, 1
+  %80 = getelementptr inbounds double, double* %0, i64 %79
+  %81 = load double, double* %80, align 8
+  %82 = fadd double %78, %81
+  %83 = add nsw i64 %75, 2
+  %84 = getelementptr inbounds double, double* %0, i64 %83
+  %85 = load double, double* %84, align 8
+  %86 = fadd double %82, %85
+  %87 = add nsw i64 %75, 3
+  %88 = getelementptr inbounds double, double* %0, i64 %87
+  %89 = load double, double* %88, align 8
+  %90 = fadd double %86, %89
+  store double %90, double* %25, align 8
+  %91 = add nsw i64 %24, %7
+  %92 = icmp slt i64 %91, %13
+  br i1 %92, label %23, label %93
+
+93:                                               ; preds = %23, %10
+  %94 = phi i64 [ %12, %10 ], [ %91, %23 ]
+  %95 = add nsw i64 %11, %6
+  %96 = icmp slt i64 %95, %4
+  br i1 %96, label %10, label %97
+
+97:                                               ; preds = %93, %8
+  ret void
+}
diff --git a/llvm/test/CodeGen/PowerPC/vsx_builtins.ll b/llvm/test/CodeGen/PowerPC/vsx_builtins.ll
index 2ab747384b698..0aae50af26490 100644
--- a/llvm/test/CodeGen/PowerPC/vsx_builtins.ll
+++ b/llvm/test/CodeGen/PowerPC/vsx_builtins.ll
@@ -106,3 +106,37 @@ define i32 @test_vec_test_swsqrts(<4 x float> %a) {
     ret i32 %0
 }
 declare i32 @llvm.ppc.vsx.xvtsqrtsp(<4 x float>)
+
+define i32 @xvtdivdp_andi(<2 x double> %a, <2 x double> %b) {
+; CHECK-LABEL: xvtdivdp_andi:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    xvtdivdp cr0, v2, v3
+; CHECK-NEXT:    li r4, 222
+; CHECK-NEXT:    mfocrf r3, 128
+; CHECK-NEXT:    srwi r3, r3, 28
+; CHECK-NEXT:    andi. r3, r3, 2
+; CHECK-NEXT:    li r3, 22
+; CHECK-NEXT:    iseleq r3, r4, r3
+; CHECK-NEXT:    blr
+  entry:
+    %0 = tail call i32 @llvm.ppc.vsx.xvtdivdp(<2 x double> %a, <2 x double> %b)
+    %1 = and i32 %0, 2
+    %cmp.not = icmp eq i32 %1, 0
+    %retval.0 = select i1 %cmp.not, i32 222, i32 22
+    ret i32 %retval.0
+}
+
+define i32 @xvtdivdp_shift(<2 x double> %a, <2 x double> %b) {
+; CHECK-LABEL: xvtdivdp_shift:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    xvtdivdp cr0, v2, v3
+; CHECK-NEXT:    mfocrf r3, 128
+; CHECK-NEXT:    srwi r3, r3, 28
+; CHECK-NEXT:    rlwinm r3, r3, 28, 31, 31
+; CHECK-NEXT:    blr
+entry:
+  %0 = tail call i32 @llvm.ppc.vsx.xvtdivdp(<2 x double> %a, <2 x double> %b)
+  %1 = lshr i32 %0, 4
+  %.lobit = and i32 %1, 1
+  ret i32 %.lobit
+}
diff --git a/llvm/test/CodeGen/RISCV/imm.ll b/llvm/test/CodeGen/RISCV/imm.ll
index d80dc5f41e445..0b393d464d940 100644
--- a/llvm/test/CodeGen/RISCV/imm.ll
+++ b/llvm/test/CodeGen/RISCV/imm.ll
@@ -105,6 +105,57 @@ define signext i32 @neg_i32_hi20_only() nounwind {
   ret i32 -65536 ; -0x10000
 }
 
+; This can be materialized with ADDI+SLLI, improving compressibility.
+
+define signext i32 @imm_left_shifted_addi() nounwind {
+; RV32I-LABEL: imm_left_shifted_addi:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    lui a0, 32
+; RV32I-NEXT:    addi a0, a0, -64
+; RV32I-NEXT:    ret
+;
+; RV64I-LABEL: imm_left_shifted_addi:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    lui a0, 32
+; RV64I-NEXT:    addiw a0, a0, -64
+; RV64I-NEXT:    ret
+  ret i32 131008 ; 0x1FFC0
+}
+
+; This can be materialized with ADDI+SRLI, improving compressibility.
+
+define signext i32 @imm_right_shifted_addi() nounwind {
+; RV32I-LABEL: imm_right_shifted_addi:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    lui a0, 524288
+; RV32I-NEXT:    addi a0, a0, -1
+; RV32I-NEXT:    ret
+;
+; RV64I-LABEL: imm_right_shifted_addi:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    lui a0, 524288
+; RV64I-NEXT:    addiw a0, a0, -1
+; RV64I-NEXT:    ret
+  ret i32 2147483647 ; 0x7FFFFFFF
+}
+
+; This can be materialized with LUI+SRLI, improving compressibility.
+
+define signext i32 @imm_right_shifted_lui() nounwind {
+; RV32I-LABEL: imm_right_shifted_lui:
+; RV32I:       # %bb.0:
+; RV32I-NEXT:    lui a0, 56
+; RV32I-NEXT:    addi a0, a0, 580
+; RV32I-NEXT:    ret
+;
+; RV64I-LABEL: imm_right_shifted_lui:
+; RV64I:       # %bb.0:
+; RV64I-NEXT:    lui a0, 56
+; RV64I-NEXT:    addiw a0, a0, 580
+; RV64I-NEXT:    ret
+  ret i32 229956 ; 0x38244
+}
+
 define i64 @imm64_1() nounwind {
 ; RV32I-LABEL: imm64_1:
 ; RV32I:       # %bb.0:
diff --git a/llvm/test/CodeGen/SystemZ/fp-strict-conv-08.ll b/llvm/test/CodeGen/SystemZ/fp-strict-conv-08.ll
index a151c15e34d15..c323086b6e135 100644
--- a/llvm/test/CodeGen/SystemZ/fp-strict-conv-08.ll
+++ b/llvm/test/CodeGen/SystemZ/fp-strict-conv-08.ll
@@ -22,7 +22,7 @@ define float @f1(i64 %i) #0 {
 ; Test i64->f64.
 define double @f2(i64 %i) #0 {
 ; CHECK-LABEL: f2:
-; CHECK: ldgr
+; CHECK: cdgbr
 ; CHECK: adbr
 ; CHECK: br %r14
   %conv = call double @llvm.experimental.constrained.uitofp.f64.i64(i64 %i,
diff --git a/llvm/test/CodeGen/SystemZ/frame-26.mir b/llvm/test/CodeGen/SystemZ/frame-26.mir
new file mode 100644
index 0000000000000..e4ff1bd3dcb60
--- /dev/null
+++ b/llvm/test/CodeGen/SystemZ/frame-26.mir
@@ -0,0 +1,199 @@
+# RUN: llc -mtriple=s390x-linux-gnu -start-before=prologepilog %s -o - -print-after=prologepilog \
+# RUN:  -verify-machineinstrs 2>&1 | FileCheck %s
+# REQUIRES: asserts
+#
+# Test that R6 when used for an argument is modelled as being live throughout
+# the function when not saved in the prologue..
+
+# CHECK:      # *** IR Dump After Prologue/Epilogue Insertion & Frame Finalization ***:
+# CHECK-NEXT: # Machine code for function fun0:
+# CHECK-LABEL: bb.0:
+# CHECK: liveins:{{.*}} $r6d
+# CHECK: STMG killed $r7d, killed $r15d
+# CHECK: STG renamable $r6d
+# CHECK: End machine code for function fun0.
+
+
+--- |
+  
+  @g_181 = external dso_local global i32, align 4
+  @g_1390 = external dso_local constant i64*, align 8
+  
+  define internal i8 @fun0(i8 %arg, i8 %arg1, i32 %arg2, i8 %arg3, i32* %arg4, float %F0, float %F1) #0 {
+    ret i8 0
+  }
+  
+  ; Same function but in a single block which will make the verifier complain
+  ; if R6 is killed by the original store before the point where the
+  ; RegScavenger inserts its (killing) store of R6.
+  define internal i8 @fun1(i8 %arg, i8 %arg1, i32 %arg2, i8 %arg3, i32* %arg4) #0 {
+    ret i8 0
+  }
+  
+  attributes #0 = { "frame-pointer"="all" }
+
+...
+---
+name:            fun0
+alignment:       16
+tracksRegLiveness: true
+liveins:
+  - { reg: '$r6d' }
+  - { reg: '$f0s' }
+  - { reg: '$f2s' }
+frameInfo:
+  maxAlignment:    8
+stack:
+  - { id: 0, size: 96, alignment: 8 }
+  - { id: 1, size: 1960, alignment: 8 }
+  - { id: 2, size: 8, alignment: 8 }
+  - { id: 3, size: 320, alignment: 8 }
+  - { id: 4, size: 4, alignment: 4 }
+  - { id: 5, size: 8, alignment: 8 }
+  - { id: 6, size: 8, alignment: 8 }
+  - { id: 7, size: 4, alignment: 4 }
+  - { id: 8, size: 8, alignment: 8 }
+  - { id: 9, size: 4, alignment: 4 }
+  - { id: 10, size: 8, alignment: 8 }
+  - { id: 11, size: 8, alignment: 8 }
+  - { id: 12, size: 8, alignment: 8 }
+  - { id: 13, size: 8, alignment: 8 }
+  - { id: 14, size: 24, alignment: 4 }
+  - { id: 15, size: 4, alignment: 4 }
+  - { id: 16, size: 1792, alignment: 8 }
+  - { id: 17, size: 8, alignment: 8 }
+  - { id: 18, size: 8, alignment: 8 }
+  - { id: 19, size: 8, alignment: 8 }
+  - { id: 20, size: 1, alignment: 2 }
+  - { id: 21, size: 672, alignment: 8 }
+  - { id: 22, size: 4, alignment: 4 }
+  - { id: 23, size: 4, alignment: 4 }
+  - { id: 24, size: 4, alignment: 4 }
+  - { id: 25, size: 64, alignment: 8 }
+machineFunctionInfo: {}
+body:             |
+  bb.0:
+    liveins: $f0s, $f2s, $r6d
+  
+    STG killed renamable $r6d, undef renamable $r1d, 0, $noreg :: (store 8 into `i32** undef`)
+    renamable $r0d = LARL @g_181
+    nofpexcept CEBR renamable $f0s, renamable $f2s, implicit-def $cc, implicit $fpc
+    STG renamable $r0d, undef renamable $r1d, 0, $noreg :: (store 8 into `i32** undef`)
+    BRC 15, 4, %bb.2, implicit killed $cc
+  
+  bb.1:
+    liveins: $f2s, $r0d
+  
+    renamable $f0s = COPY killed renamable $f2s
+  
+  bb.2:
+    liveins: $f0s, $r0d
+  
+    STE killed renamable $f0s, undef renamable $r1d, 0, $noreg :: (volatile store 4 into `float* undef`)
+    renamable $r1d = nuw LA %stack.0, 16, $noreg
+    renamable $r2d = nuw LA %stack.0, 24, $noreg
+    renamable $r3d = LA %stack.0, 40, $noreg
+    renamable $r4d = LARL @g_1390
+    STG renamable $r4d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    renamable $r5d = nuw LA %stack.0, 48, $noreg
+    renamable $r14d = LA %stack.0, 72, $noreg
+    renamable $r13d = LA %stack.0, 80, $noreg
+    renamable $r12d = LA %stack.0, 56, $noreg
+    renamable $r10d = LA %stack.0, 0, $noreg
+    STG renamable $r10d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    renamable $r9d = LA %stack.0, 64, $noreg
+    renamable $r8d = LA %stack.0, 88, $noreg
+    renamable $r7d = nuw LA %stack.0, 8, $noreg
+    MVGHI %stack.1, 904, 0
+    STG killed renamable $r9d, $noreg, 0, $noreg :: (store 8 into `i64*** null`)
+    STG killed renamable $r3d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r14d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r7d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r1d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r4d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r2d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r5d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r8d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r12d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r13d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r10d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    $r2l = LHI 0
+    STG killed renamable $r0d, undef renamable $r1d, 0, $noreg :: (store 8 into `i32** undef`)
+    Return implicit $r2l
+
+...
+---
+name:            fun1
+alignment:       16
+tracksRegLiveness: true
+liveins:
+  - { reg: '$r6d' }
+frameInfo:
+  maxAlignment:    8
+stack:
+  - { id: 0, size: 96, alignment: 8 }
+  - { id: 1, size: 1960, alignment: 8 }
+  - { id: 2, size: 8, alignment: 8 }
+  - { id: 3, size: 320, alignment: 8 }
+  - { id: 4, size: 4, alignment: 4 }
+  - { id: 5, size: 8, alignment: 8 }
+  - { id: 6, size: 8, alignment: 8 }
+  - { id: 7, size: 4, alignment: 4 }
+  - { id: 8, size: 8, alignment: 8 }
+  - { id: 9, size: 4, alignment: 4 }
+  - { id: 10, size: 8, alignment: 8 }
+  - { id: 11, size: 8, alignment: 8 }
+  - { id: 12, size: 8, alignment: 8 }
+  - { id: 13, size: 8, alignment: 8 }
+  - { id: 14, size: 24, alignment: 4 }
+  - { id: 15, size: 4, alignment: 4 }
+  - { id: 16, size: 1792, alignment: 8 }
+  - { id: 17, size: 8, alignment: 8 }
+  - { id: 18, size: 8, alignment: 8 }
+  - { id: 19, size: 8, alignment: 8 }
+  - { id: 20, size: 1, alignment: 2 }
+  - { id: 21, size: 672, alignment: 8 }
+  - { id: 22, size: 4, alignment: 4 }
+  - { id: 23, size: 4, alignment: 4 }
+  - { id: 24, size: 4, alignment: 4 }
+  - { id: 25, size: 64, alignment: 8 }
+machineFunctionInfo: {}
+body:             |
+  bb.0:
+    liveins: $r6d
+  
+    STG killed renamable $r6d, undef renamable $r1d, 0, $noreg :: (store 8 into `i32** undef`)
+    renamable $r0d = LARL @g_181
+    STG renamable $r0d, undef renamable $r1d, 0, $noreg :: (store 8 into `i32** undef`)
+    renamable $r1d = nuw LA %stack.0, 16, $noreg
+    renamable $r2d = nuw LA %stack.0, 24, $noreg
+    renamable $r3d = LA %stack.0, 40, $noreg
+    renamable $r4d = LARL @g_1390
+    STG renamable $r4d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    renamable $r5d = nuw LA %stack.0, 48, $noreg
+    renamable $r14d = LA %stack.0, 72, $noreg
+    renamable $r13d = LA %stack.0, 80, $noreg
+    renamable $r12d = LA %stack.0, 56, $noreg
+    renamable $r10d = LA %stack.0, 0, $noreg
+    STG renamable $r10d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    renamable $r9d = LA %stack.0, 64, $noreg
+    renamable $r8d = LA %stack.0, 88, $noreg
+    renamable $r7d = nuw LA %stack.0, 8, $noreg
+    MVGHI %stack.1, 904, 0
+    STG killed renamable $r9d, $noreg, 0, $noreg :: (store 8 into `i64*** null`)
+    STG killed renamable $r3d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r14d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r7d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r1d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r4d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r2d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r5d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r8d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r12d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r13d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    STG killed renamable $r10d, undef renamable $r1d, 0, $noreg :: (store 8 into `i64*** undef`)
+    $r2l = LHI 0
+    STG killed renamable $r0d, undef renamable $r1d, 0, $noreg :: (store 8 into `i32** undef`)
+    Return implicit $r2l
+
+...
diff --git a/llvm/test/CodeGen/X86/2009-06-05-VariableIndexInsert.ll b/llvm/test/CodeGen/X86/2009-06-05-VariableIndexInsert.ll
index 71a560a63ec52..535450a52ff60 100644
--- a/llvm/test/CodeGen/X86/2009-06-05-VariableIndexInsert.ll
+++ b/llvm/test/CodeGen/X86/2009-06-05-VariableIndexInsert.ll
@@ -1,7 +1,32 @@
-; RUN: llc < %s
-; REQUIRES: default_triple
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=i686-- -mattr=+sse2 | FileCheck %s -check-prefix=X86
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2 | FileCheck %s -check-prefix=X64
 
 define <2 x i64> @_mm_insert_epi16(<2 x i64> %a, i32 %b, i32 %imm) nounwind readnone {
+; X86-LABEL: _mm_insert_epi16:
+; X86:       # %bb.0: # %entry
+; X86-NEXT:    pushl %ebp
+; X86-NEXT:    movl %esp, %ebp
+; X86-NEXT:    andl $-16, %esp
+; X86-NEXT:    subl $32, %esp
+; X86-NEXT:    movzwl 8(%ebp), %eax
+; X86-NEXT:    movl 12(%ebp), %ecx
+; X86-NEXT:    andl $7, %ecx
+; X86-NEXT:    movaps %xmm0, (%esp)
+; X86-NEXT:    movw %ax, (%esp,%ecx,2)
+; X86-NEXT:    movaps (%esp), %xmm0
+; X86-NEXT:    movl %ebp, %esp
+; X86-NEXT:    popl %ebp
+; X86-NEXT:    retl
+;
+; X64-LABEL: _mm_insert_epi16:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    # kill: def $esi killed $esi def $rsi
+; X64-NEXT:    andl $7, %esi
+; X64-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; X64-NEXT:    movw %di, -24(%rsp,%rsi,2)
+; X64-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; X64-NEXT:    retq
 entry:
 	%conv = bitcast <2 x i64> %a to <8 x i16>		; <<8 x i16>> [#uses=1]
 	%conv2 = trunc i32 %b to i16		; <i16> [#uses=1]
diff --git a/llvm/test/CodeGen/X86/fp-intrinsics.ll b/llvm/test/CodeGen/X86/fp-intrinsics.ll
index b8a974f795b2c..7bef1d6780290 100644
--- a/llvm/test/CodeGen/X86/fp-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/fp-intrinsics.ll
@@ -2492,38 +2492,54 @@ define double @uifdl(i64 %x) #0 {
 ;
 ; X86-SSE-LABEL: uifdl:
 ; X86-SSE:       # %bb.0: # %entry
-; X86-SSE-NEXT:    subl $12, %esp
-; X86-SSE-NEXT:    .cfi_def_cfa_offset 16
+; X86-SSE-NEXT:    subl $28, %esp
+; X86-SSE-NEXT:    .cfi_def_cfa_offset 32
+; X86-SSE-NEXT:    movl {{[0-9]+}}(%esp), %eax
 ; X86-SSE-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
-; X86-SSE-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; X86-SSE-NEXT:    subpd {{\.LCPI.*}}, %xmm0
-; X86-SSE-NEXT:    movapd %xmm0, %xmm1
-; X86-SSE-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; X86-SSE-NEXT:    addpd %xmm0, %xmm1
-; X86-SSE-NEXT:    movlpd %xmm1, (%esp)
+; X86-SSE-NEXT:    movlps %xmm0, {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    shrl $31, %eax
+; X86-SSE-NEXT:    fildll {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; X86-SSE-NEXT:    fstpl {{[0-9]+}}(%esp)
+; X86-SSE-NEXT:    wait
+; X86-SSE-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; X86-SSE-NEXT:    movsd %xmm0, (%esp)
 ; X86-SSE-NEXT:    fldl (%esp)
 ; X86-SSE-NEXT:    wait
-; X86-SSE-NEXT:    addl $12, %esp
+; X86-SSE-NEXT:    addl $28, %esp
 ; X86-SSE-NEXT:    .cfi_def_cfa_offset 4
 ; X86-SSE-NEXT:    retl
 ;
 ; SSE-LABEL: uifdl:
 ; SSE:       # %bb.0: # %entry
-; SSE-NEXT:    movq %rdi, %xmm1
-; SSE-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; SSE-NEXT:    subpd {{.*}}(%rip), %xmm1
-; SSE-NEXT:    movapd %xmm1, %xmm0
-; SSE-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-NEXT:    addpd %xmm1, %xmm0
+; SSE-NEXT:    movq %rdi, %rax
+; SSE-NEXT:    shrq %rax
+; SSE-NEXT:    movl %edi, %ecx
+; SSE-NEXT:    andl $1, %ecx
+; SSE-NEXT:    orq %rax, %rcx
+; SSE-NEXT:    testq %rdi, %rdi
+; SSE-NEXT:    cmovnsq %rdi, %rcx
+; SSE-NEXT:    cvtsi2sd %rcx, %xmm0
+; SSE-NEXT:    jns .LBB48_2
+; SSE-NEXT:  # %bb.1:
+; SSE-NEXT:    addsd %xmm0, %xmm0
+; SSE-NEXT:  .LBB48_2: # %entry
 ; SSE-NEXT:    retq
 ;
 ; AVX1-LABEL: uifdl:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vmovq %rdi, %xmm0
-; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; AVX1-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    movq %rdi, %rax
+; AVX1-NEXT:    shrq %rax
+; AVX1-NEXT:    movl %edi, %ecx
+; AVX1-NEXT:    andl $1, %ecx
+; AVX1-NEXT:    orq %rax, %rcx
+; AVX1-NEXT:    testq %rdi, %rdi
+; AVX1-NEXT:    cmovnsq %rdi, %rcx
+; AVX1-NEXT:    vcvtsi2sd %rcx, %xmm0, %xmm0
+; AVX1-NEXT:    jns .LBB48_2
+; AVX1-NEXT:  # %bb.1:
+; AVX1-NEXT:    vaddsd %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:  .LBB48_2: # %entry
 ; AVX1-NEXT:    retq
 ;
 ; AVX512-LABEL: uifdl:
diff --git a/llvm/test/CodeGen/X86/fp-strict-scalar-inttofp.ll b/llvm/test/CodeGen/X86/fp-strict-scalar-inttofp.ll
index 21b09e2d25079..49e238df4aa2a 100644
--- a/llvm/test/CodeGen/X86/fp-strict-scalar-inttofp.ll
+++ b/llvm/test/CodeGen/X86/fp-strict-scalar-inttofp.ll
@@ -1262,14 +1262,17 @@ define double @uitofp_i64tof64(i64 %x) #0 {
 ; SSE-X86-NEXT:    movl %esp, %ebp
 ; SSE-X86-NEXT:    .cfi_def_cfa_register %ebp
 ; SSE-X86-NEXT:    andl $-8, %esp
-; SSE-X86-NEXT:    subl $8, %esp
+; SSE-X86-NEXT:    subl $24, %esp
+; SSE-X86-NEXT:    movl 12(%ebp), %eax
 ; SSE-X86-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
-; SSE-X86-NEXT:    unpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; SSE-X86-NEXT:    subpd {{\.LCPI.*}}, %xmm0
-; SSE-X86-NEXT:    movapd %xmm0, %xmm1
-; SSE-X86-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm0[1]
-; SSE-X86-NEXT:    addpd %xmm0, %xmm1
-; SSE-X86-NEXT:    movlpd %xmm1, (%esp)
+; SSE-X86-NEXT:    movlps %xmm0, {{[0-9]+}}(%esp)
+; SSE-X86-NEXT:    shrl $31, %eax
+; SSE-X86-NEXT:    fildll {{[0-9]+}}(%esp)
+; SSE-X86-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; SSE-X86-NEXT:    fstpl {{[0-9]+}}(%esp)
+; SSE-X86-NEXT:    wait
+; SSE-X86-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; SSE-X86-NEXT:    movsd %xmm0, (%esp)
 ; SSE-X86-NEXT:    fldl (%esp)
 ; SSE-X86-NEXT:    wait
 ; SSE-X86-NEXT:    movl %ebp, %esp
@@ -1279,12 +1282,18 @@ define double @uitofp_i64tof64(i64 %x) #0 {
 ;
 ; SSE-X64-LABEL: uitofp_i64tof64:
 ; SSE-X64:       # %bb.0:
-; SSE-X64-NEXT:    movq %rdi, %xmm1
-; SSE-X64-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; SSE-X64-NEXT:    subpd {{.*}}(%rip), %xmm1
-; SSE-X64-NEXT:    movapd %xmm1, %xmm0
-; SSE-X64-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; SSE-X64-NEXT:    addpd %xmm1, %xmm0
+; SSE-X64-NEXT:    movq %rdi, %rax
+; SSE-X64-NEXT:    shrq %rax
+; SSE-X64-NEXT:    movl %edi, %ecx
+; SSE-X64-NEXT:    andl $1, %ecx
+; SSE-X64-NEXT:    orq %rax, %rcx
+; SSE-X64-NEXT:    testq %rdi, %rdi
+; SSE-X64-NEXT:    cmovnsq %rdi, %rcx
+; SSE-X64-NEXT:    cvtsi2sd %rcx, %xmm0
+; SSE-X64-NEXT:    jns .LBB18_2
+; SSE-X64-NEXT:  # %bb.1:
+; SSE-X64-NEXT:    addsd %xmm0, %xmm0
+; SSE-X64-NEXT:  .LBB18_2:
 ; SSE-X64-NEXT:    retq
 ;
 ; AVX-X86-LABEL: uitofp_i64tof64:
@@ -1295,13 +1304,17 @@ define double @uitofp_i64tof64(i64 %x) #0 {
 ; AVX-X86-NEXT:    movl %esp, %ebp
 ; AVX-X86-NEXT:    .cfi_def_cfa_register %ebp
 ; AVX-X86-NEXT:    andl $-8, %esp
-; AVX-X86-NEXT:    subl $8, %esp
+; AVX-X86-NEXT:    subl $24, %esp
+; AVX-X86-NEXT:    movl 12(%ebp), %eax
 ; AVX-X86-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
-; AVX-X86-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; AVX-X86-NEXT:    vsubpd {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX-X86-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX-X86-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX-X86-NEXT:    vmovlpd %xmm0, (%esp)
+; AVX-X86-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
+; AVX-X86-NEXT:    shrl $31, %eax
+; AVX-X86-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-X86-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-X86-NEXT:    fstpl {{[0-9]+}}(%esp)
+; AVX-X86-NEXT:    wait
+; AVX-X86-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX-X86-NEXT:    vmovsd %xmm0, (%esp)
 ; AVX-X86-NEXT:    fldl (%esp)
 ; AVX-X86-NEXT:    wait
 ; AVX-X86-NEXT:    movl %ebp, %esp
@@ -1311,11 +1324,18 @@ define double @uitofp_i64tof64(i64 %x) #0 {
 ;
 ; AVX1-X64-LABEL: uitofp_i64tof64:
 ; AVX1-X64:       # %bb.0:
-; AVX1-X64-NEXT:    vmovq %rdi, %xmm0
-; AVX1-X64-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; AVX1-X64-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-X64-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX1-X64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX1-X64-NEXT:    movq %rdi, %rax
+; AVX1-X64-NEXT:    shrq %rax
+; AVX1-X64-NEXT:    movl %edi, %ecx
+; AVX1-X64-NEXT:    andl $1, %ecx
+; AVX1-X64-NEXT:    orq %rax, %rcx
+; AVX1-X64-NEXT:    testq %rdi, %rdi
+; AVX1-X64-NEXT:    cmovnsq %rdi, %rcx
+; AVX1-X64-NEXT:    vcvtsi2sd %rcx, %xmm0, %xmm0
+; AVX1-X64-NEXT:    jns .LBB18_2
+; AVX1-X64-NEXT:  # %bb.1:
+; AVX1-X64-NEXT:    vaddsd %xmm0, %xmm0, %xmm0
+; AVX1-X64-NEXT:  .LBB18_2:
 ; AVX1-X64-NEXT:    retq
 ;
 ; AVX512-X64-LABEL: uitofp_i64tof64:
diff --git a/llvm/test/CodeGen/X86/insertelement-var-index.ll b/llvm/test/CodeGen/X86/insertelement-var-index.ll
index 75ed52588d505..9acc259dc3254 100644
--- a/llvm/test/CodeGen/X86/insertelement-var-index.ll
+++ b/llvm/test/CodeGen/X86/insertelement-var-index.ll
@@ -1,7 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.1 | FileCheck %s --check-prefixes=ALL,SSE
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse2   | FileCheck %s --check-prefixes=ALL,SSE,SSE2
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+sse4.1 | FileCheck %s --check-prefixes=ALL,SSE,SSE41
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx    | FileCheck %s --check-prefixes=ALL,AVX,AVX1
 ; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx2   | FileCheck %s --check-prefixes=ALL,AVX,AVX2
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512f,+avx512vl  | FileCheck %s --check-prefixes=ALL,AVX,AVX512,AVX512F
+; RUN: llc < %s -mtriple=x86_64-- -mattr=+avx512bw,+avx512vl | FileCheck %s --check-prefixes=ALL,AVX,AVX512,AVX512BW
 
 define <16 x i8> @undef_index(i8 %x) nounwind {
 ; ALL-LABEL: undef_index:
@@ -19,124 +22,178 @@ define <16 x i8> @undef_scalar(<16 x i8> %x, i32 %index) nounwind {
   ret <16 x i8> %ins
 }
 
-define <16 x i8> @arg_i8_v16i8(i8 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i8_v16i8:
-; SSE:       # %bb.0:
-; SSE-NEXT:    movd %edi, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pshufb %xmm1, %xmm0
-; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i8_v16i8:
+; Insertion into undef vectors
+;
+
+define <16 x i8> @arg_i8_v16i8_undef(i8 %x, i32 %y) nounwind {
+; SSE2-LABEL: arg_i8_v16i8_undef:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movd %edi, %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE2-NEXT:    retq
+;
+; SSE41-LABEL: arg_i8_v16i8_undef:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movd %edi, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: arg_i8_v16i8_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i8_v16i8:
+; AVX2-LABEL: arg_i8_v16i8_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovd %edi, %xmm0
 ; AVX2-NEXT:    vpbroadcastb %xmm0, %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: arg_i8_v16i8_undef:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovd %edi, %xmm0
+; AVX512F-NEXT:    vpbroadcastb %xmm0, %xmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: arg_i8_v16i8_undef:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpbroadcastb %edi, %xmm0
+; AVX512BW-NEXT:    retq
   %ins = insertelement <16 x i8> undef, i8 %x, i32 %y
   ret <16 x i8> %ins
 }
 
-define <8 x i16> @arg_i16_v8i16(i16 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i16_v8i16:
+define <8 x i16> @arg_i16_v8i16_undef(i16 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i16_v8i16_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movd %edi, %xmm0
 ; SSE-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i16_v8i16:
+; AVX1-LABEL: arg_i16_v8i16_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i16_v8i16:
+; AVX2-LABEL: arg_i16_v8i16_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovd %edi, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: arg_i16_v8i16_undef:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovd %edi, %xmm0
+; AVX512F-NEXT:    vpbroadcastw %xmm0, %xmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: arg_i16_v8i16_undef:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpbroadcastw %edi, %xmm0
+; AVX512BW-NEXT:    retq
   %ins = insertelement <8 x i16> undef, i16 %x, i32 %y
   ret <8 x i16> %ins
 }
 
-define <4 x i32> @arg_i32_v4i32(i32 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i32_v4i32:
+define <4 x i32> @arg_i32_v4i32_undef(i32 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i32_v4i32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movd %edi, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i32_v4i32:
+; AVX1-LABEL: arg_i32_v4i32_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i32_v4i32:
+; AVX2-LABEL: arg_i32_v4i32_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovd %edi, %xmm0
 ; AVX2-NEXT:    vpbroadcastd %xmm0, %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_i32_v4i32_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastd %edi, %xmm0
+; AVX512-NEXT:    retq
   %ins = insertelement <4 x i32> undef, i32 %x, i32 %y
   ret <4 x i32> %ins
 }
 
-define <2 x i64> @arg_i64_v2i64(i64 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i64_v2i64:
+define <2 x i64> @arg_i64_v2i64_undef(i64 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i64_v2i64_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq %rdi, %xmm0
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i64_v2i64:
+; AVX1-LABEL: arg_i64_v2i64_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovq %rdi, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i64_v2i64:
+; AVX2-LABEL: arg_i64_v2i64_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovq %rdi, %xmm0
 ; AVX2-NEXT:    vpbroadcastq %xmm0, %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_i64_v2i64_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastq %rdi, %xmm0
+; AVX512-NEXT:    retq
   %ins = insertelement <2 x i64> undef, i64 %x, i32 %y
   ret <2 x i64> %ins
 }
 
-define <4 x float> @arg_f32_v4f32(float %x, i32 %y) nounwind {
-; SSE-LABEL: arg_f32_v4f32:
+define <4 x float> @arg_f32_v4f32_undef(float %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f32_v4f32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_f32_v4f32:
+; AVX1-LABEL: arg_f32_v4f32_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_f32_v4f32:
+; AVX2-LABEL: arg_f32_v4f32_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vbroadcastss %xmm0, %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_f32_v4f32_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vbroadcastss %xmm0, %xmm0
+; AVX512-NEXT:    retq
   %ins = insertelement <4 x float> undef, float %x, i32 %y
   ret <4 x float> %ins
 }
 
-define <2 x double> @arg_f64_v2f64(double %x, i32 %y) nounwind {
-; SSE-LABEL: arg_f64_v2f64:
-; SSE:       # %bb.0:
-; SSE-NEXT:    movddup {{.*#+}} xmm0 = xmm0[0,0]
-; SSE-NEXT:    retq
+define <2 x double> @arg_f64_v2f64_undef(double %x, i32 %y) nounwind {
+; SSE2-LABEL: arg_f64_v2f64_undef:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0,0]
+; SSE2-NEXT:    retq
 ;
-; AVX-LABEL: arg_f64_v2f64:
+; SSE41-LABEL: arg_f64_v2f64_undef:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movddup {{.*#+}} xmm0 = xmm0[0,0]
+; SSE41-NEXT:    retq
+;
+; AVX-LABEL: arg_f64_v2f64_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX-NEXT:    retq
@@ -144,16 +201,25 @@ define <2 x double> @arg_f64_v2f64(double %x, i32 %y) nounwind {
   ret <2 x double> %ins
 }
 
-define <16 x i8> @load_i8_v16i8(i8* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i8_v16i8:
-; SSE:       # %bb.0:
-; SSE-NEXT:    movzbl (%rdi), %eax
-; SSE-NEXT:    movd %eax, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pshufb %xmm1, %xmm0
-; SSE-NEXT:    retq
+define <16 x i8> @load_i8_v16i8_undef(i8* %p, i32 %y) nounwind {
+; SSE2-LABEL: load_i8_v16i8_undef:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movzbl (%rdi), %eax
+; SSE2-NEXT:    movd %eax, %xmm0
+; SSE2-NEXT:    punpcklbw {{.*#+}} xmm0 = xmm0[0,0,1,1,2,2,3,3,4,4,5,5,6,6,7,7]
+; SSE2-NEXT:    pshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
+; SSE2-NEXT:    retq
 ;
-; AVX1-LABEL: load_i8_v16i8:
+; SSE41-LABEL: load_i8_v16i8_undef:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movzbl (%rdi), %eax
+; SSE41-NEXT:    movd %eax, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pshufb %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: load_i8_v16i8_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    movzbl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
@@ -161,17 +227,22 @@ define <16 x i8> @load_i8_v16i8(i8* %p, i32 %y) nounwind {
 ; AVX1-NEXT:    vpshufb %xmm1, %xmm0, %xmm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: load_i8_v16i8:
+; AVX2-LABEL: load_i8_v16i8_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastb (%rdi), %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: load_i8_v16i8_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastb (%rdi), %xmm0
+; AVX512-NEXT:    retq
   %x = load i8, i8* %p
   %ins = insertelement <16 x i8> undef, i8 %x, i32 %y
   ret <16 x i8> %ins
 }
 
-define <8 x i16> @load_i16_v8i16(i16* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i16_v8i16:
+define <8 x i16> @load_i16_v8i16_undef(i16* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i16_v8i16_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movzwl (%rdi), %eax
 ; SSE-NEXT:    movd %eax, %xmm0
@@ -179,7 +250,7 @@ define <8 x i16> @load_i16_v8i16(i16* %p, i32 %y) nounwind {
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: load_i16_v8i16:
+; AVX1-LABEL: load_i16_v8i16_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    movzwl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
@@ -187,23 +258,28 @@ define <8 x i16> @load_i16_v8i16(i16* %p, i32 %y) nounwind {
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: load_i16_v8i16:
+; AVX2-LABEL: load_i16_v8i16_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastw (%rdi), %xmm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: load_i16_v8i16_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastw (%rdi), %xmm0
+; AVX512-NEXT:    retq
   %x = load i16, i16* %p
   %ins = insertelement <8 x i16> undef, i16 %x, i32 %y
   ret <8 x i16> %ins
 }
 
-define <4 x i32> @load_i32_v4i32(i32* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i32_v4i32:
+define <4 x i32> @load_i32_v4i32_undef(i32* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i32_v4i32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movd {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_i32_v4i32:
+; AVX-LABEL: load_i32_v4i32_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastss (%rdi), %xmm0
 ; AVX-NEXT:    retq
@@ -212,14 +288,14 @@ define <4 x i32> @load_i32_v4i32(i32* %p, i32 %y) nounwind {
   ret <4 x i32> %ins
 }
 
-define <2 x i64> @load_i64_v2i64(i64* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i64_v2i64:
+define <2 x i64> @load_i64_v2i64_undef(i64* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i64_v2i64_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movq {{.*#+}} xmm0 = mem[0],zero
 ; SSE-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_i64_v2i64:
+; AVX-LABEL: load_i64_v2i64_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX-NEXT:    retq
@@ -228,14 +304,14 @@ define <2 x i64> @load_i64_v2i64(i64* %p, i32 %y) nounwind {
   ret <2 x i64> %ins
 }
 
-define <4 x float> @load_f32_v4f32(float* %p, i32 %y) nounwind {
-; SSE-LABEL: load_f32_v4f32:
+define <4 x float> @load_f32_v4f32_undef(float* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f32_v4f32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
 ; SSE-NEXT:    shufps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_f32_v4f32:
+; AVX-LABEL: load_f32_v4f32_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastss (%rdi), %xmm0
 ; AVX-NEXT:    retq
@@ -244,13 +320,19 @@ define <4 x float> @load_f32_v4f32(float* %p, i32 %y) nounwind {
   ret <4 x float> %ins
 }
 
-define <2 x double> @load_f64_v2f64(double* %p, i32 %y) nounwind {
-; SSE-LABEL: load_f64_v2f64:
-; SSE:       # %bb.0:
-; SSE-NEXT:    movddup {{.*#+}} xmm0 = mem[0,0]
-; SSE-NEXT:    retq
+define <2 x double> @load_f64_v2f64_undef(double* %p, i32 %y) nounwind {
+; SSE2-LABEL: load_f64_v2f64_undef:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; SSE2-NEXT:    movlhps {{.*#+}} xmm0 = xmm0[0,0]
+; SSE2-NEXT:    retq
 ;
-; AVX-LABEL: load_f64_v2f64:
+; SSE41-LABEL: load_f64_v2f64_undef:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movddup {{.*#+}} xmm0 = mem[0,0]
+; SSE41-NEXT:    retq
+;
+; AVX-LABEL: load_f64_v2f64_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vmovddup {{.*#+}} xmm0 = mem[0,0]
 ; AVX-NEXT:    retq
@@ -259,8 +341,8 @@ define <2 x double> @load_f64_v2f64(double* %p, i32 %y) nounwind {
   ret <2 x double> %ins
 }
 
-define <32 x i8> @arg_i8_v32i8(i8 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i8_v32i8:
+define <32 x i8> @arg_i8_v32i8_undef(i8 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i8_v32i8_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    andl $31, %esi
@@ -269,7 +351,7 @@ define <32 x i8> @arg_i8_v32i8(i8 %x, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i8_v32i8:
+; AVX1-LABEL: arg_i8_v32i8_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
@@ -277,17 +359,28 @@ define <32 x i8> @arg_i8_v32i8(i8 %x, i32 %y) nounwind {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i8_v32i8:
+; AVX2-LABEL: arg_i8_v32i8_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovd %edi, %xmm0
 ; AVX2-NEXT:    vpbroadcastb %xmm0, %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: arg_i8_v32i8_undef:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovd %edi, %xmm0
+; AVX512F-NEXT:    vpbroadcastb %xmm0, %ymm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: arg_i8_v32i8_undef:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpbroadcastb %edi, %ymm0
+; AVX512BW-NEXT:    retq
   %ins = insertelement <32 x i8> undef, i8 %x, i32 %y
   ret <32 x i8> %ins
 }
 
-define <16 x i16> @arg_i16_v16i16(i16 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i16_v16i16:
+define <16 x i16> @arg_i16_v16i16_undef(i16 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i16_v16i16_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    andl $15, %esi
@@ -296,7 +389,7 @@ define <16 x i16> @arg_i16_v16i16(i16 %x, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i16_v16i16:
+; AVX1-LABEL: arg_i16_v16i16_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpshuflw {{.*#+}} xmm0 = xmm0[0,0,0,0,4,5,6,7]
@@ -304,17 +397,28 @@ define <16 x i16> @arg_i16_v16i16(i16 %x, i32 %y) nounwind {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i16_v16i16:
+; AVX2-LABEL: arg_i16_v16i16_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovd %edi, %xmm0
 ; AVX2-NEXT:    vpbroadcastw %xmm0, %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512F-LABEL: arg_i16_v16i16_undef:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovd %edi, %xmm0
+; AVX512F-NEXT:    vpbroadcastw %xmm0, %ymm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: arg_i16_v16i16_undef:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpbroadcastw %edi, %ymm0
+; AVX512BW-NEXT:    retq
   %ins = insertelement <16 x i16> undef, i16 %x, i32 %y
   ret <16 x i16> %ins
 }
 
-define <8 x i32> @arg_i32_v8i32(i32 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i32_v8i32:
+define <8 x i32> @arg_i32_v8i32_undef(i32 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i32_v8i32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    andl $7, %esi
@@ -323,24 +427,29 @@ define <8 x i32> @arg_i32_v8i32(i32 %x, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i32_v8i32:
+; AVX1-LABEL: arg_i32_v8i32_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovd %edi, %xmm0
 ; AVX1-NEXT:    vpshufd {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i32_v8i32:
+; AVX2-LABEL: arg_i32_v8i32_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovd %edi, %xmm0
 ; AVX2-NEXT:    vpbroadcastd %xmm0, %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_i32_v8i32_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastd %edi, %ymm0
+; AVX512-NEXT:    retq
   %ins = insertelement <8 x i32> undef, i32 %x, i32 %y
   ret <8 x i32> %ins
 }
 
-define <4 x i64> @arg_i64_v4i64(i64 %x, i32 %y) nounwind {
-; SSE-LABEL: arg_i64_v4i64:
+define <4 x i64> @arg_i64_v4i64_undef(i64 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i64_v4i64_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    andl $3, %esi
@@ -349,24 +458,29 @@ define <4 x i64> @arg_i64_v4i64(i64 %x, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_i64_v4i64:
+; AVX1-LABEL: arg_i64_v4i64_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovq %rdi, %xmm0
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_i64_v4i64:
+; AVX2-LABEL: arg_i64_v4i64_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vmovq %rdi, %xmm0
 ; AVX2-NEXT:    vpbroadcastq %xmm0, %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_i64_v4i64_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastq %rdi, %ymm0
+; AVX512-NEXT:    retq
   %ins = insertelement <4 x i64> undef, i64 %x, i32 %y
   ret <4 x i64> %ins
 }
 
-define <8 x float> @arg_f32_v8f32(float %x, i32 %y) nounwind {
-; SSE-LABEL: arg_f32_v8f32:
+define <8 x float> @arg_f32_v8f32_undef(float %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f32_v8f32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
 ; SSE-NEXT:    andl $7, %edi
@@ -375,22 +489,27 @@ define <8 x float> @arg_f32_v8f32(float %x, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_f32_v8f32:
+; AVX1-LABEL: arg_f32_v8f32_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vpermilps {{.*#+}} xmm0 = xmm0[0,0,0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_f32_v8f32:
+; AVX2-LABEL: arg_f32_v8f32_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vbroadcastss %xmm0, %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_f32_v8f32_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vbroadcastss %xmm0, %ymm0
+; AVX512-NEXT:    retq
   %ins = insertelement <8 x float> undef, float %x, i32 %y
   ret <8 x float> %ins
 }
 
-define <4 x double> @arg_f64_v4f64(double %x, i32 %y) nounwind {
-; SSE-LABEL: arg_f64_v4f64:
+define <4 x double> @arg_f64_v4f64_undef(double %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f64_v4f64_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
 ; SSE-NEXT:    andl $3, %edi
@@ -399,22 +518,27 @@ define <4 x double> @arg_f64_v4f64(double %x, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: arg_f64_v4f64:
+; AVX1-LABEL: arg_f64_v4f64_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    vmovddup {{.*#+}} xmm0 = xmm0[0,0]
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: arg_f64_v4f64:
+; AVX2-LABEL: arg_f64_v4f64_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vbroadcastsd %xmm0, %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: arg_f64_v4f64_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vbroadcastsd %xmm0, %ymm0
+; AVX512-NEXT:    retq
   %ins = insertelement <4 x double> undef, double %x, i32 %y
   ret <4 x double> %ins
 }
 
-define <32 x i8> @load_i8_v32i8(i8* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i8_v32i8:
+define <32 x i8> @load_i8_v32i8_undef(i8* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i8_v32i8_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    movb (%rdi), %al
@@ -424,7 +548,7 @@ define <32 x i8> @load_i8_v32i8(i8* %p, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: load_i8_v32i8:
+; AVX1-LABEL: load_i8_v32i8_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    movzbl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
@@ -433,17 +557,22 @@ define <32 x i8> @load_i8_v32i8(i8* %p, i32 %y) nounwind {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: load_i8_v32i8:
+; AVX2-LABEL: load_i8_v32i8_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastb (%rdi), %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: load_i8_v32i8_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastb (%rdi), %ymm0
+; AVX512-NEXT:    retq
   %x = load i8, i8* %p
   %ins = insertelement <32 x i8> undef, i8 %x, i32 %y
   ret <32 x i8> %ins
 }
 
-define <16 x i16> @load_i16_v16i16(i16* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i16_v16i16:
+define <16 x i16> @load_i16_v16i16_undef(i16* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i16_v16i16_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    movzwl (%rdi), %eax
@@ -453,7 +582,7 @@ define <16 x i16> @load_i16_v16i16(i16* %p, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX1-LABEL: load_i16_v16i16:
+; AVX1-LABEL: load_i16_v16i16_undef:
 ; AVX1:       # %bb.0:
 ; AVX1-NEXT:    movzwl (%rdi), %eax
 ; AVX1-NEXT:    vmovd %eax, %xmm0
@@ -462,17 +591,22 @@ define <16 x i16> @load_i16_v16i16(i16* %p, i32 %y) nounwind {
 ; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
-; AVX2-LABEL: load_i16_v16i16:
+; AVX2-LABEL: load_i16_v16i16_undef:
 ; AVX2:       # %bb.0:
 ; AVX2-NEXT:    vpbroadcastw (%rdi), %ymm0
 ; AVX2-NEXT:    retq
+;
+; AVX512-LABEL: load_i16_v16i16_undef:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vpbroadcastw (%rdi), %ymm0
+; AVX512-NEXT:    retq
   %x = load i16, i16* %p
   %ins = insertelement <16 x i16> undef, i16 %x, i32 %y
   ret <16 x i16> %ins
 }
 
-define <8 x i32> @load_i32_v8i32(i32* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i32_v8i32:
+define <8 x i32> @load_i32_v8i32_undef(i32* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i32_v8i32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    movl (%rdi), %eax
@@ -482,7 +616,7 @@ define <8 x i32> @load_i32_v8i32(i32* %p, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_i32_v8i32:
+; AVX-LABEL: load_i32_v8i32_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastss (%rdi), %ymm0
 ; AVX-NEXT:    retq
@@ -491,8 +625,8 @@ define <8 x i32> @load_i32_v8i32(i32* %p, i32 %y) nounwind {
   ret <8 x i32> %ins
 }
 
-define <4 x i64> @load_i64_v4i64(i64* %p, i32 %y) nounwind {
-; SSE-LABEL: load_i64_v4i64:
+define <4 x i64> @load_i64_v4i64_undef(i64* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i64_v4i64_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    movq (%rdi), %rax
@@ -502,7 +636,7 @@ define <4 x i64> @load_i64_v4i64(i64* %p, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_i64_v4i64:
+; AVX-LABEL: load_i64_v4i64_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastsd (%rdi), %ymm0
 ; AVX-NEXT:    retq
@@ -511,8 +645,8 @@ define <4 x i64> @load_i64_v4i64(i64* %p, i32 %y) nounwind {
   ret <4 x i64> %ins
 }
 
-define <8 x float> @load_f32_v8f32(float* %p, i32 %y) nounwind {
-; SSE-LABEL: load_f32_v8f32:
+define <8 x float> @load_f32_v8f32_undef(float* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f32_v8f32_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    movss {{.*#+}} xmm0 = mem[0],zero,zero,zero
@@ -522,7 +656,7 @@ define <8 x float> @load_f32_v8f32(float* %p, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_f32_v8f32:
+; AVX-LABEL: load_f32_v8f32_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastss (%rdi), %ymm0
 ; AVX-NEXT:    retq
@@ -531,8 +665,8 @@ define <8 x float> @load_f32_v8f32(float* %p, i32 %y) nounwind {
   ret <8 x float> %ins
 }
 
-define <4 x double> @load_f64_v4f64(double* %p, i32 %y) nounwind {
-; SSE-LABEL: load_f64_v4f64:
+define <4 x double> @load_f64_v4f64_undef(double* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f64_v4f64_undef:
 ; SSE:       # %bb.0:
 ; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
 ; SSE-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
@@ -542,7 +676,7 @@ define <4 x double> @load_f64_v4f64(double* %p, i32 %y) nounwind {
 ; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
 ; SSE-NEXT:    retq
 ;
-; AVX-LABEL: load_f64_v4f64:
+; AVX-LABEL: load_f64_v4f64_undef:
 ; AVX:       # %bb.0:
 ; AVX-NEXT:    vbroadcastsd (%rdi), %ymm0
 ; AVX-NEXT:    retq
@@ -551,6 +685,670 @@ define <4 x double> @load_f64_v4f64(double* %p, i32 %y) nounwind {
   ret <4 x double> %ins
 }
 
+;
+; Insertion into arg vectors
+;
+
+define <16 x i8> @arg_i8_v16i8(<16 x i8> %v, i8 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i8_v16i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $15, %esi
+; SSE-NEXT:    movb %dil, -24(%rsp,%rsi)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i8_v16i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $15, %esi
+; AVX-NEXT:    movb %dil, -24(%rsp,%rsi)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %ins = insertelement <16 x i8> %v, i8 %x, i32 %y
+  ret <16 x i8> %ins
+}
+
+define <8 x i16> @arg_i16_v8i16(<8 x i16> %v, i16 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i16_v8i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $7, %esi
+; SSE-NEXT:    movw %di, -24(%rsp,%rsi,2)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i16_v8i16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $7, %esi
+; AVX-NEXT:    movw %di, -24(%rsp,%rsi,2)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %ins = insertelement <8 x i16> %v, i16 %x, i32 %y
+  ret <8 x i16> %ins
+}
+
+define <4 x i32> @arg_i32_v4i32(<4 x i32> %v, i32 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i32_v4i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %esi
+; SSE-NEXT:    movl %edi, -24(%rsp,%rsi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i32_v4i32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $3, %esi
+; AVX-NEXT:    movl %edi, -24(%rsp,%rsi,4)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %ins = insertelement <4 x i32> %v, i32 %x, i32 %y
+  ret <4 x i32> %ins
+}
+
+define <2 x i64> @arg_i64_v2i64(<2 x i64> %v, i64 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i64_v2i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $1, %esi
+; SSE-NEXT:    movq %rdi, -24(%rsp,%rsi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i64_v2i64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $1, %esi
+; AVX-NEXT:    movq %rdi, -24(%rsp,%rsi,8)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %ins = insertelement <2 x i64> %v, i64 %x, i32 %y
+  ret <2 x i64> %ins
+}
+
+define <4 x float> @arg_f32_v4f32(<4 x float> %v, float %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f32_v4f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %edi
+; SSE-NEXT:    movss %xmm1, -24(%rsp,%rdi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_f32_v4f32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $edi killed $edi def $rdi
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $3, %edi
+; AVX-NEXT:    vmovss %xmm1, -24(%rsp,%rdi,4)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %ins = insertelement <4 x float> %v, float %x, i32 %y
+  ret <4 x float> %ins
+}
+
+define <2 x double> @arg_f64_v2f64(<2 x double> %v, double %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f64_v2f64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $1, %edi
+; SSE-NEXT:    movsd %xmm1, -24(%rsp,%rdi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_f64_v2f64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $edi killed $edi def $rdi
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $1, %edi
+; AVX-NEXT:    vmovsd %xmm1, -24(%rsp,%rdi,8)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %ins = insertelement <2 x double> %v, double %x, i32 %y
+  ret <2 x double> %ins
+}
+
+define <16 x i8> @load_i8_v16i8(<16 x i8> %v, i8* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i8_v16i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movb (%rdi), %al
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $15, %esi
+; SSE-NEXT:    movb %al, -24(%rsp,%rsi)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i8_v16i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movb (%rdi), %al
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $15, %esi
+; AVX-NEXT:    movb %al, -24(%rsp,%rsi)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %x = load i8, i8* %p
+  %ins = insertelement <16 x i8> %v, i8 %x, i32 %y
+  ret <16 x i8> %ins
+}
+
+define <8 x i16> @load_i16_v8i16(<8 x i16> %v, i16* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i16_v8i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movzwl (%rdi), %eax
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $7, %esi
+; SSE-NEXT:    movw %ax, -24(%rsp,%rsi,2)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i16_v8i16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movzwl (%rdi), %eax
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $7, %esi
+; AVX-NEXT:    movw %ax, -24(%rsp,%rsi,2)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %x = load i16, i16* %p
+  %ins = insertelement <8 x i16> %v, i16 %x, i32 %y
+  ret <8 x i16> %ins
+}
+
+define <4 x i32> @load_i32_v4i32(<4 x i32> %v, i32* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i32_v4i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movl (%rdi), %eax
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %esi
+; SSE-NEXT:    movl %eax, -24(%rsp,%rsi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i32_v4i32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movl (%rdi), %eax
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $3, %esi
+; AVX-NEXT:    movl %eax, -24(%rsp,%rsi,4)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %x = load i32, i32* %p
+  %ins = insertelement <4 x i32> %v, i32 %x, i32 %y
+  ret <4 x i32> %ins
+}
+
+define <2 x i64> @load_i64_v2i64(<2 x i64> %v, i64* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i64_v2i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movq (%rdi), %rax
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $1, %esi
+; SSE-NEXT:    movq %rax, -24(%rsp,%rsi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i64_v2i64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movq (%rdi), %rax
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $1, %esi
+; AVX-NEXT:    movq %rax, -24(%rsp,%rsi,8)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %x = load i64, i64* %p
+  %ins = insertelement <2 x i64> %v, i64 %x, i32 %y
+  ret <2 x i64> %ins
+}
+
+define <4 x float> @load_f32_v4f32(<4 x float> %v, float* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f32_v4f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %esi
+; SSE-NEXT:    movss %xmm1, -24(%rsp,%rsi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_f32_v4f32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $3, %esi
+; AVX-NEXT:    vmovss %xmm1, -24(%rsp,%rsi,4)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %x = load float, float* %p
+  %ins = insertelement <4 x float> %v, float %x, i32 %y
+  ret <4 x float> %ins
+}
+
+define <2 x double> @load_f64_v2f64(<2 x double> %v, double* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f64_v2f64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movsd {{.*#+}} xmm1 = mem[0],zero
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $1, %esi
+; SSE-NEXT:    movsd %xmm1, -24(%rsp,%rsi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_f64_v2f64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX-NEXT:    vmovaps %xmm0, -{{[0-9]+}}(%rsp)
+; AVX-NEXT:    andl $1, %esi
+; AVX-NEXT:    vmovsd %xmm1, -24(%rsp,%rsi,8)
+; AVX-NEXT:    vmovaps -{{[0-9]+}}(%rsp), %xmm0
+; AVX-NEXT:    retq
+  %x = load double, double* %p
+  %ins = insertelement <2 x double> %v, double %x, i32 %y
+  ret <2 x double> %ins
+}
+
+define <32 x i8> @arg_i8_v32i8(<32 x i8> %v, i8 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i8_v32i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $31, %esi
+; SSE-NEXT:    movb %dil, -40(%rsp,%rsi)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i8_v32i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $31, %esi
+; AVX-NEXT:    movb %dil, (%rsp,%rsi)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %ins = insertelement <32 x i8> %v, i8 %x, i32 %y
+  ret <32 x i8> %ins
+}
+
+define <16 x i16> @arg_i16_v16i16(<16 x i16> %v, i16 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i16_v16i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $15, %esi
+; SSE-NEXT:    movw %di, -40(%rsp,%rsi,2)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i16_v16i16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $15, %esi
+; AVX-NEXT:    movw %di, (%rsp,%rsi,2)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %ins = insertelement <16 x i16> %v, i16 %x, i32 %y
+  ret <16 x i16> %ins
+}
+
+define <8 x i32> @arg_i32_v8i32(<8 x i32> %v, i32 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i32_v8i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $7, %esi
+; SSE-NEXT:    movl %edi, -40(%rsp,%rsi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i32_v8i32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $7, %esi
+; AVX-NEXT:    movl %edi, (%rsp,%rsi,4)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %ins = insertelement <8 x i32> %v, i32 %x, i32 %y
+  ret <8 x i32> %ins
+}
+
+define <4 x i64> @arg_i64_v4i64(<4 x i64> %v, i64 %x, i32 %y) nounwind {
+; SSE-LABEL: arg_i64_v4i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %esi
+; SSE-NEXT:    movq %rdi, -40(%rsp,%rsi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_i64_v4i64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $3, %esi
+; AVX-NEXT:    movq %rdi, (%rsp,%rsi,8)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %ins = insertelement <4 x i64> %v, i64 %x, i32 %y
+  ret <4 x i64> %ins
+}
+
+define <8 x float> @arg_f32_v8f32(<8 x float> %v, float %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f32_v8f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $7, %edi
+; SSE-NEXT:    movss %xmm2, -40(%rsp,%rdi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_f32_v8f32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $edi killed $edi def $rdi
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $7, %edi
+; AVX-NEXT:    vmovss %xmm1, (%rsp,%rdi,4)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %ins = insertelement <8 x float> %v, float %x, i32 %y
+  ret <8 x float> %ins
+}
+
+define <4 x double> @arg_f64_v4f64(<4 x double> %v, double %x, i32 %y) nounwind {
+; SSE-LABEL: arg_f64_v4f64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $edi killed $edi def $rdi
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %edi
+; SSE-NEXT:    movsd %xmm2, -40(%rsp,%rdi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: arg_f64_v4f64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $edi killed $edi def $rdi
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $3, %edi
+; AVX-NEXT:    vmovsd %xmm1, (%rsp,%rdi,8)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %ins = insertelement <4 x double> %v, double %x, i32 %y
+  ret <4 x double> %ins
+}
+
+define <32 x i8> @load_i8_v32i8(<32 x i8> %v, i8* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i8_v32i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movb (%rdi), %al
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $31, %esi
+; SSE-NEXT:    movb %al, -40(%rsp,%rsi)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i8_v32i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movb (%rdi), %al
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $31, %esi
+; AVX-NEXT:    movb %al, (%rsp,%rsi)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %x = load i8, i8* %p
+  %ins = insertelement <32 x i8> %v, i8 %x, i32 %y
+  ret <32 x i8> %ins
+}
+
+define <16 x i16> @load_i16_v16i16(<16 x i16> %v, i16* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i16_v16i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movzwl (%rdi), %eax
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $15, %esi
+; SSE-NEXT:    movw %ax, -40(%rsp,%rsi,2)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i16_v16i16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movzwl (%rdi), %eax
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $15, %esi
+; AVX-NEXT:    movw %ax, (%rsp,%rsi,2)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %x = load i16, i16* %p
+  %ins = insertelement <16 x i16> %v, i16 %x, i32 %y
+  ret <16 x i16> %ins
+}
+
+define <8 x i32> @load_i32_v8i32(<8 x i32> %v, i32* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i32_v8i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movl (%rdi), %eax
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $7, %esi
+; SSE-NEXT:    movl %eax, -40(%rsp,%rsi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i32_v8i32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movl (%rdi), %eax
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $7, %esi
+; AVX-NEXT:    movl %eax, (%rsp,%rsi,4)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %x = load i32, i32* %p
+  %ins = insertelement <8 x i32> %v, i32 %x, i32 %y
+  ret <8 x i32> %ins
+}
+
+define <4 x i64> @load_i64_v4i64(<4 x i64> %v, i64* %p, i32 %y) nounwind {
+; SSE-LABEL: load_i64_v4i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movq (%rdi), %rax
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %esi
+; SSE-NEXT:    movq %rax, -40(%rsp,%rsi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_i64_v4i64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    movq (%rdi), %rax
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $3, %esi
+; AVX-NEXT:    movq %rax, (%rsp,%rsi,8)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %x = load i64, i64* %p
+  %ins = insertelement <4 x i64> %v, i64 %x, i32 %y
+  ret <4 x i64> %ins
+}
+
+define <8 x float> @load_f32_v8f32(<8 x float> %v, float* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f32_v8f32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movss {{.*#+}} xmm2 = mem[0],zero,zero,zero
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $7, %esi
+; SSE-NEXT:    movss %xmm2, -40(%rsp,%rsi,4)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_f32_v8f32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovss {{.*#+}} xmm1 = mem[0],zero,zero,zero
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $7, %esi
+; AVX-NEXT:    vmovss %xmm1, (%rsp,%rsi,4)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %x = load float, float* %p
+  %ins = insertelement <8 x float> %v, float %x, i32 %y
+  ret <8 x float> %ins
+}
+
+define <4 x double> @load_f64_v4f64(<4 x double> %v, double* %p, i32 %y) nounwind {
+; SSE-LABEL: load_f64_v4f64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    # kill: def $esi killed $esi def $rsi
+; SSE-NEXT:    movsd {{.*#+}} xmm2 = mem[0],zero
+; SSE-NEXT:    movaps %xmm1, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    movaps %xmm0, -{{[0-9]+}}(%rsp)
+; SSE-NEXT:    andl $3, %esi
+; SSE-NEXT:    movsd %xmm2, -40(%rsp,%rsi,8)
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm0
+; SSE-NEXT:    movaps -{{[0-9]+}}(%rsp), %xmm1
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: load_f64_v4f64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    pushq %rbp
+; AVX-NEXT:    movq %rsp, %rbp
+; AVX-NEXT:    andq $-32, %rsp
+; AVX-NEXT:    subq $64, %rsp
+; AVX-NEXT:    # kill: def $esi killed $esi def $rsi
+; AVX-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX-NEXT:    vmovaps %ymm0, (%rsp)
+; AVX-NEXT:    andl $3, %esi
+; AVX-NEXT:    vmovsd %xmm1, (%rsp,%rsi,8)
+; AVX-NEXT:    vmovaps (%rsp), %ymm0
+; AVX-NEXT:    movq %rbp, %rsp
+; AVX-NEXT:    popq %rbp
+; AVX-NEXT:    retq
+  %x = load double, double* %p
+  %ins = insertelement <4 x double> %v, double %x, i32 %y
+  ret <4 x double> %ins
+}
+
 ; Don't die trying to insert to an invalid index.
 
 define i32 @PR44139(<16 x i64>* %p) {
diff --git a/llvm/test/CodeGen/X86/stack-protector-3.ll b/llvm/test/CodeGen/X86/stack-protector-3.ll
new file mode 100644
index 0000000000000..96b1a0860e6b1
--- /dev/null
+++ b/llvm/test/CodeGen/X86/stack-protector-3.ll
@@ -0,0 +1,67 @@
+; RUN: llc -mtriple=x86_64-pc-linux-gnu -o - < %s | FileCheck --check-prefix=CHECK-TLS-FS-40 %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnu -stack-protector-guard=tls -o - < %s | FileCheck --check-prefix=CHECK-TLS-FS-40 %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnu -stack-protector-guard=global -o - < %s | FileCheck --check-prefix=CHECK-GLOBAL %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnu -stack-protector-guard-reg=fs -o - < %s | FileCheck --check-prefix=CHECK-TLS-FS-40 %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnu -stack-protector-guard-reg=gs -o - < %s | FileCheck --check-prefix=CHECK-GS %s
+; RUN: llc -mtriple=x86_64-pc-linux-gnu -stack-protector-guard-offset=20 -o - < %s | FileCheck --check-prefix=CHECK-OFFSET %s
+
+; CHECK-TLS-FS-40:       movq    %fs:40, %rax
+; CHECK-TLS-FS-40:       movq    %fs:40, %rax
+; CHECK-TLS-FS-40-NEXT:  cmpq    16(%rsp), %rax
+; CHECK-TLS-FS-40-NEXT:  jne     .LBB0_2
+; CHECK-TLS-FS-40:       .LBB0_2:
+; CHECK-TLS-FS-40-NEXT:  .cfi_def_cfa_offset 32
+; CHECK-TLS-FS-40-NEXT:  callq   __stack_chk_fail
+
+; CHECK-GS:       movq    %gs:40, %rax
+; CHECK-GS:       movq    %gs:40, %rax
+; CHECK-GS-NEXT:  cmpq    16(%rsp), %rax
+; CHECK-GS-NEXT:  jne     .LBB0_2
+; CHECK-GS:       .LBB0_2:
+; CHECK-GS-NEXT:  .cfi_def_cfa_offset 32
+; CHECK-GS-NEXT:  callq   __stack_chk_fail
+
+; CHECK-OFFSET:       movq    %fs:20, %rax
+; CHECK-OFFSET:       movq    %fs:20, %rax
+; CHECK-OFFSET-NEXT:  cmpq    16(%rsp), %rax
+; CHECK-OFFSET-NEXT:  jne     .LBB0_2
+; CHECK-OFFSET:       .LBB0_2:
+; CHECK-OFFSET-NEXT:  .cfi_def_cfa_offset 32
+; CHECK-OFFSET-NEXT:  callq   __stack_chk_fail
+
+; CHECK-GLOBAL:       movq    __stack_chk_guard(%rip), %rax
+; CHECK-GLOBAL:       movq    __stack_chk_guard(%rip), %rax
+; CHECK-GLOBAL-NEXT:  cmpq    16(%rsp), %rax
+; CHECK-GLOBAL-NEXT:  jne     .LBB0_2
+; CHECK-GLOBAL:       .LBB0_2:
+; CHECK-GLOBAL-NEXT:  .cfi_def_cfa_offset 32
+; CHECK-GLOBAL-NEXT:  callq   __stack_chk_fail
+
+; ModuleID = 't.c'
+@.str = private unnamed_addr constant [14 x i8] c"stackoverflow\00", align 1
+@a = dso_local local_unnamed_addr global i8* null, align 8
+
+; Function Attrs: nounwind sspreq uwtable writeonly
+define dso_local i32 @main() local_unnamed_addr #0 {
+entry:
+  %array = alloca [5 x i8], align 1
+  %0 = getelementptr inbounds [5 x i8], [5 x i8]* %array, i64 0, i64 0
+  call void @llvm.lifetime.start.p0i8(i64 5, i8* nonnull %0) #2
+  call void @llvm.memcpy.p0i8.p0i8.i64(i8* nonnull align 1 dereferenceable(14) %0, i8* nonnull align 1 dereferenceable(14) getelementptr inbounds ([14 x i8], [14 x i8]* @.str, i64 0, i64 0), i64 14, i1 false) #2
+  store i8* %0, i8** @a, align 8
+  call void @llvm.lifetime.end.p0i8(i64 5, i8* nonnull %0) #2
+  ret i32 0
+}
+
+; Function Attrs: argmemonly nounwind willreturn
+declare void @llvm.lifetime.start.p0i8(i64 immarg, i8* nocapture) #1
+
+; Function Attrs: argmemonly nounwind willreturn
+declare void @llvm.lifetime.end.p0i8(i64 immarg, i8* nocapture) #1
+
+; Function Attrs: argmemonly nounwind willreturn
+declare void @llvm.memcpy.p0i8.p0i8.i64(i8* noalias nocapture writeonly, i8* noalias nocapture readonly, i64, i1 immarg) #1
+
+attributes #0 = { nounwind sspreq uwtable writeonly "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { argmemonly nounwind willreturn }
+attributes #2 = { nounwind }
diff --git a/llvm/test/CodeGen/X86/uintr-intrinsics.ll b/llvm/test/CodeGen/X86/uintr-intrinsics.ll
new file mode 100644
index 0000000000000..cab0a2831e5e7
--- /dev/null
+++ b/llvm/test/CodeGen/X86/uintr-intrinsics.ll
@@ -0,0 +1,34 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+uintr | FileCheck %s --check-prefix=X64
+; RUN: llc < %s -mtriple=x86_64-linux-gnux32 -mattr=+uintr | FileCheck %s --check-prefix=X32
+
+define i8 @test_uintr(i64 %arg) {
+; X64-LABEL: test_uintr:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    clui
+; X64-NEXT:    stui
+; X64-NEXT:    senduipi %rdi
+; X64-NEXT:    testui
+; X64-NEXT:    setb %al
+; X64-NEXT:    retq
+
+; X32-LABEL: test_uintr:
+; X32:       # %bb.0: # %entry
+; X32-NEXT:    clui
+; X32-NEXT:    stui
+; X32-NEXT:    senduipi %rdi
+; X32-NEXT:    testui
+; X32-NEXT:    setb %al
+; X32-NEXT:    retq
+entry:
+  call void @llvm.x86.clui()
+  call void @llvm.x86.stui()
+  call void @llvm.x86.senduipi(i64 %arg)
+  %0 = call i8 @llvm.x86.testui()
+  ret i8 %0
+}
+
+declare void @llvm.x86.clui()
+declare void @llvm.x86.stui()
+declare i8 @llvm.x86.testui()
+declare void @llvm.x86.senduipi(i64 %arg)
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll
index 0f8881ad25c37..8ead5e2f4fa47 100644
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-128.ll
@@ -1262,112 +1262,218 @@ define <2 x double> @sitofp_v2i64_v2f64(<2 x i64> %x) #0 {
 define <2 x double> @uitofp_v2i64_v2f64(<2 x i64> %x) #0 {
 ; SSE-32-LABEL: uitofp_v2i64_v2f64:
 ; SSE-32:       # %bb.0:
-; SSE-32-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,0,4294967295,0]
-; SSE-32-NEXT:    pand %xmm0, %xmm1
-; SSE-32-NEXT:    por {{\.LCPI.*}}, %xmm1
-; SSE-32-NEXT:    psrlq $32, %xmm0
-; SSE-32-NEXT:    por {{\.LCPI.*}}, %xmm0
-; SSE-32-NEXT:    subpd {{\.LCPI.*}}, %xmm0
-; SSE-32-NEXT:    addpd %xmm1, %xmm0
+; SSE-32-NEXT:    pushl %ebp
+; SSE-32-NEXT:    .cfi_def_cfa_offset 8
+; SSE-32-NEXT:    .cfi_offset %ebp, -8
+; SSE-32-NEXT:    movl %esp, %ebp
+; SSE-32-NEXT:    .cfi_def_cfa_register %ebp
+; SSE-32-NEXT:    andl $-8, %esp
+; SSE-32-NEXT:    subl $32, %esp
+; SSE-32-NEXT:    movq %xmm0, {{[0-9]+}}(%esp)
+; SSE-32-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; SSE-32-NEXT:    movq %xmm1, {{[0-9]+}}(%esp)
+; SSE-32-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE-32-NEXT:    movd %xmm1, %eax
+; SSE-32-NEXT:    shrl $31, %eax
+; SSE-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; SSE-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; SSE-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; SSE-32-NEXT:    wait
+; SSE-32-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE-32-NEXT:    movd %xmm0, %eax
+; SSE-32-NEXT:    shrl $31, %eax
+; SSE-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; SSE-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; SSE-32-NEXT:    fstpl (%esp)
+; SSE-32-NEXT:    wait
+; SSE-32-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; SSE-32-NEXT:    movhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; SSE-32-NEXT:    movl %ebp, %esp
+; SSE-32-NEXT:    popl %ebp
+; SSE-32-NEXT:    .cfi_def_cfa %esp, 4
 ; SSE-32-NEXT:    retl
 ;
 ; SSE-64-LABEL: uitofp_v2i64_v2f64:
 ; SSE-64:       # %bb.0:
-; SSE-64-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,4294967295]
-; SSE-64-NEXT:    pand %xmm0, %xmm1
-; SSE-64-NEXT:    por {{.*}}(%rip), %xmm1
-; SSE-64-NEXT:    psrlq $32, %xmm0
-; SSE-64-NEXT:    por {{.*}}(%rip), %xmm0
-; SSE-64-NEXT:    subpd {{.*}}(%rip), %xmm0
-; SSE-64-NEXT:    addpd %xmm1, %xmm0
+; SSE-64-NEXT:    movdqa %xmm0, %xmm1
+; SSE-64-NEXT:    movq %xmm0, %rax
+; SSE-64-NEXT:    movq %rax, %rcx
+; SSE-64-NEXT:    shrq %rcx
+; SSE-64-NEXT:    movl %eax, %edx
+; SSE-64-NEXT:    andl $1, %edx
+; SSE-64-NEXT:    orq %rcx, %rdx
+; SSE-64-NEXT:    testq %rax, %rax
+; SSE-64-NEXT:    cmovnsq %rax, %rdx
+; SSE-64-NEXT:    xorps %xmm0, %xmm0
+; SSE-64-NEXT:    cvtsi2sd %rdx, %xmm0
+; SSE-64-NEXT:    jns .LBB21_2
+; SSE-64-NEXT:  # %bb.1:
+; SSE-64-NEXT:    addsd %xmm0, %xmm0
+; SSE-64-NEXT:  .LBB21_2:
+; SSE-64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE-64-NEXT:    movq %xmm1, %rax
+; SSE-64-NEXT:    movq %rax, %rcx
+; SSE-64-NEXT:    shrq %rcx
+; SSE-64-NEXT:    movl %eax, %edx
+; SSE-64-NEXT:    andl $1, %edx
+; SSE-64-NEXT:    orq %rcx, %rdx
+; SSE-64-NEXT:    testq %rax, %rax
+; SSE-64-NEXT:    cmovnsq %rax, %rdx
+; SSE-64-NEXT:    xorps %xmm1, %xmm1
+; SSE-64-NEXT:    cvtsi2sd %rdx, %xmm1
+; SSE-64-NEXT:    jns .LBB21_4
+; SSE-64-NEXT:  # %bb.3:
+; SSE-64-NEXT:    addsd %xmm1, %xmm1
+; SSE-64-NEXT:  .LBB21_4:
+; SSE-64-NEXT:    unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE-64-NEXT:    retq
 ;
 ; SSE41-32-LABEL: uitofp_v2i64_v2f64:
 ; SSE41-32:       # %bb.0:
-; SSE41-32-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,0,4294967295,0]
-; SSE41-32-NEXT:    pand %xmm0, %xmm1
-; SSE41-32-NEXT:    por {{\.LCPI.*}}, %xmm1
-; SSE41-32-NEXT:    psrlq $32, %xmm0
-; SSE41-32-NEXT:    por {{\.LCPI.*}}, %xmm0
-; SSE41-32-NEXT:    subpd {{\.LCPI.*}}, %xmm0
-; SSE41-32-NEXT:    addpd %xmm1, %xmm0
+; SSE41-32-NEXT:    pushl %ebp
+; SSE41-32-NEXT:    .cfi_def_cfa_offset 8
+; SSE41-32-NEXT:    .cfi_offset %ebp, -8
+; SSE41-32-NEXT:    movl %esp, %ebp
+; SSE41-32-NEXT:    .cfi_def_cfa_register %ebp
+; SSE41-32-NEXT:    andl $-8, %esp
+; SSE41-32-NEXT:    subl $32, %esp
+; SSE41-32-NEXT:    movq %xmm0, {{[0-9]+}}(%esp)
+; SSE41-32-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; SSE41-32-NEXT:    movq %xmm1, {{[0-9]+}}(%esp)
+; SSE41-32-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,1,1]
+; SSE41-32-NEXT:    movd %xmm1, %eax
+; SSE41-32-NEXT:    shrl $31, %eax
+; SSE41-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; SSE41-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; SSE41-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; SSE41-32-NEXT:    wait
+; SSE41-32-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[3,3,3,3]
+; SSE41-32-NEXT:    movd %xmm0, %eax
+; SSE41-32-NEXT:    shrl $31, %eax
+; SSE41-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; SSE41-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; SSE41-32-NEXT:    fstpl (%esp)
+; SSE41-32-NEXT:    wait
+; SSE41-32-NEXT:    movsd {{.*#+}} xmm0 = mem[0],zero
+; SSE41-32-NEXT:    movhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; SSE41-32-NEXT:    movl %ebp, %esp
+; SSE41-32-NEXT:    popl %ebp
+; SSE41-32-NEXT:    .cfi_def_cfa %esp, 4
 ; SSE41-32-NEXT:    retl
 ;
 ; SSE41-64-LABEL: uitofp_v2i64_v2f64:
 ; SSE41-64:       # %bb.0:
-; SSE41-64-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,4294967295]
-; SSE41-64-NEXT:    pand %xmm0, %xmm1
-; SSE41-64-NEXT:    por {{.*}}(%rip), %xmm1
-; SSE41-64-NEXT:    psrlq $32, %xmm0
-; SSE41-64-NEXT:    por {{.*}}(%rip), %xmm0
-; SSE41-64-NEXT:    subpd {{.*}}(%rip), %xmm0
-; SSE41-64-NEXT:    addpd %xmm1, %xmm0
+; SSE41-64-NEXT:    movdqa %xmm0, %xmm1
+; SSE41-64-NEXT:    movq %xmm0, %rax
+; SSE41-64-NEXT:    movq %rax, %rcx
+; SSE41-64-NEXT:    shrq %rcx
+; SSE41-64-NEXT:    movl %eax, %edx
+; SSE41-64-NEXT:    andl $1, %edx
+; SSE41-64-NEXT:    orq %rcx, %rdx
+; SSE41-64-NEXT:    testq %rax, %rax
+; SSE41-64-NEXT:    cmovnsq %rax, %rdx
+; SSE41-64-NEXT:    xorps %xmm0, %xmm0
+; SSE41-64-NEXT:    cvtsi2sd %rdx, %xmm0
+; SSE41-64-NEXT:    jns .LBB21_2
+; SSE41-64-NEXT:  # %bb.1:
+; SSE41-64-NEXT:    addsd %xmm0, %xmm0
+; SSE41-64-NEXT:  .LBB21_2:
+; SSE41-64-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; SSE41-64-NEXT:    movq %xmm1, %rax
+; SSE41-64-NEXT:    movq %rax, %rcx
+; SSE41-64-NEXT:    shrq %rcx
+; SSE41-64-NEXT:    movl %eax, %edx
+; SSE41-64-NEXT:    andl $1, %edx
+; SSE41-64-NEXT:    orq %rcx, %rdx
+; SSE41-64-NEXT:    testq %rax, %rax
+; SSE41-64-NEXT:    cmovnsq %rax, %rdx
+; SSE41-64-NEXT:    xorps %xmm1, %xmm1
+; SSE41-64-NEXT:    cvtsi2sd %rdx, %xmm1
+; SSE41-64-NEXT:    jns .LBB21_4
+; SSE41-64-NEXT:  # %bb.3:
+; SSE41-64-NEXT:    addsd %xmm1, %xmm1
+; SSE41-64-NEXT:  .LBB21_4:
+; SSE41-64-NEXT:    unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; SSE41-64-NEXT:    retq
 ;
-; AVX1-32-LABEL: uitofp_v2i64_v2f64:
-; AVX1-32:       # %bb.0:
-; AVX1-32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-32-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
-; AVX1-32-NEXT:    vpor {{\.LCPI.*}}, %xmm1, %xmm1
-; AVX1-32-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX1-32-NEXT:    vpor {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX1-32-NEXT:    vsubpd {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX1-32-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX1-32-NEXT:    retl
+; AVX-32-LABEL: uitofp_v2i64_v2f64:
+; AVX-32:       # %bb.0:
+; AVX-32-NEXT:    pushl %ebp
+; AVX-32-NEXT:    .cfi_def_cfa_offset 8
+; AVX-32-NEXT:    .cfi_offset %ebp, -8
+; AVX-32-NEXT:    movl %esp, %ebp
+; AVX-32-NEXT:    .cfi_def_cfa_register %ebp
+; AVX-32-NEXT:    andl $-8, %esp
+; AVX-32-NEXT:    subl $32, %esp
+; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    vextractps $1, %xmm0, %eax
+; AVX-32-NEXT:    shrl $31, %eax
+; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    wait
+; AVX-32-NEXT:    vextractps $3, %xmm0, %eax
+; AVX-32-NEXT:    shrl $31, %eax
+; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-32-NEXT:    fstpl (%esp)
+; AVX-32-NEXT:    wait
+; AVX-32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX-32-NEXT:    vmovhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; AVX-32-NEXT:    movl %ebp, %esp
+; AVX-32-NEXT:    popl %ebp
+; AVX-32-NEXT:    .cfi_def_cfa %esp, 4
+; AVX-32-NEXT:    retl
 ;
 ; AVX1-64-LABEL: uitofp_v2i64_v2f64:
 ; AVX1-64:       # %bb.0:
-; AVX1-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-64-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
-; AVX1-64-NEXT:    vpor {{.*}}(%rip), %xmm1, %xmm1
-; AVX1-64-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX1-64-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-64-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX1-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-64-NEXT:    movq %rax, %rcx
+; AVX1-64-NEXT:    shrq %rcx
+; AVX1-64-NEXT:    movl %eax, %edx
+; AVX1-64-NEXT:    andl $1, %edx
+; AVX1-64-NEXT:    orq %rcx, %rdx
+; AVX1-64-NEXT:    testq %rax, %rax
+; AVX1-64-NEXT:    cmovnsq %rax, %rdx
+; AVX1-64-NEXT:    vcvtsi2sd %rdx, %xmm1, %xmm1
+; AVX1-64-NEXT:    jns .LBB21_2
+; AVX1-64-NEXT:  # %bb.1:
+; AVX1-64-NEXT:    vaddsd %xmm1, %xmm1, %xmm1
+; AVX1-64-NEXT:  .LBB21_2:
+; AVX1-64-NEXT:    vmovq %xmm0, %rax
+; AVX1-64-NEXT:    movq %rax, %rcx
+; AVX1-64-NEXT:    shrq %rcx
+; AVX1-64-NEXT:    movl %eax, %edx
+; AVX1-64-NEXT:    andl $1, %edx
+; AVX1-64-NEXT:    orq %rcx, %rdx
+; AVX1-64-NEXT:    testq %rax, %rax
+; AVX1-64-NEXT:    cmovnsq %rax, %rdx
+; AVX1-64-NEXT:    vcvtsi2sd %rdx, %xmm2, %xmm0
+; AVX1-64-NEXT:    jns .LBB21_4
+; AVX1-64-NEXT:  # %bb.3:
+; AVX1-64-NEXT:    vaddsd %xmm0, %xmm0, %xmm0
+; AVX1-64-NEXT:  .LBB21_4:
+; AVX1-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-64-NEXT:    retq
 ;
-; AVX512F-32-LABEL: uitofp_v2i64_v2f64:
-; AVX512F-32:       # %bb.0:
-; AVX512F-32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-32-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX512F-32-NEXT:    vpor {{\.LCPI.*}}, %xmm1, %xmm1
-; AVX512F-32-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX512F-32-NEXT:    vpor {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX512F-32-NEXT:    vsubpd {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX512F-32-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX512F-32-NEXT:    retl
-;
 ; AVX512F-64-LABEL: uitofp_v2i64_v2f64:
 ; AVX512F-64:       # %bb.0:
-; AVX512F-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-64-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX512F-64-NEXT:    vpor {{.*}}(%rip), %xmm1, %xmm1
-; AVX512F-64-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX512F-64-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
-; AVX512F-64-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX512F-64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX512F-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512F-64-NEXT:    vcvtusi2sd %rax, %xmm1, %xmm1
+; AVX512F-64-NEXT:    vmovq %xmm0, %rax
+; AVX512F-64-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm0
+; AVX512F-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX512F-64-NEXT:    retq
 ;
-; AVX512VL-32-LABEL: uitofp_v2i64_v2f64:
-; AVX512VL-32:       # %bb.0:
-; AVX512VL-32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-32-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX512VL-32-NEXT:    vpor {{\.LCPI.*}}, %xmm1, %xmm1
-; AVX512VL-32-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX512VL-32-NEXT:    vpor {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX512VL-32-NEXT:    vsubpd {{\.LCPI.*}}, %xmm0, %xmm0
-; AVX512VL-32-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX512VL-32-NEXT:    retl
-;
 ; AVX512VL-64-LABEL: uitofp_v2i64_v2f64:
 ; AVX512VL-64:       # %bb.0:
-; AVX512VL-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-64-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX512VL-64-NEXT:    vpor {{.*}}(%rip), %xmm1, %xmm1
-; AVX512VL-64-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX512VL-64-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
-; AVX512VL-64-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX512VL-64-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX512VL-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512VL-64-NEXT:    vcvtusi2sd %rax, %xmm1, %xmm1
+; AVX512VL-64-NEXT:    vmovq %xmm0, %rax
+; AVX512VL-64-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm0
+; AVX512VL-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX512VL-64-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: uitofp_v2i64_v2f64:
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
index 00be3baedac89..82af829506bda 100644
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-256.ll
@@ -748,106 +748,154 @@ define <4 x double> @sitofp_v4i64_v4f64(<4 x i64> %x) #0 {
 }
 
 define <4 x double> @uitofp_v4i64_v4f64(<4 x i64> %x) #0 {
-; AVX1-32-LABEL: uitofp_v4i64_v4f64:
-; AVX1-32:       # %bb.0:
-; AVX1-32-NEXT:    vxorps %xmm1, %xmm1, %xmm1
-; AVX1-32-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX1-32-NEXT:    vorps {{\.LCPI.*}}, %ymm1, %ymm1
-; AVX1-32-NEXT:    vpsrlq $32, %xmm0, %xmm2
-; AVX1-32-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-32-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX1-32-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-32-NEXT:    vorpd {{\.LCPI.*}}, %ymm0, %ymm0
-; AVX1-32-NEXT:    vsubpd {{\.LCPI.*}}, %ymm0, %ymm0
-; AVX1-32-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
-; AVX1-32-NEXT:    retl
+; AVX-32-LABEL: uitofp_v4i64_v4f64:
+; AVX-32:       # %bb.0:
+; AVX-32-NEXT:    pushl %ebp
+; AVX-32-NEXT:    .cfi_def_cfa_offset 8
+; AVX-32-NEXT:    .cfi_offset %ebp, -8
+; AVX-32-NEXT:    movl %esp, %ebp
+; AVX-32-NEXT:    .cfi_def_cfa_register %ebp
+; AVX-32-NEXT:    andl $-8, %esp
+; AVX-32-NEXT:    subl $64, %esp
+; AVX-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    vpermilps {{.*#+}} xmm2 = xmm1[2,3,2,3]
+; AVX-32-NEXT:    vmovlps %xmm2, {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    vextractps $1, %xmm0, %eax
+; AVX-32-NEXT:    shrl $31, %eax
+; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-32-NEXT:    fstpl (%esp)
+; AVX-32-NEXT:    wait
+; AVX-32-NEXT:    vextractps $3, %xmm0, %eax
+; AVX-32-NEXT:    shrl $31, %eax
+; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    wait
+; AVX-32-NEXT:    vextractps $1, %xmm1, %eax
+; AVX-32-NEXT:    shrl $31, %eax
+; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    wait
+; AVX-32-NEXT:    vextractps $3, %xmm1, %eax
+; AVX-32-NEXT:    shrl $31, %eax
+; AVX-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; AVX-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; AVX-32-NEXT:    wait
+; AVX-32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; AVX-32-NEXT:    vmovhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; AVX-32-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; AVX-32-NEXT:    vmovhps {{.*#+}} xmm1 = xmm1[0,1],mem[0,1]
+; AVX-32-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX-32-NEXT:    movl %ebp, %esp
+; AVX-32-NEXT:    popl %ebp
+; AVX-32-NEXT:    .cfi_def_cfa %esp, 4
+; AVX-32-NEXT:    retl
 ;
 ; AVX1-64-LABEL: uitofp_v4i64_v4f64:
 ; AVX1-64:       # %bb.0:
 ; AVX1-64-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-64-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX1-64-NEXT:    vorps {{.*}}(%rip), %ymm1, %ymm1
-; AVX1-64-NEXT:    vpsrlq $32, %xmm0, %xmm2
-; AVX1-64-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-64-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-64-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm3, %xmm3
+; AVX1-64-NEXT:    vmovq %xmm2, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm2
+; AVX1-64-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX1-64-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX1-64-NEXT:    vmovq %xmm1, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm1
+; AVX1-64-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; AVX1-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-64-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-64-NEXT:    vpsrlq $32, %xmm2, %xmm2
+; AVX1-64-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX1-64-NEXT:    vmovq %xmm2, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm2
+; AVX1-64-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
 ; AVX1-64-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX1-64-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-64-NEXT:    vorpd {{.*}}(%rip), %ymm0, %ymm0
-; AVX1-64-NEXT:    vsubpd {{.*}}(%rip), %ymm0, %ymm0
-; AVX1-64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
+; AVX1-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX1-64-NEXT:    vmovq %xmm0, %rax
+; AVX1-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm0
+; AVX1-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; AVX1-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-64-NEXT:    vmulpd {{.*}}(%rip), %ymm0, %ymm0
+; AVX1-64-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
 ; AVX1-64-NEXT:    retq
 ;
-; AVX2-32-LABEL: uitofp_v4i64_v4f64:
-; AVX2-32:       # %bb.0:
-; AVX2-32-NEXT:    vpsrlq $32, %ymm0, %ymm1
-; AVX2-32-NEXT:    vpor {{\.LCPI.*}}, %ymm1, %ymm1
-; AVX2-32-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25]
-; AVX2-32-NEXT:    vsubpd %ymm2, %ymm1, %ymm1
-; AVX2-32-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
-; AVX2-32-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2],ymm2[3],ymm0[4],ymm2[5],ymm0[6],ymm2[7]
-; AVX2-32-NEXT:    vpor {{\.LCPI.*}}, %ymm0, %ymm0
-; AVX2-32-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
-; AVX2-32-NEXT:    retl
-;
 ; AVX2-64-LABEL: uitofp_v4i64_v4f64:
 ; AVX2-64:       # %bb.0:
-; AVX2-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-64-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX2-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200]
-; AVX2-64-NEXT:    vpor %ymm2, %ymm1, %ymm1
-; AVX2-64-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX2-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072]
-; AVX2-64-NEXT:    vpor %ymm2, %ymm0, %ymm0
-; AVX2-64-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25]
-; AVX2-64-NEXT:    vsubpd %ymm2, %ymm0, %ymm0
+; AVX2-64-NEXT:    vpsrlq $32, %ymm0, %ymm1
+; AVX2-64-NEXT:    vextracti128 $1, %ymm1, %xmm2
+; AVX2-64-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm3, %xmm3
+; AVX2-64-NEXT:    vmovq %xmm2, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm2
+; AVX2-64-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX2-64-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX2-64-NEXT:    vmovq %xmm1, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm1
+; AVX2-64-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; AVX2-64-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [4.294967296E+9,4.294967296E+9,4.294967296E+9,4.294967296E+9]
+; AVX2-64-NEXT:    vmulpd %ymm2, %ymm1, %ymm1
+; AVX2-64-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
+; AVX2-64-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2],ymm2[3],ymm0[4],ymm2[5],ymm0[6],ymm2[7]
+; AVX2-64-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; AVX2-64-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX2-64-NEXT:    vmovq %xmm2, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm2
+; AVX2-64-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX2-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX2-64-NEXT:    vmovq %xmm0, %rax
+; AVX2-64-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm0
+; AVX2-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; AVX2-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
 ; AVX2-64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
 ; AVX2-64-NEXT:    retq
 ;
-; AVX512F-32-LABEL: uitofp_v4i64_v4f64:
-; AVX512F-32:       # %bb.0:
-; AVX512F-32-NEXT:    vpsrlq $32, %ymm0, %ymm1
-; AVX512F-32-NEXT:    vpor {{\.LCPI.*}}, %ymm1, %ymm1
-; AVX512F-32-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25]
-; AVX512F-32-NEXT:    vsubpd %ymm2, %ymm1, %ymm1
-; AVX512F-32-NEXT:    vxorpd %xmm2, %xmm2, %xmm2
-; AVX512F-32-NEXT:    vpblendd {{.*#+}} ymm0 = ymm0[0],ymm2[1],ymm0[2],ymm2[3],ymm0[4],ymm2[5],ymm0[6],ymm2[7]
-; AVX512F-32-NEXT:    vpor {{\.LCPI.*}}, %ymm0, %ymm0
-; AVX512F-32-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
-; AVX512F-32-NEXT:    retl
-;
 ; AVX512F-64-LABEL: uitofp_v4i64_v4f64:
 ; AVX512F-64:       # %bb.0:
-; AVX512F-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-64-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX512F-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200]
-; AVX512F-64-NEXT:    vpor %ymm2, %ymm1, %ymm1
-; AVX512F-64-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX512F-64-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072]
-; AVX512F-64-NEXT:    vpor %ymm2, %ymm0, %ymm0
-; AVX512F-64-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25]
-; AVX512F-64-NEXT:    vsubpd %ymm2, %ymm0, %ymm0
-; AVX512F-64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
+; AVX512F-64-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX512F-64-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX512F-64-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm2
+; AVX512F-64-NEXT:    vmovq %xmm1, %rax
+; AVX512F-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm1
+; AVX512F-64-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX512F-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512F-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm2
+; AVX512F-64-NEXT:    vmovq %xmm0, %rax
+; AVX512F-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm0
+; AVX512F-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX512F-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512F-64-NEXT:    retq
 ;
-; AVX512VL-32-LABEL: uitofp_v4i64_v4f64:
-; AVX512VL-32:       # %bb.0:
-; AVX512VL-32-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-32-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX512VL-32-NEXT:    vpor {{\.LCPI.*}}, %ymm1, %ymm1
-; AVX512VL-32-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX512VL-32-NEXT:    vpor {{\.LCPI.*}}, %ymm0, %ymm0
-; AVX512VL-32-NEXT:    vsubpd {{\.LCPI.*}}{1to4}, %ymm0, %ymm0
-; AVX512VL-32-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
-; AVX512VL-32-NEXT:    retl
-;
 ; AVX512VL-64-LABEL: uitofp_v4i64_v4f64:
 ; AVX512VL-64:       # %bb.0:
-; AVX512VL-64-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VL-64-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX512VL-64-NEXT:    vporq {{.*}}(%rip){1to4}, %ymm1, %ymm1
-; AVX512VL-64-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX512VL-64-NEXT:    vporq {{.*}}(%rip){1to4}, %ymm0, %ymm0
-; AVX512VL-64-NEXT:    vsubpd {{.*}}(%rip){1to4}, %ymm0, %ymm0
-; AVX512VL-64-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
+; AVX512VL-64-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX512VL-64-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX512VL-64-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm2
+; AVX512VL-64-NEXT:    vmovq %xmm1, %rax
+; AVX512VL-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm1
+; AVX512VL-64-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX512VL-64-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512VL-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm2
+; AVX512VL-64-NEXT:    vmovq %xmm0, %rax
+; AVX512VL-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm0
+; AVX512VL-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX512VL-64-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512VL-64-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: uitofp_v4i64_v4f64:
diff --git a/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll b/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll
index 328f3c15fc48b..00be17a4d162c 100644
--- a/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll
+++ b/llvm/test/CodeGen/X86/vec-strict-inttofp-512.ll
@@ -362,22 +362,120 @@ define <8 x double> @sitofp_v8i64_v8f64(<8 x i64> %x) #0 {
 define <8 x double> @uitofp_v8i64_v8f64(<8 x i64> %x) #0 {
 ; NODQ-32-LABEL: uitofp_v8i64_v8f64:
 ; NODQ-32:       # %bb.0:
-; NODQ-32-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [0,1127219200,0,1127219200,0,1127219200,0,1127219200,0,1127219200,0,1127219200,0,1127219200,0,1127219200]
-; NODQ-32-NEXT:    vpternlogq $248, {{\.LCPI.*}}, %zmm0, %zmm1
-; NODQ-32-NEXT:    vpsrlq $32, %zmm0, %zmm0
-; NODQ-32-NEXT:    vporq {{\.LCPI.*}}, %zmm0, %zmm0
-; NODQ-32-NEXT:    vsubpd {{\.LCPI.*}}{1to8}, %zmm0, %zmm0
-; NODQ-32-NEXT:    vaddpd %zmm0, %zmm1, %zmm0
+; NODQ-32-NEXT:    pushl %ebp
+; NODQ-32-NEXT:    .cfi_def_cfa_offset 8
+; NODQ-32-NEXT:    .cfi_offset %ebp, -8
+; NODQ-32-NEXT:    movl %esp, %ebp
+; NODQ-32-NEXT:    .cfi_def_cfa_register %ebp
+; NODQ-32-NEXT:    andl $-8, %esp
+; NODQ-32-NEXT:    subl $128, %esp
+; NODQ-32-NEXT:    vextractf32x4 $2, %zmm0, %xmm2
+; NODQ-32-NEXT:    vmovlps %xmm2, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm2[2,3,2,3]
+; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vextractf32x4 $3, %zmm0, %xmm3
+; NODQ-32-NEXT:    vmovlps %xmm3, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm3[2,3,2,3]
+; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vmovlps %xmm0, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm1 = xmm0[2,3,2,3]
+; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; NODQ-32-NEXT:    vmovlps %xmm1, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vpermilps {{.*#+}} xmm4 = xmm1[2,3,2,3]
+; NODQ-32-NEXT:    vmovlps %xmm4, {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    vextractps $1, %xmm2, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $3, %xmm2, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $1, %xmm3, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $3, %xmm3, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $1, %xmm0, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl (%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $3, %xmm0, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $1, %xmm1, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vextractps $3, %xmm1, %eax
+; NODQ-32-NEXT:    shrl $31, %eax
+; NODQ-32-NEXT:    fildll {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    fadds {{\.LCPI.*}}(,%eax,4)
+; NODQ-32-NEXT:    fstpl {{[0-9]+}}(%esp)
+; NODQ-32-NEXT:    wait
+; NODQ-32-NEXT:    vmovsd {{.*#+}} xmm0 = mem[0],zero
+; NODQ-32-NEXT:    vmovhps {{.*#+}} xmm0 = xmm0[0,1],mem[0,1]
+; NODQ-32-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; NODQ-32-NEXT:    vmovhps {{.*#+}} xmm1 = xmm1[0,1],mem[0,1]
+; NODQ-32-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; NODQ-32-NEXT:    vmovsd {{.*#+}} xmm1 = mem[0],zero
+; NODQ-32-NEXT:    vmovhps {{.*#+}} xmm1 = xmm1[0,1],mem[0,1]
+; NODQ-32-NEXT:    vmovsd {{.*#+}} xmm2 = mem[0],zero
+; NODQ-32-NEXT:    vmovhps {{.*#+}} xmm2 = xmm2[0,1],mem[0,1]
+; NODQ-32-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; NODQ-32-NEXT:    vinsertf64x4 $1, %ymm0, %zmm1, %zmm0
+; NODQ-32-NEXT:    movl %ebp, %esp
+; NODQ-32-NEXT:    popl %ebp
+; NODQ-32-NEXT:    .cfi_def_cfa %esp, 4
 ; NODQ-32-NEXT:    retl
 ;
 ; NODQ-64-LABEL: uitofp_v8i64_v8f64:
 ; NODQ-64:       # %bb.0:
-; NODQ-64-NEXT:    vpbroadcastq {{.*#+}} zmm1 = [4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200]
-; NODQ-64-NEXT:    vpternlogq $248, {{.*}}(%rip){1to8}, %zmm0, %zmm1
-; NODQ-64-NEXT:    vpsrlq $32, %zmm0, %zmm0
-; NODQ-64-NEXT:    vporq {{.*}}(%rip){1to8}, %zmm0, %zmm0
-; NODQ-64-NEXT:    vsubpd {{.*}}(%rip){1to8}, %zmm0, %zmm0
-; NODQ-64-NEXT:    vaddpd %zmm0, %zmm1, %zmm0
+; NODQ-64-NEXT:    vextracti32x4 $3, %zmm0, %xmm1
+; NODQ-64-NEXT:    vpextrq $1, %xmm1, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm2
+; NODQ-64-NEXT:    vmovq %xmm1, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm1
+; NODQ-64-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; NODQ-64-NEXT:    vextracti32x4 $2, %zmm0, %xmm2
+; NODQ-64-NEXT:    vpextrq $1, %xmm2, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm3
+; NODQ-64-NEXT:    vmovq %xmm2, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm4, %xmm2
+; NODQ-64-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; NODQ-64-NEXT:    vinsertf128 $1, %xmm1, %ymm2, %ymm1
+; NODQ-64-NEXT:    vextracti128 $1, %ymm0, %xmm2
+; NODQ-64-NEXT:    vpextrq $1, %xmm2, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm4, %xmm3
+; NODQ-64-NEXT:    vmovq %xmm2, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm4, %xmm2
+; NODQ-64-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; NODQ-64-NEXT:    vpextrq $1, %xmm0, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm4, %xmm3
+; NODQ-64-NEXT:    vmovq %xmm0, %rax
+; NODQ-64-NEXT:    vcvtusi2sd %rax, %xmm4, %xmm0
+; NODQ-64-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; NODQ-64-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; NODQ-64-NEXT:    vinsertf64x4 $1, %ymm1, %zmm0, %zmm0
 ; NODQ-64-NEXT:    retq
 ;
 ; DQ-LABEL: uitofp_v8i64_v8f64:
diff --git a/llvm/test/CodeGen/X86/vector-constrained-fp-intrinsics.ll b/llvm/test/CodeGen/X86/vector-constrained-fp-intrinsics.ll
index 1e3bc0c61c913..5f8b3b33bfdce 100644
--- a/llvm/test/CodeGen/X86/vector-constrained-fp-intrinsics.ll
+++ b/llvm/test/CodeGen/X86/vector-constrained-fp-intrinsics.ll
@@ -6761,21 +6761,34 @@ entry:
 define <1 x double> @constrained_vector_uitofp_v1f64_v1i64(<1 x i64> %x) #0 {
 ; CHECK-LABEL: constrained_vector_uitofp_v1f64_v1i64:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movq %rdi, %xmm1
-; CHECK-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],mem[0],xmm1[1],mem[1]
-; CHECK-NEXT:    subpd {{.*}}(%rip), %xmm1
-; CHECK-NEXT:    movapd %xmm1, %xmm0
-; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; CHECK-NEXT:    addpd %xmm1, %xmm0
+; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    movl %edi, %ecx
+; CHECK-NEXT:    andl $1, %ecx
+; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    testq %rdi, %rdi
+; CHECK-NEXT:    cmovnsq %rdi, %rcx
+; CHECK-NEXT:    cvtsi2sd %rcx, %xmm0
+; CHECK-NEXT:    jns .LBB169_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    addsd %xmm0, %xmm0
+; CHECK-NEXT:  .LBB169_2: # %entry
 ; CHECK-NEXT:    retq
 ;
 ; AVX1-LABEL: constrained_vector_uitofp_v1f64_v1i64:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vmovq %rdi, %xmm0
-; AVX1-NEXT:    vpunpckldq {{.*#+}} xmm0 = xmm0[0],mem[0],xmm0[1],mem[1]
-; AVX1-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    movq %rdi, %rax
+; AVX1-NEXT:    shrq %rax
+; AVX1-NEXT:    movl %edi, %ecx
+; AVX1-NEXT:    andl $1, %ecx
+; AVX1-NEXT:    orq %rax, %rcx
+; AVX1-NEXT:    testq %rdi, %rdi
+; AVX1-NEXT:    cmovnsq %rdi, %rcx
+; AVX1-NEXT:    vcvtsi2sd %rcx, %xmm0, %xmm0
+; AVX1-NEXT:    jns .LBB169_2
+; AVX1-NEXT:  # %bb.1:
+; AVX1-NEXT:    vaddsd %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:  .LBB169_2: # %entry
 ; AVX1-NEXT:    retq
 ;
 ; AVX512-LABEL: constrained_vector_uitofp_v1f64_v1i64:
@@ -6906,35 +6919,77 @@ entry:
 define <2 x double> @constrained_vector_uitofp_v2f64_v2i64(<2 x i64> %x) #0 {
 ; CHECK-LABEL: constrained_vector_uitofp_v2f64_v2i64:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movdqa {{.*#+}} xmm1 = [4294967295,4294967295]
-; CHECK-NEXT:    pand %xmm0, %xmm1
-; CHECK-NEXT:    por {{.*}}(%rip), %xmm1
-; CHECK-NEXT:    psrlq $32, %xmm0
-; CHECK-NEXT:    por {{.*}}(%rip), %xmm0
-; CHECK-NEXT:    subpd {{.*}}(%rip), %xmm0
-; CHECK-NEXT:    addpd %xmm1, %xmm0
+; CHECK-NEXT:    movdqa %xmm0, %xmm1
+; CHECK-NEXT:    movq %xmm0, %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    shrq %rcx
+; CHECK-NEXT:    movl %eax, %edx
+; CHECK-NEXT:    andl $1, %edx
+; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmovnsq %rax, %rdx
+; CHECK-NEXT:    xorps %xmm0, %xmm0
+; CHECK-NEXT:    cvtsi2sd %rdx, %xmm0
+; CHECK-NEXT:    jns .LBB173_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    addsd %xmm0, %xmm0
+; CHECK-NEXT:  .LBB173_2: # %entry
+; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; CHECK-NEXT:    movq %xmm1, %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    shrq %rcx
+; CHECK-NEXT:    movl %eax, %edx
+; CHECK-NEXT:    andl $1, %edx
+; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmovnsq %rax, %rdx
+; CHECK-NEXT:    xorps %xmm1, %xmm1
+; CHECK-NEXT:    cvtsi2sd %rdx, %xmm1
+; CHECK-NEXT:    jns .LBB173_4
+; CHECK-NEXT:  # %bb.3:
+; CHECK-NEXT:    addsd %xmm1, %xmm1
+; CHECK-NEXT:  .LBB173_4: # %entry
+; CHECK-NEXT:    unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; CHECK-NEXT:    retq
 ;
 ; AVX1-LABEL: constrained_vector_uitofp_v2f64_v2i64:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpblendw {{.*#+}} xmm1 = xmm0[0,1],xmm1[2,3],xmm0[4,5],xmm1[6,7]
-; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm1, %xmm1
-; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX1-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    shrq %rcx
+; AVX1-NEXT:    movl %eax, %edx
+; AVX1-NEXT:    andl $1, %edx
+; AVX1-NEXT:    orq %rcx, %rdx
+; AVX1-NEXT:    testq %rax, %rax
+; AVX1-NEXT:    cmovnsq %rax, %rdx
+; AVX1-NEXT:    vcvtsi2sd %rdx, %xmm1, %xmm1
+; AVX1-NEXT:    jns .LBB173_2
+; AVX1-NEXT:  # %bb.1:
+; AVX1-NEXT:    vaddsd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:  .LBB173_2: # %entry
+; AVX1-NEXT:    vmovq %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    shrq %rcx
+; AVX1-NEXT:    movl %eax, %edx
+; AVX1-NEXT:    andl $1, %edx
+; AVX1-NEXT:    orq %rcx, %rdx
+; AVX1-NEXT:    testq %rax, %rax
+; AVX1-NEXT:    cmovnsq %rax, %rdx
+; AVX1-NEXT:    vcvtsi2sd %rdx, %xmm2, %xmm0
+; AVX1-NEXT:    jns .LBB173_4
+; AVX1-NEXT:  # %bb.3:
+; AVX1-NEXT:    vaddsd %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:  .LBB173_4: # %entry
+; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX1-NEXT:    retq
 ;
 ; AVX512F-LABEL: constrained_vector_uitofp_v2f64_v2i64:
 ; AVX512F:       # %bb.0: # %entry
-; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT:    vpblendd {{.*#+}} xmm1 = xmm0[0],xmm1[1],xmm0[2],xmm1[3]
-; AVX512F-NEXT:    vpor {{.*}}(%rip), %xmm1, %xmm1
-; AVX512F-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX512F-NEXT:    vpor {{.*}}(%rip), %xmm0, %xmm0
-; AVX512F-NEXT:    vsubpd {{.*}}(%rip), %xmm0, %xmm0
-; AVX512F-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
+; AVX512F-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512F-NEXT:    vcvtusi2sd %rax, %xmm1, %xmm1
+; AVX512F-NEXT:    vmovq %xmm0, %rax
+; AVX512F-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm0
+; AVX512F-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm1[0]
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: constrained_vector_uitofp_v2f64_v2i64:
@@ -7124,51 +7179,91 @@ entry:
 define <3 x double> @constrained_vector_uitofp_v3f64_v3i64(<3 x i64> %x) #0 {
 ; CHECK-LABEL: constrained_vector_uitofp_v3f64_v3i64:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movq %rdi, %xmm1
-; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [1127219200,1160773632,0,0]
-; CHECK-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1]
-; CHECK-NEXT:    movapd {{.*#+}} xmm3 = [4.503599627370496E+15,1.9342813113834067E+25]
-; CHECK-NEXT:    subpd %xmm3, %xmm1
-; CHECK-NEXT:    movapd %xmm1, %xmm0
-; CHECK-NEXT:    unpckhpd {{.*#+}} xmm0 = xmm0[1],xmm1[1]
-; CHECK-NEXT:    addpd %xmm1, %xmm0
-; CHECK-NEXT:    movq %rsi, %xmm4
-; CHECK-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; CHECK-NEXT:    subpd %xmm3, %xmm4
-; CHECK-NEXT:    movapd %xmm4, %xmm1
-; CHECK-NEXT:    unpckhpd {{.*#+}} xmm1 = xmm1[1],xmm4[1]
-; CHECK-NEXT:    addpd %xmm4, %xmm1
-; CHECK-NEXT:    movq %rdx, %xmm4
-; CHECK-NEXT:    punpckldq {{.*#+}} xmm4 = xmm4[0],xmm2[0],xmm4[1],xmm2[1]
-; CHECK-NEXT:    subpd %xmm3, %xmm4
-; CHECK-NEXT:    movapd %xmm4, %xmm2
-; CHECK-NEXT:    unpckhpd {{.*#+}} xmm2 = xmm2[1],xmm4[1]
-; CHECK-NEXT:    addpd %xmm4, %xmm2
-; CHECK-NEXT:    movlpd %xmm2, -{{[0-9]+}}(%rsp)
+; CHECK-NEXT:    movq %rdi, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    movl %edi, %ecx
+; CHECK-NEXT:    andl $1, %ecx
+; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    testq %rdi, %rdi
+; CHECK-NEXT:    cmovnsq %rdi, %rcx
+; CHECK-NEXT:    cvtsi2sd %rcx, %xmm0
+; CHECK-NEXT:    jns .LBB177_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    addsd %xmm0, %xmm0
+; CHECK-NEXT:  .LBB177_2: # %entry
+; CHECK-NEXT:    movq %rsi, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    movl %esi, %ecx
+; CHECK-NEXT:    andl $1, %ecx
+; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    testq %rsi, %rsi
+; CHECK-NEXT:    cmovnsq %rsi, %rcx
+; CHECK-NEXT:    cvtsi2sd %rcx, %xmm1
+; CHECK-NEXT:    jns .LBB177_4
+; CHECK-NEXT:  # %bb.3:
+; CHECK-NEXT:    addsd %xmm1, %xmm1
+; CHECK-NEXT:  .LBB177_4: # %entry
+; CHECK-NEXT:    movq %rdx, %rax
+; CHECK-NEXT:    shrq %rax
+; CHECK-NEXT:    movl %edx, %ecx
+; CHECK-NEXT:    andl $1, %ecx
+; CHECK-NEXT:    orq %rax, %rcx
+; CHECK-NEXT:    testq %rdx, %rdx
+; CHECK-NEXT:    cmovnsq %rdx, %rcx
+; CHECK-NEXT:    cvtsi2sd %rcx, %xmm2
+; CHECK-NEXT:    jns .LBB177_6
+; CHECK-NEXT:  # %bb.5:
+; CHECK-NEXT:    addsd %xmm2, %xmm2
+; CHECK-NEXT:  .LBB177_6: # %entry
+; CHECK-NEXT:    movsd %xmm2, -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    fldl -{{[0-9]+}}(%rsp)
 ; CHECK-NEXT:    wait
 ; CHECK-NEXT:    retq
 ;
 ; AVX1-LABEL: constrained_vector_uitofp_v3f64_v3i64:
 ; AVX1:       # %bb.0: # %entry
-; AVX1-NEXT:    vmovapd {{.*#+}} xmm1 = [1127219200,1160773632,0,0]
-; AVX1-NEXT:    vunpcklps {{.*#+}} xmm2 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX1-NEXT:    vmovapd {{.*#+}} xmm3 = [4.503599627370496E+15,1.9342813113834067E+25]
-; AVX1-NEXT:    vsubpd %xmm3, %xmm2, %xmm2
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm4 = xmm2[1,0]
-; AVX1-NEXT:    vaddpd %xmm2, %xmm4, %xmm2
-; AVX1-NEXT:    vpermilps {{.*#+}} xmm4 = xmm0[2,3,2,3]
-; AVX1-NEXT:    vunpcklps {{.*#+}} xmm4 = xmm4[0],xmm1[0],xmm4[1],xmm1[1]
-; AVX1-NEXT:    vsubpd %xmm3, %xmm4, %xmm4
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm5 = xmm4[1,0]
-; AVX1-NEXT:    vaddpd %xmm4, %xmm5, %xmm4
-; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm4[0]
+; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    shrq %rcx
+; AVX1-NEXT:    movl %eax, %edx
+; AVX1-NEXT:    andl $1, %edx
+; AVX1-NEXT:    orq %rcx, %rdx
+; AVX1-NEXT:    testq %rax, %rax
+; AVX1-NEXT:    cmovnsq %rax, %rdx
+; AVX1-NEXT:    vcvtsi2sd %rdx, %xmm1, %xmm1
+; AVX1-NEXT:    jns .LBB177_2
+; AVX1-NEXT:  # %bb.1:
+; AVX1-NEXT:    vaddsd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:  .LBB177_2: # %entry
+; AVX1-NEXT:    vmovq %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    shrq %rcx
+; AVX1-NEXT:    movl %eax, %edx
+; AVX1-NEXT:    andl $1, %edx
+; AVX1-NEXT:    orq %rcx, %rdx
+; AVX1-NEXT:    testq %rax, %rax
+; AVX1-NEXT:    cmovnsq %rax, %rdx
+; AVX1-NEXT:    vcvtsi2sd %rdx, %xmm2, %xmm2
+; AVX1-NEXT:    jns .LBB177_4
+; AVX1-NEXT:  # %bb.3:
+; AVX1-NEXT:    vaddsd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:  .LBB177_4: # %entry
+; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm2[0],xmm1[0]
 ; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
-; AVX1-NEXT:    vunpcklps {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1]
-; AVX1-NEXT:    vsubpd %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpermilpd {{.*#+}} xmm1 = xmm0[1,0]
-; AVX1-NEXT:    vaddpd %xmm0, %xmm1, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
+; AVX1-NEXT:    vmovq %xmm0, %rax
+; AVX1-NEXT:    movq %rax, %rcx
+; AVX1-NEXT:    shrq %rcx
+; AVX1-NEXT:    movl %eax, %edx
+; AVX1-NEXT:    andl $1, %edx
+; AVX1-NEXT:    orq %rcx, %rdx
+; AVX1-NEXT:    testq %rax, %rax
+; AVX1-NEXT:    cmovnsq %rax, %rdx
+; AVX1-NEXT:    vcvtsi2sd %rdx, %xmm3, %xmm0
+; AVX1-NEXT:    jns .LBB177_6
+; AVX1-NEXT:  # %bb.5:
+; AVX1-NEXT:    vaddsd %xmm0, %xmm0, %xmm0
+; AVX1-NEXT:  .LBB177_6: # %entry
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX512-LABEL: constrained_vector_uitofp_v3f64_v3i64:
@@ -7381,51 +7476,117 @@ entry:
 define <4 x double> @constrained_vector_uitofp_v4f64_v4i64(<4 x i64> %x) #0 {
 ; CHECK-LABEL: constrained_vector_uitofp_v4f64_v4i64:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    movdqa {{.*#+}} xmm2 = [4294967295,4294967295]
-; CHECK-NEXT:    movdqa %xmm1, %xmm3
-; CHECK-NEXT:    pand %xmm2, %xmm3
-; CHECK-NEXT:    movdqa {{.*#+}} xmm4 = [4841369599423283200,4841369599423283200]
-; CHECK-NEXT:    por %xmm4, %xmm3
-; CHECK-NEXT:    psrlq $32, %xmm1
-; CHECK-NEXT:    movdqa {{.*#+}} xmm5 = [4985484787499139072,4985484787499139072]
-; CHECK-NEXT:    por %xmm5, %xmm1
-; CHECK-NEXT:    movapd {{.*#+}} xmm6 = [1.9342813118337666E+25,1.9342813118337666E+25]
-; CHECK-NEXT:    subpd %xmm6, %xmm1
-; CHECK-NEXT:    addpd %xmm3, %xmm1
-; CHECK-NEXT:    pand %xmm0, %xmm2
-; CHECK-NEXT:    por %xmm4, %xmm2
-; CHECK-NEXT:    psrlq $32, %xmm0
-; CHECK-NEXT:    por %xmm5, %xmm0
-; CHECK-NEXT:    subpd %xmm6, %xmm0
-; CHECK-NEXT:    addpd %xmm2, %xmm0
+; CHECK-NEXT:    movdqa %xmm0, %xmm2
+; CHECK-NEXT:    movq %xmm0, %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    shrq %rcx
+; CHECK-NEXT:    movl %eax, %edx
+; CHECK-NEXT:    andl $1, %edx
+; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmovnsq %rax, %rdx
+; CHECK-NEXT:    xorps %xmm0, %xmm0
+; CHECK-NEXT:    cvtsi2sd %rdx, %xmm0
+; CHECK-NEXT:    jns .LBB181_2
+; CHECK-NEXT:  # %bb.1:
+; CHECK-NEXT:    addsd %xmm0, %xmm0
+; CHECK-NEXT:  .LBB181_2: # %entry
+; CHECK-NEXT:    pshufd {{.*#+}} xmm2 = xmm2[2,3,2,3]
+; CHECK-NEXT:    movq %xmm2, %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    shrq %rcx
+; CHECK-NEXT:    movl %eax, %edx
+; CHECK-NEXT:    andl $1, %edx
+; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmovnsq %rax, %rdx
+; CHECK-NEXT:    cvtsi2sd %rdx, %xmm3
+; CHECK-NEXT:    jns .LBB181_4
+; CHECK-NEXT:  # %bb.3:
+; CHECK-NEXT:    addsd %xmm3, %xmm3
+; CHECK-NEXT:  .LBB181_4: # %entry
+; CHECK-NEXT:    movq %xmm1, %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    shrq %rcx
+; CHECK-NEXT:    movl %eax, %edx
+; CHECK-NEXT:    andl $1, %edx
+; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmovnsq %rax, %rdx
+; CHECK-NEXT:    xorps %xmm2, %xmm2
+; CHECK-NEXT:    cvtsi2sd %rdx, %xmm2
+; CHECK-NEXT:    jns .LBB181_6
+; CHECK-NEXT:  # %bb.5:
+; CHECK-NEXT:    addsd %xmm2, %xmm2
+; CHECK-NEXT:  .LBB181_6: # %entry
+; CHECK-NEXT:    unpcklpd {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; CHECK-NEXT:    pshufd {{.*#+}} xmm1 = xmm1[2,3,2,3]
+; CHECK-NEXT:    movq %xmm1, %rax
+; CHECK-NEXT:    movq %rax, %rcx
+; CHECK-NEXT:    shrq %rcx
+; CHECK-NEXT:    movl %eax, %edx
+; CHECK-NEXT:    andl $1, %edx
+; CHECK-NEXT:    orq %rcx, %rdx
+; CHECK-NEXT:    testq %rax, %rax
+; CHECK-NEXT:    cmovnsq %rax, %rdx
+; CHECK-NEXT:    xorps %xmm1, %xmm1
+; CHECK-NEXT:    cvtsi2sd %rdx, %xmm1
+; CHECK-NEXT:    jns .LBB181_8
+; CHECK-NEXT:  # %bb.7:
+; CHECK-NEXT:    addsd %xmm1, %xmm1
+; CHECK-NEXT:  .LBB181_8: # %entry
+; CHECK-NEXT:    unpcklpd {{.*#+}} xmm2 = xmm2[0],xmm1[0]
+; CHECK-NEXT:    movapd %xmm2, %xmm1
 ; CHECK-NEXT:    retq
 ;
 ; AVX1-LABEL: constrained_vector_uitofp_v4f64_v4i64:
 ; AVX1:       # %bb.0: # %entry
 ; AVX1-NEXT:    vxorps %xmm1, %xmm1, %xmm1
 ; AVX1-NEXT:    vblendps {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX1-NEXT:    vorps {{.*}}(%rip), %ymm1, %ymm1
-; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm2
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vextractf128 $1, %ymm1, %xmm2
+; AVX1-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm3, %xmm3
+; AVX1-NEXT:    vmovq %xmm2, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm2
+; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
+; AVX1-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX1-NEXT:    vmovq %xmm1, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm1
+; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm3[0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm1, %ymm1
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm2
+; AVX1-NEXT:    vpsrlq $32, %xmm2, %xmm2
+; AVX1-NEXT:    vpextrq $1, %xmm2, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX1-NEXT:    vmovq %xmm2, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm2
+; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm2 = xmm2[0],xmm3[0]
 ; AVX1-NEXT:    vpsrlq $32, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm2, %ymm0
-; AVX1-NEXT:    vorpd {{.*}}(%rip), %ymm0, %ymm0
-; AVX1-NEXT:    vsubpd {{.*}}(%rip), %ymm0, %ymm0
-; AVX1-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
+; AVX1-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm3
+; AVX1-NEXT:    vmovq %xmm0, %rax
+; AVX1-NEXT:    vcvtsi2sd %rax, %xmm4, %xmm0
+; AVX1-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm3[0]
+; AVX1-NEXT:    vinsertf128 $1, %xmm2, %ymm0, %ymm0
+; AVX1-NEXT:    vmulpd {{.*}}(%rip), %ymm0, %ymm0
+; AVX1-NEXT:    vaddpd %ymm1, %ymm0, %ymm0
 ; AVX1-NEXT:    retq
 ;
 ; AVX512F-LABEL: constrained_vector_uitofp_v4f64_v4i64:
 ; AVX512F:       # %bb.0: # %entry
-; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512F-NEXT:    vpblendd {{.*#+}} ymm1 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
-; AVX512F-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4841369599423283200,4841369599423283200,4841369599423283200,4841369599423283200]
-; AVX512F-NEXT:    vpor %ymm2, %ymm1, %ymm1
-; AVX512F-NEXT:    vpsrlq $32, %ymm0, %ymm0
-; AVX512F-NEXT:    vpbroadcastq {{.*#+}} ymm2 = [4985484787499139072,4985484787499139072,4985484787499139072,4985484787499139072]
-; AVX512F-NEXT:    vpor %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vbroadcastsd {{.*#+}} ymm2 = [1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25,1.9342813118337666E+25]
-; AVX512F-NEXT:    vsubpd %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vaddpd %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vextracti128 $1, %ymm0, %xmm1
+; AVX512F-NEXT:    vpextrq $1, %xmm1, %rax
+; AVX512F-NEXT:    vcvtusi2sd %rax, %xmm2, %xmm2
+; AVX512F-NEXT:    vmovq %xmm1, %rax
+; AVX512F-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm1
+; AVX512F-NEXT:    vunpcklpd {{.*#+}} xmm1 = xmm1[0],xmm2[0]
+; AVX512F-NEXT:    vpextrq $1, %xmm0, %rax
+; AVX512F-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm2
+; AVX512F-NEXT:    vmovq %xmm0, %rax
+; AVX512F-NEXT:    vcvtusi2sd %rax, %xmm3, %xmm0
+; AVX512F-NEXT:    vunpcklpd {{.*#+}} xmm0 = xmm0[0],xmm2[0]
+; AVX512F-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
 ; AVX512F-NEXT:    retq
 ;
 ; AVX512DQ-LABEL: constrained_vector_uitofp_v4f64_v4i64:
diff --git a/llvm/test/CodeGen/X86/vector-popcnt-128-ult-ugt.ll b/llvm/test/CodeGen/X86/vector-popcnt-128-ult-ugt.ll
new file mode 100644
index 0000000000000..09a15f2f29e2a
--- /dev/null
+++ b/llvm/test/CodeGen/X86/vector-popcnt-128-ult-ugt.ll
@@ -0,0 +1,44072 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+ssse3 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSSE3
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+sse4.1 | FileCheck %s --check-prefix=ALL --check-prefix=SSE --check-prefix=SSE41
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vpopcntdq | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX512VPOPCNTDQ
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vpopcntdq,+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX512VPOPCNTDQVL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bitalg | FileCheck %s --check-prefix=ALL --check-prefix=BITALG_NOVLX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bitalg,+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=BITALG
+
+
+define <16 x i8> @ult_0_v16i8(<16 x i8> %0) {
+; SSE-LABEL: ult_0_v16i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    xorps %xmm0, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: ult_0_v16i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_0_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_0_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_0_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_0_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    pcmpeqb %xmm0, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE2-NEXT:    pxor %xmm0, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_0_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    pcmpeqb %xmm0, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE3-NEXT:    pxor %xmm0, %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_0_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm1, %xmm1
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm1
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSSE3-NEXT:    pxor %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_0_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm1
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_0_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_1_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_1_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    pcmpeqb %xmm0, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_1_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    pcmpeqb %xmm0, %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_1_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_1_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_1_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_1_v16i8(<16 x i8> %0) {
+; SSE-LABEL: ugt_1_v16i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    paddb %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pcmpeqb %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: ugt_1_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddb %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddb %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_2_v16i8(<16 x i8> %0) {
+; SSE-LABEL: ult_2_v16i8:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE-NEXT:    paddb %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pcmpeqb %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: ult_2_v16i8:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_2_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_2_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_2_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_2_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_2_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_2_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_3_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_3_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_3_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_3_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_3_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_3_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_3_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_3_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_3_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_3_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_3_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_3_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_4_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_4_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_4_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_4_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_4_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_4_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_4_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_4_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_4_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_4_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_4_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_4_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_5_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_5_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_5_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_5_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_5_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_5_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_5_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_5_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_5_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_5_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_5_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_5_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_6_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_6_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_6_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_6_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_6_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_6_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_6_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_6_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_6_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_6_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_6_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_6_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_7_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_7_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_7_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_7_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_7_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_7_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_7_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_7_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_7_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_7_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_7_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_7_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_8_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_8_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_8_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_8_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_8_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_8_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ugt_8_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ugt_8_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_8_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pcmpgtb {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_8_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; SSSE3-NEXT:    pmaxub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_8_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; SSE41-NEXT:    pmaxub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_8_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ugt <16 x i8> %2, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <16 x i8> @ult_9_v16i8(<16 x i8> %0) {
+; SSE2-LABEL: ult_9_v16i8:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; SSE2-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_9_v16i8:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; SSE3-NEXT:    pcmpgtb %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_9_v16i8:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSSE3-NEXT:    pminub %xmm3, %xmm0
+; SSSE3-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_9_v16i8:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; SSE41-NEXT:    pminub %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqb %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_9_v16i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v16i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %xmm0, %xmm1
+; AVX2-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v16i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v16i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxbd {{.*#+}} zmm0 = xmm0[0],zero,zero,zero,xmm0[1],zero,zero,zero,xmm0[2],zero,zero,zero,xmm0[3],zero,zero,zero,xmm0[4],zero,zero,zero,xmm0[5],zero,zero,zero,xmm0[6],zero,zero,zero,xmm0[7],zero,zero,zero,xmm0[8],zero,zero,zero,xmm0[9],zero,zero,zero,xmm0[10],zero,zero,zero,xmm0[11],zero,zero,zero,xmm0[12],zero,zero,zero,xmm0[13],zero,zero,zero,xmm0[14],zero,zero,zero,xmm0[15],zero,zero,zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdb %zmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v16i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v16i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
+  %3 = icmp ult <16 x i8> %2, <i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9>
+  %4 = sext <16 x i1> %3 to <16 x i8>
+  ret <16 x i8> %4
+}
+
+define <8 x i16> @ult_0_v8i16(<8 x i16> %0) {
+; SSE-LABEL: ult_0_v8i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    xorps %xmm0, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: ult_0_v8i16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_0_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_0_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_0_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_0_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE2-NEXT:    pxor %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_0_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pxor %xmm1, %xmm1
+; SSE3-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE3-NEXT:    pxor %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_0_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pxor %xmm1, %xmm1
+; SSSE3-NEXT:    pcmpeqw %xmm0, %xmm1
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSSE3-NEXT:    pxor %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_0_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pcmpeqw %xmm0, %xmm1
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_0_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_1_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_1_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm1
+; SSE2-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_1_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pxor %xmm1, %xmm1
+; SSE3-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_1_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_1_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_1_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_1_v8i16(<8 x i16> %0) {
+; SSE-LABEL: ugt_1_v8i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    paddw %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pcmpeqw %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: ugt_1_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddw %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpeqw %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqw %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpaddw %xmm1, %xmm0, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpaddw %xmm1, %xmm0, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_2_v8i16(<8 x i16> %0) {
+; SSE-LABEL: ult_2_v8i16:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE-NEXT:    paddw %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pcmpeqw %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: ult_2_v8i16:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm1
+; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_2_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_2_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_2_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_2_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_2_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_2_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_3_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_3_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_3_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_3_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_3_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3,3,3,3,3]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_3_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_3_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_3_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_3_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_3_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_3_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_3_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_4_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_4_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_4_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_4_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_4_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4,4,4,4,4]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_4_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_4_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_4_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_4_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_4_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_4_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_4_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_5_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_5_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_5_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_5_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_5_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5,5,5,5,5]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_5_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_5_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_5_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_5_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_5_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_5_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_5_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_6_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_6_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_6_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_6_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_6_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6,6,6,6,6]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_6_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_6_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_6_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_6_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_6_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_6_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_6_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_7_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_7_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_7_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_7_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_7_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7,7,7,7,7]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_7_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_7_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_7_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_7_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_7_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_7_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_7_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_8_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_8_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_8_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_8_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_8_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8,8,8,8,8]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_8_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_8_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_8_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_8_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_8_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_8_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_8_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_9_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_9_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_9_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_9_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_9_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9,9,9,9,9]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_9_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_9_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_9_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_9_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_9_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_9_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_9_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_9_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_9_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_9_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_10_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_10_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10,10,10,10,10]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_10_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10,10,10,10,10]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_10_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10,10,10,10,10]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_10_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10,10,10,10,10]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_10_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_10_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_10_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_10_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_10_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_10_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_10_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_10_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_10_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_10_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_10_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_10_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_10_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_11_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_11_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11,11,11,11,11]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_11_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11,11,11,11,11]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_11_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11,11,11,11,11]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_11_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11,11,11,11,11]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_11_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_11_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_11_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_11_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_11_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_11_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_11_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_11_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_11_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_11_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_11_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_11_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_11_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_12_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_12_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12,12,12,12,12]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_12_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12,12,12,12,12]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_12_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12,12,12,12,12]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_12_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12,12,12,12,12]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_12_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_12_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_12_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_12_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_12_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_12_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_12_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_12_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_12_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_12_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_12_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_12_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_12_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_13_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_13_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13,13,13,13,13]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_13_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13,13,13,13,13]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_13_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13,13,13,13,13]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_13_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13,13,13,13,13]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_13_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_13_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_13_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_13_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_13_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_13_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_13_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_13_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_13_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_13_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_13_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_13_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_13_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_14_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_14_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14,14,14,14,14]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_14_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14,14,14,14,14]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_14_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14,14,14,14,14]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_14_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14,14,14,14,14]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_14_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_14_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_14_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_14_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_14_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_14_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_14_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_14_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_14_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_14_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_14_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_14_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_14_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_15_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_15_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15,15,15,15,15]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_15_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15,15,15,15,15]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_15_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_15_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_15_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_15_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_15_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_15_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_15_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_15_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_15_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_15_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_15_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_15_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_15_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_15_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_15_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_16_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_16_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16,16,16,16,16]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_16_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16,16,16,16,16]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_16_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16,16,16,16,16]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_16_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16,16,16,16,16]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_16_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_16_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_16_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_16_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ugt_16_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ugt_16_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    psllw $8, %xmm0
+; SSE2-NEXT:    paddb %xmm1, %xmm0
+; SSE2-NEXT:    psrlw $8, %xmm0
+; SSE2-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_16_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    psllw $8, %xmm0
+; SSE3-NEXT:    paddb %xmm1, %xmm0
+; SSE3-NEXT:    psrlw $8, %xmm0
+; SSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_16_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm0
+; SSSE3-NEXT:    psllw $8, %xmm0
+; SSSE3-NEXT:    paddb %xmm3, %xmm0
+; SSSE3-NEXT:    psrlw $8, %xmm0
+; SSSE3-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_16_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm0
+; SSE41-NEXT:    psllw $8, %xmm0
+; SSE41-NEXT:    paddb %xmm3, %xmm0
+; SSE41-NEXT:    psrlw $8, %xmm0
+; SSE41-NEXT:    pcmpgtw {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_16_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_16_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_16_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_16_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ugt <8 x i16> %2, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <8 x i16> @ult_17_v8i16(<8 x i16> %0) {
+; SSE2-LABEL: ult_17_v8i16:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    psllw $8, %xmm2
+; SSE2-NEXT:    paddb %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $8, %xmm2
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17,17,17,17,17]
+; SSE2-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_17_v8i16:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    psllw $8, %xmm2
+; SSE3-NEXT:    paddb %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $8, %xmm2
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17,17,17,17,17]
+; SSE3-NEXT:    pcmpgtw %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_17_v8i16:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    psllw $8, %xmm1
+; SSSE3-NEXT:    paddb %xmm3, %xmm1
+; SSSE3-NEXT:    psrlw $8, %xmm1
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17,17,17,17,17]
+; SSSE3-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_17_v8i16:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    movdqa %xmm3, %xmm1
+; SSE41-NEXT:    psllw $8, %xmm1
+; SSE41-NEXT:    paddb %xmm3, %xmm1
+; SSE41-NEXT:    psrlw $8, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17,17,17,17,17]
+; SSE41-NEXT:    pcmpgtw %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_17_v8i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17,17,17,17,17]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_17_v8i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX2-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17,17,17,17,17]
+; AVX2-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v8i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_17_v8i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} ymm0 = xmm0[0],zero,xmm0[1],zero,xmm0[2],zero,xmm0[3],zero,xmm0[4],zero,xmm0[5],zero,xmm0[6],zero,xmm0[7],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %ymm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vzeroupper
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_17_v8i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17,17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v8i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
+  %3 = icmp ult <8 x i16> %2, <i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17>
+  %4 = sext <8 x i1> %3 to <8 x i16>
+  ret <8 x i16> %4
+}
+
+define <4 x i32> @ult_0_v4i32(<4 x i32> %0) {
+; SSE-LABEL: ult_0_v4i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    xorps %xmm0, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: ult_0_v4i32:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_0_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_0_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 0, i32 0, i32 0, i32 0>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_0_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_0_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE2-NEXT:    pxor %xmm0, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_0_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE3-NEXT:    pxor %xmm0, %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_0_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm3
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSSE3-NEXT:    pxor %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_0_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_0_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 0, i32 0, i32 0, i32 0>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_1_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_1_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_1_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_1_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_1_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_1_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 1, i32 1, i32 1, i32 1>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_1_v4i32(<4 x i32> %0) {
+; SSE-LABEL: ugt_1_v4i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE-NEXT:    movdqa %xmm0, %xmm1
+; SSE-NEXT:    paddd %xmm2, %xmm1
+; SSE-NEXT:    pand %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm0, %xmm0
+; SSE-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE-NEXT:    pxor %xmm2, %xmm1
+; SSE-NEXT:    movdqa %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: ugt_1_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqd %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 1, i32 1, i32 1, i32 1>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_2_v4i32(<4 x i32> %0) {
+; SSE-LABEL: ult_2_v4i32:
+; SSE:       # %bb.0:
+; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE-NEXT:    paddd %xmm0, %xmm1
+; SSE-NEXT:    pand %xmm1, %xmm0
+; SSE-NEXT:    pxor %xmm1, %xmm1
+; SSE-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX1-LABEL: ult_2_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_2_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [2,2,2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_2_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
+; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 2, i32 2, i32 2, i32 2>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_2_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_2_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_2_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_2_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_2_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_2_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [2,2,2,2]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [2,2,2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 2, i32 2, i32 2, i32 2>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_3_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_3_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_3_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_3_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_3_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [3,3,3,3]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_3_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3,3,3,3]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 3, i32 3, i32 3, i32 3>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_3_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_3_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_3_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_3_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_3_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_3_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3,3,3,3]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 3, i32 3, i32 3, i32 3>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_4_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_4_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_4_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_4_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_4_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [4,4,4,4]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_4_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 4, i32 4, i32 4, i32 4>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_4_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_4_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_4_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_4_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_4_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_4_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 4, i32 4, i32 4, i32 4>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_5_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_5_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_5_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_5_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_5_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [5,5,5,5]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_5_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [5,5,5,5]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 5, i32 5, i32 5, i32 5>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_5_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_5_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_5_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_5_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_5_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_5_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [5,5,5,5]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 5, i32 5, i32 5, i32 5>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_6_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_6_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_6_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_6_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_6_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [6,6,6,6]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_6_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [6,6,6,6]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 6, i32 6, i32 6, i32 6>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_6_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_6_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_6_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_6_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_6_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_6_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [6,6,6,6]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 6, i32 6, i32 6, i32 6>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_7_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_7_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_7_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_7_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_7_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [7,7,7,7]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_7_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 7, i32 7, i32 7, i32 7>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_7_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_7_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_7_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_7_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_7_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_7_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 7, i32 7, i32 7, i32 7>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_8_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_8_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_8_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_8_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_8_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [8,8,8,8]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_8_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [8,8,8,8]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 8, i32 8, i32 8, i32 8>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_8_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_8_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_8_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_8_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_8_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_8_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [8,8,8,8]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 8, i32 8, i32 8, i32 8>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_9_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_9_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_9_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_9_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_9_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [9,9,9,9]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_9_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [9,9,9,9]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 9, i32 9, i32 9, i32 9>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_9_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_9_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_9_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_9_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_9_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_9_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_9_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [9,9,9,9]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_9_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_9_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 9, i32 9, i32 9, i32 9>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_10_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_10_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_10_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_10_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_10_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [10,10,10,10]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_10_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_10_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [10,10,10,10]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_10_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_10_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 10, i32 10, i32 10, i32 10>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_10_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_10_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_10_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_10_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_10_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_10_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_10_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [10,10,10,10]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_10_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_10_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 10, i32 10, i32 10, i32 10>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_11_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_11_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_11_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_11_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_11_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [11,11,11,11]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_11_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_11_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [11,11,11,11]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_11_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_11_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 11, i32 11, i32 11, i32 11>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_11_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_11_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_11_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_11_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_11_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_11_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_11_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [11,11,11,11]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_11_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_11_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 11, i32 11, i32 11, i32 11>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_12_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_12_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_12_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_12_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_12_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [12,12,12,12]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_12_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_12_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [12,12,12,12]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_12_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_12_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 12, i32 12, i32 12, i32 12>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_12_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_12_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_12_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_12_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_12_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_12_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_12_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [12,12,12,12]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_12_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_12_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 12, i32 12, i32 12, i32 12>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_13_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_13_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_13_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_13_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_13_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [13,13,13,13]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_13_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_13_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [13,13,13,13]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_13_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_13_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 13, i32 13, i32 13, i32 13>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_13_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_13_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_13_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_13_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_13_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_13_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_13_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [13,13,13,13]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_13_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_13_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 13, i32 13, i32 13, i32 13>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_14_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_14_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_14_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_14_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_14_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [14,14,14,14]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_14_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_14_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [14,14,14,14]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_14_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_14_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 14, i32 14, i32 14, i32 14>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_14_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_14_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_14_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_14_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_14_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_14_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_14_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [14,14,14,14]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_14_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_14_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 14, i32 14, i32 14, i32 14>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_15_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_15_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_15_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_15_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_15_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [15,15,15,15]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_15_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_15_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_15_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_15_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 15, i32 15, i32 15, i32 15>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_15_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_15_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_15_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_15_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_15_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_15_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_15_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_15_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_15_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 15, i32 15, i32 15, i32 15>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_16_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_16_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_16_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_16_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_16_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [16,16,16,16]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_16_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_16_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [16,16,16,16]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_16_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_16_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 16, i32 16, i32 16, i32 16>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_16_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_16_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_16_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_16_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_16_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_16_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_16_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [16,16,16,16]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_16_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_16_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 16, i32 16, i32 16, i32 16>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_17_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_17_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_17_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_17_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_17_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [17,17,17,17]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_17_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_17_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_17_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_17_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 17, i32 17, i32 17, i32 17>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_17_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_17_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_17_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_17_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_17_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_17_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_17_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_17_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_17_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_17_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_17_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 17, i32 17, i32 17, i32 17>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_18_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_18_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [18,18,18,18]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_18_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [18,18,18,18]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_18_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [18,18,18,18]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_18_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [18,18,18,18]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_18_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18,18,18]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_18_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [18,18,18,18]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_18_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [18,18,18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_18_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_18_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [18,18,18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_18_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 18, i32 18, i32 18, i32 18>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_18_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_18_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_18_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_18_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_18_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_18_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_18_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [18,18,18,18]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_18_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [18,18,18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_18_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_18_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [18,18,18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_18_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 18, i32 18, i32 18, i32 18>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_19_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_19_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [19,19,19,19]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_19_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [19,19,19,19]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_19_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [19,19,19,19]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_19_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [19,19,19,19]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_19_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19,19,19]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_19_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [19,19,19,19]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_19_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [19,19,19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_19_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_19_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [19,19,19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_19_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 19, i32 19, i32 19, i32 19>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_19_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_19_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_19_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_19_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_19_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_19_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_19_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [19,19,19,19]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_19_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [19,19,19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_19_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_19_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [19,19,19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_19_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 19, i32 19, i32 19, i32 19>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_20_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_20_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [20,20,20,20]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_20_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [20,20,20,20]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_20_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [20,20,20,20]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_20_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [20,20,20,20]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_20_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20,20,20]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_20_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [20,20,20,20]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_20_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [20,20,20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_20_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_20_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [20,20,20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_20_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 20, i32 20, i32 20, i32 20>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_20_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_20_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_20_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_20_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_20_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_20_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_20_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [20,20,20,20]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_20_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [20,20,20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_20_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_20_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [20,20,20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_20_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 20, i32 20, i32 20, i32 20>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_21_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_21_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [21,21,21,21]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_21_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [21,21,21,21]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_21_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [21,21,21,21]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_21_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [21,21,21,21]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_21_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21,21,21]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_21_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [21,21,21,21]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_21_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [21,21,21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_21_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_21_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [21,21,21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_21_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 21, i32 21, i32 21, i32 21>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_21_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_21_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_21_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_21_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_21_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_21_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_21_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [21,21,21,21]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_21_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [21,21,21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_21_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_21_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [21,21,21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_21_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 21, i32 21, i32 21, i32 21>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_22_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_22_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [22,22,22,22]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_22_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [22,22,22,22]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_22_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [22,22,22,22]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_22_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [22,22,22,22]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_22_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22,22,22]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_22_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [22,22,22,22]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_22_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [22,22,22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_22_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_22_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [22,22,22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_22_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 22, i32 22, i32 22, i32 22>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_22_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_22_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_22_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_22_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_22_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_22_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_22_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [22,22,22,22]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_22_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [22,22,22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_22_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_22_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [22,22,22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_22_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 22, i32 22, i32 22, i32 22>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_23_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_23_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [23,23,23,23]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_23_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [23,23,23,23]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_23_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [23,23,23,23]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_23_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [23,23,23,23]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_23_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23,23,23]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_23_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [23,23,23,23]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_23_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [23,23,23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_23_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_23_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [23,23,23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_23_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 23, i32 23, i32 23, i32 23>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_23_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_23_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_23_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_23_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_23_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_23_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_23_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [23,23,23,23]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_23_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [23,23,23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_23_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_23_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [23,23,23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_23_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 23, i32 23, i32 23, i32 23>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_24_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_24_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [24,24,24,24]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_24_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [24,24,24,24]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_24_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [24,24,24,24]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_24_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [24,24,24,24]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_24_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24,24,24]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_24_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [24,24,24,24]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_24_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [24,24,24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_24_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_24_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [24,24,24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_24_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 24, i32 24, i32 24, i32 24>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_24_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_24_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_24_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_24_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_24_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_24_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_24_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [24,24,24,24]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_24_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [24,24,24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_24_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_24_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [24,24,24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_24_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 24, i32 24, i32 24, i32 24>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_25_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_25_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [25,25,25,25]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_25_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [25,25,25,25]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_25_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [25,25,25,25]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_25_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [25,25,25,25]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_25_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25,25,25]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_25_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [25,25,25,25]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_25_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [25,25,25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_25_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_25_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [25,25,25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_25_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 25, i32 25, i32 25, i32 25>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_25_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_25_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_25_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_25_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_25_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_25_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_25_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [25,25,25,25]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_25_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [25,25,25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_25_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_25_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [25,25,25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_25_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 25, i32 25, i32 25, i32 25>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_26_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_26_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [26,26,26,26]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_26_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [26,26,26,26]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_26_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [26,26,26,26]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_26_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [26,26,26,26]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_26_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26,26,26]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_26_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [26,26,26,26]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_26_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [26,26,26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_26_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_26_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [26,26,26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_26_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 26, i32 26, i32 26, i32 26>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_26_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_26_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_26_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_26_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_26_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_26_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_26_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [26,26,26,26]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_26_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [26,26,26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_26_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_26_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [26,26,26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_26_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 26, i32 26, i32 26, i32 26>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_27_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_27_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [27,27,27,27]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_27_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [27,27,27,27]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_27_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [27,27,27,27]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_27_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [27,27,27,27]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_27_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27,27,27]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_27_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [27,27,27,27]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_27_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [27,27,27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_27_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_27_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [27,27,27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_27_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 27, i32 27, i32 27, i32 27>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_27_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_27_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_27_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_27_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_27_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_27_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_27_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [27,27,27,27]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_27_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [27,27,27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_27_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_27_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [27,27,27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_27_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 27, i32 27, i32 27, i32 27>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_28_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_28_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [28,28,28,28]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_28_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [28,28,28,28]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_28_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [28,28,28,28]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_28_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [28,28,28,28]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_28_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28,28,28]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_28_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [28,28,28,28]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_28_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [28,28,28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_28_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_28_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [28,28,28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_28_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 28, i32 28, i32 28, i32 28>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_28_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_28_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_28_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_28_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_28_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_28_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_28_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [28,28,28,28]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_28_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [28,28,28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_28_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_28_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [28,28,28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_28_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 28, i32 28, i32 28, i32 28>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_29_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_29_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [29,29,29,29]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_29_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [29,29,29,29]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_29_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [29,29,29,29]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_29_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [29,29,29,29]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_29_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29,29,29]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_29_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [29,29,29,29]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_29_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [29,29,29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_29_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_29_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [29,29,29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_29_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 29, i32 29, i32 29, i32 29>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_29_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_29_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_29_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_29_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_29_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_29_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_29_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [29,29,29,29]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_29_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [29,29,29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_29_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_29_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [29,29,29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_29_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 29, i32 29, i32 29, i32 29>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_30_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_30_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [30,30,30,30]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_30_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [30,30,30,30]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_30_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [30,30,30,30]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_30_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [30,30,30,30]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_30_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30,30,30]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_30_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [30,30,30,30]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_30_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [30,30,30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_30_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_30_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [30,30,30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_30_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 30, i32 30, i32 30, i32 30>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_30_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_30_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_30_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_30_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_30_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_30_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_30_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [30,30,30,30]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_30_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [30,30,30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_30_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_30_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [30,30,30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_30_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 30, i32 30, i32 30, i32 30>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_31_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_31_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [31,31,31,31]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_31_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [31,31,31,31]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_31_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [31,31,31,31]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_31_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [31,31,31,31]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_31_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31,31,31]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_31_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [31,31,31,31]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_31_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [31,31,31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_31_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_31_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [31,31,31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_31_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 31, i32 31, i32 31, i32 31>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_31_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_31_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_31_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_31_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_31_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_31_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_31_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [31,31,31,31]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_31_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [31,31,31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_31_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_31_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [31,31,31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_31_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 31, i32 31, i32 31, i32 31>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_32_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_32_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [32,32,32,32]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_32_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [32,32,32,32]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_32_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [32,32,32,32]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_32_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [32,32,32,32]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_32_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32,32,32]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_32_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [32,32,32,32]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_32_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [32,32,32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_32_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_32_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [32,32,32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_32_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 32, i32 32, i32 32, i32 32>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ugt_32_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ugt_32_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_32_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_32_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm3
+; SSSE3-NEXT:    pand %xmm2, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm4
+; SSSE3-NEXT:    pshufb %xmm3, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm1
+; SSSE3-NEXT:    paddb %xmm4, %xmm1
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSSE3-NEXT:    pcmpgtd {{.*}}(%rip), %xmm1
+; SSSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_32_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm0 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm1[2],xmm3[3],xmm1[3]
+; SSE41-NEXT:    psadbw %xmm1, %xmm3
+; SSE41-NEXT:    psadbw %xmm1, %xmm0
+; SSE41-NEXT:    packuswb %xmm3, %xmm0
+; SSE41-NEXT:    pcmpgtd {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_32_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_32_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [32,32,32,32]
+; AVX2-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_32_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [32,32,32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_32_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_32_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [32,32,32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_32_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ugt <4 x i32> %2, <i32 32, i32 32, i32 32, i32 32>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <4 x i32> @ult_33_v4i32(<4 x i32> %0) {
+; SSE2-LABEL: ult_33_v4i32:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE2-NEXT:    psadbw %xmm0, %xmm2
+; SSE2-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    packuswb %xmm2, %xmm1
+; SSE2-NEXT:    movdqa {{.*#+}} xmm0 = [33,33,33,33]
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_33_v4i32:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    punpckhdq {{.*#+}} xmm2 = xmm2[2],xmm0[2],xmm2[3],xmm0[3]
+; SSE3-NEXT:    psadbw %xmm0, %xmm2
+; SSE3-NEXT:    punpckldq {{.*#+}} xmm1 = xmm1[0],xmm0[0],xmm1[1],xmm0[1]
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    packuswb %xmm2, %xmm1
+; SSE3-NEXT:    movdqa {{.*#+}} xmm0 = [33,33,33,33]
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_33_v4i32:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    movdqa %xmm3, %xmm1
+; SSSE3-NEXT:    punpckhdq {{.*#+}} xmm1 = xmm1[2],xmm0[2],xmm1[3],xmm0[3]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSSE3-NEXT:    punpckldq {{.*#+}} xmm3 = xmm3[0],xmm0[0],xmm3[1],xmm0[1]
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    packuswb %xmm1, %xmm3
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm0 = [33,33,33,33]
+; SSSE3-NEXT:    pcmpgtd %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_33_v4i32:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pmovzxdq {{.*#+}} xmm1 = xmm3[0],zero,xmm3[1],zero
+; SSE41-NEXT:    punpckhdq {{.*#+}} xmm3 = xmm3[2],xmm0[2],xmm3[3],xmm0[3]
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    packuswb %xmm3, %xmm1
+; SSE41-NEXT:    movdqa {{.*#+}} xmm0 = [33,33,33,33]
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_33_v4i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33,33,33]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_33_v4i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; AVX2-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [33,33,33,33]
+; AVX2-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_33_v4i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [33,33,33,33]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_33_v4i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_33_v4i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG_NOVLX-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [33,33,33,33]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_33_v4i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} xmm2 = xmm0[2],xmm1[2],xmm0[3],xmm1[3]
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm2, %xmm2
+; BITALG-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpackuswb %xmm2, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
+  %3 = icmp ult <4 x i32> %2, <i32 33, i32 33, i32 33, i32 33>
+  %4 = sext <4 x i1> %3 to <4 x i32>
+  ret <4 x i32> %4
+}
+
+define <2 x i64> @ult_0_v2i64(<2 x i64> %0) {
+; SSE-LABEL: ult_0_v2i64:
+; SSE:       # %bb.0:
+; SSE-NEXT:    xorps %xmm0, %xmm0
+; SSE-NEXT:    retq
+;
+; AVX-LABEL: ult_0_v2i64:
+; AVX:       # %bb.0:
+; AVX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_0_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_0_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 0, i64 0>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_0_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_0_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
+; SSE2-NEXT:    pand %xmm0, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE2-NEXT:    pxor %xmm0, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_0_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
+; SSE3-NEXT:    pand %xmm0, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE3-NEXT:    pxor %xmm0, %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_0_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm3
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm3[1,0,3,2]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSSE3-NEXT:    pxor %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_0_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm0, %xmm3
+; SSE41-NEXT:    pcmpeqq %xmm0, %xmm3
+; SSE41-NEXT:    pcmpeqd %xmm0, %xmm0
+; SSE41-NEXT:    pxor %xmm3, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_0_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 0, i64 0>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_1_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_1_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm0, %xmm1
+; SSE2-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
+; SSE2-NEXT:    pand %xmm0, %xmm1
+; SSE2-NEXT:    movdqa %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_1_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm0, %xmm1
+; SSE3-NEXT:    pcmpeqd %xmm0, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm1[1,0,3,2]
+; SSE3-NEXT:    pand %xmm0, %xmm1
+; SSE3-NEXT:    movdqa %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_1_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm0, %xmm3
+; SSSE3-NEXT:    pcmpeqd %xmm0, %xmm3
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSSE3-NEXT:    pand %xmm3, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_1_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm3
+; SSE41-NEXT:    pand %xmm2, %xmm3
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm1, %xmm4
+; SSE41-NEXT:    pshufb %xmm3, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm2, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm1
+; SSE41-NEXT:    paddb %xmm4, %xmm1
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm0, %xmm1
+; SSE41-NEXT:    pcmpeqq %xmm0, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_1_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 1, i64 1>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_1_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_1_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    paddq %xmm1, %xmm2
+; SSE2-NEXT:    pand %xmm0, %xmm2
+; SSE2-NEXT:    pxor %xmm3, %xmm3
+; SSE2-NEXT:    pcmpeqd %xmm2, %xmm3
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSE2-NEXT:    pand %xmm3, %xmm0
+; SSE2-NEXT:    pxor %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_1_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    paddq %xmm1, %xmm2
+; SSE3-NEXT:    pand %xmm0, %xmm2
+; SSE3-NEXT:    pxor %xmm3, %xmm3
+; SSE3-NEXT:    pcmpeqd %xmm2, %xmm3
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSE3-NEXT:    pand %xmm3, %xmm0
+; SSE3-NEXT:    pxor %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_1_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    paddq %xmm1, %xmm2
+; SSSE3-NEXT:    pand %xmm0, %xmm2
+; SSSE3-NEXT:    pxor %xmm3, %xmm3
+; SSSE3-NEXT:    pcmpeqd %xmm2, %xmm3
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
+; SSSE3-NEXT:    pand %xmm3, %xmm0
+; SSSE3-NEXT:    pxor %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_1_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    pcmpeqd %xmm2, %xmm2
+; SSE41-NEXT:    movdqa %xmm0, %xmm1
+; SSE41-NEXT:    paddq %xmm2, %xmm1
+; SSE41-NEXT:    pand %xmm0, %xmm1
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    pcmpeqq %xmm0, %xmm1
+; SSE41-NEXT:    pxor %xmm2, %xmm1
+; SSE41-NEXT:    movdqa %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_1_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpcmpeqq %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddq %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqq %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
+; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
+; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 1, i64 1>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_2_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_2_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE2-NEXT:    paddq %xmm0, %xmm1
+; SSE2-NEXT:    pand %xmm0, %xmm1
+; SSE2-NEXT:    pxor %xmm2, %xmm2
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
+; SSE2-NEXT:    pand %xmm2, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_2_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE3-NEXT:    paddq %xmm0, %xmm1
+; SSE3-NEXT:    pand %xmm0, %xmm1
+; SSE3-NEXT:    pxor %xmm2, %xmm2
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
+; SSE3-NEXT:    pand %xmm2, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_2_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSSE3-NEXT:    paddq %xmm0, %xmm1
+; SSSE3-NEXT:    pand %xmm0, %xmm1
+; SSSE3-NEXT:    pxor %xmm2, %xmm2
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
+; SSSE3-NEXT:    pand %xmm2, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_2_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
+; SSE41-NEXT:    paddq %xmm0, %xmm1
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pxor %xmm1, %xmm1
+; SSE41-NEXT:    pcmpeqq %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_2_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_2_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_2_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
+; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
+; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 2, i64 2>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_2_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_2_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483650,2147483650]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_2_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483650,2147483650]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_2_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483650,2147483650]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_2_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483650,2147483650]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_2_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 2, i64 2>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_3_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_3_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_3_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_3_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_3_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_3_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 3, i64 3>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_3_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_3_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_3_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_3_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_3_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483651,2147483651]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_3_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 3, i64 3>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_4_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_4_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_4_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_4_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_4_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_4_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 4, i64 4>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_4_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_4_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_4_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_4_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_4_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483652,2147483652]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_4_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 4, i64 4>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_5_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_5_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_5_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_5_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_5_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_5_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 5, i64 5>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_5_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_5_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_5_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_5_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_5_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483653,2147483653]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_5_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 5, i64 5>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_6_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_6_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_6_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_6_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_6_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_6_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 6, i64 6>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_6_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_6_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_6_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_6_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_6_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483654,2147483654]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_6_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 6, i64 6>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_7_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_7_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_7_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_7_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_7_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_7_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 7, i64 7>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_7_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_7_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_7_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_7_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_7_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483655,2147483655]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_7_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 7, i64 7>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_8_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_8_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_8_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_8_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_8_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_8_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 8, i64 8>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_8_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_8_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_8_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_8_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_8_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483656,2147483656]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_8_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 8, i64 8>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_9_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_9_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_9_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_9_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_9_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_9_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 9, i64 9>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_9_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_9_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_9_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_9_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_9_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483657,2147483657]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_9_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_9_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_9_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_9_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 9, i64 9>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_10_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_10_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_10_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_10_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_10_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_10_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_10_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_10_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_10_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 10, i64 10>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_10_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_10_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_10_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_10_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_10_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483658,2147483658]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_10_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_10_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_10_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_10_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 10, i64 10>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_11_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_11_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_11_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_11_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_11_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_11_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_11_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_11_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_11_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 11, i64 11>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_11_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_11_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_11_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_11_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_11_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483659,2147483659]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_11_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_11_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_11_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_11_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 11, i64 11>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_12_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_12_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_12_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_12_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_12_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_12_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_12_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_12_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_12_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 12, i64 12>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_12_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_12_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_12_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_12_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_12_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483660,2147483660]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_12_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_12_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_12_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_12_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 12, i64 12>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_13_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_13_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_13_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_13_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_13_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_13_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_13_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_13_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_13_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 13, i64 13>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_13_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_13_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_13_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_13_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_13_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483661,2147483661]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_13_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_13_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_13_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_13_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 13, i64 13>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_14_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_14_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_14_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_14_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_14_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_14_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_14_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_14_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_14_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 14, i64 14>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_14_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_14_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_14_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_14_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_14_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483662,2147483662]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_14_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_14_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_14_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_14_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 14, i64 14>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_15_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_15_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_15_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_15_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_15_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_15_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_15_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_15_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_15_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 15, i64 15>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_15_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_15_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_15_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_15_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_15_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483663,2147483663]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_15_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_15_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_15_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_15_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 15, i64 15>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_16_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_16_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_16_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_16_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_16_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_16_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_16_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_16_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_16_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 16, i64 16>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_16_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_16_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_16_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_16_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_16_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483664,2147483664]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_16_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_16_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_16_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_16_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 16, i64 16>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_17_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_17_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_17_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_17_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_17_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_17_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_17_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_17_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_17_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 17, i64 17>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_17_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_17_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_17_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_17_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_17_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483665,2147483665]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_17_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_17_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_17_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_17_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_17_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_17_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 17, i64 17>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_18_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_18_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_18_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_18_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_18_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_18_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_18_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_18_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_18_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_18_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_18_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 18, i64 18>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_18_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_18_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_18_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_18_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_18_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483666,2147483666]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_18_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_18_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_18_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_18_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_18_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_18_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 18, i64 18>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_19_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_19_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_19_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_19_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_19_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_19_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_19_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_19_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_19_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_19_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_19_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 19, i64 19>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_19_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_19_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_19_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_19_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_19_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483667,2147483667]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_19_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_19_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_19_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_19_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_19_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_19_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 19, i64 19>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_20_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_20_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_20_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_20_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_20_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_20_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_20_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_20_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_20_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_20_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_20_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 20, i64 20>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_20_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_20_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_20_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_20_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_20_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483668,2147483668]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_20_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_20_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_20_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_20_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_20_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_20_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 20, i64 20>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_21_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_21_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_21_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_21_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_21_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_21_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_21_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_21_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_21_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_21_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_21_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 21, i64 21>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_21_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_21_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_21_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_21_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_21_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483669,2147483669]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_21_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_21_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_21_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_21_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_21_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_21_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 21, i64 21>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_22_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_22_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_22_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_22_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_22_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_22_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_22_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_22_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_22_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_22_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_22_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 22, i64 22>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_22_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_22_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_22_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_22_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_22_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483670,2147483670]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_22_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_22_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_22_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_22_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_22_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_22_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 22, i64 22>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_23_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_23_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_23_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_23_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_23_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_23_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_23_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_23_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_23_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_23_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_23_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 23, i64 23>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_23_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_23_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_23_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_23_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_23_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483671,2147483671]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_23_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_23_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_23_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_23_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_23_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_23_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 23, i64 23>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_24_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_24_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_24_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_24_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_24_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_24_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_24_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_24_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_24_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_24_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_24_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 24, i64 24>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_24_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_24_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_24_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_24_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_24_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483672,2147483672]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_24_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_24_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_24_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_24_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_24_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_24_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 24, i64 24>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_25_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_25_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_25_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_25_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_25_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_25_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_25_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_25_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_25_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_25_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_25_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 25, i64 25>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_25_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_25_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_25_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_25_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_25_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483673,2147483673]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_25_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_25_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_25_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_25_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_25_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_25_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 25, i64 25>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_26_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_26_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_26_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_26_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_26_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_26_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_26_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_26_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_26_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_26_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_26_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 26, i64 26>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_26_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_26_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_26_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_26_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_26_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483674,2147483674]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_26_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_26_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_26_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_26_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_26_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_26_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 26, i64 26>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_27_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_27_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_27_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_27_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_27_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_27_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_27_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_27_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_27_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_27_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_27_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 27, i64 27>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_27_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_27_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_27_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_27_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_27_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483675,2147483675]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_27_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_27_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_27_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_27_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_27_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_27_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 27, i64 27>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_28_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_28_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_28_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_28_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_28_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_28_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_28_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_28_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_28_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_28_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_28_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 28, i64 28>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_28_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_28_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_28_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_28_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_28_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483676,2147483676]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_28_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_28_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_28_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_28_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_28_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_28_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 28, i64 28>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_29_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_29_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_29_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_29_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_29_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_29_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_29_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_29_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_29_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_29_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_29_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 29, i64 29>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_29_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_29_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_29_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_29_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_29_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483677,2147483677]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_29_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_29_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_29_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_29_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_29_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_29_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 29, i64 29>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_30_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_30_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_30_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_30_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_30_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_30_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_30_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_30_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_30_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_30_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_30_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 30, i64 30>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_30_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_30_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_30_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_30_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_30_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483678,2147483678]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_30_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_30_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_30_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_30_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_30_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_30_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 30, i64 30>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_31_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_31_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_31_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_31_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_31_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_31_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_31_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_31_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_31_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_31_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_31_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 31, i64 31>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_31_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_31_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_31_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_31_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_31_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483679,2147483679]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_31_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_31_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_31_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_31_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_31_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_31_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 31, i64 31>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_32_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_32_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_32_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_32_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_32_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_32_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_32_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_32_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_32_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_32_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_32_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 32, i64 32>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_32_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_32_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_32_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_32_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_32_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483680,2147483680]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_32_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_32_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_32_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_32_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_32_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_32_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 32, i64 32>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_33_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_33_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_33_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_33_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_33_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_33_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_33_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_33_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_33_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_33_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_33_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 33, i64 33>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_33_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_33_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_33_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_33_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_33_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483681,2147483681]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_33_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_33_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_33_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_33_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_33_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_33_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 33, i64 33>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_34_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_34_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_34_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_34_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_34_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_34_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [34,34]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_34_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [34,34]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_34_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [34,34]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_34_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_34_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [34,34]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_34_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 34, i64 34>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_34_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_34_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_34_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_34_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_34_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483682,2147483682]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_34_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_34_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_34_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_34_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_34_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_34_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 34, i64 34>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_35_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_35_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_35_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_35_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_35_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_35_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [35,35]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_35_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [35,35]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_35_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [35,35]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_35_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_35_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [35,35]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_35_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 35, i64 35>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_35_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_35_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_35_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_35_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_35_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483683,2147483683]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_35_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_35_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_35_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_35_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_35_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_35_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 35, i64 35>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_36_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_36_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_36_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_36_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_36_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_36_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [36,36]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_36_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [36,36]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_36_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [36,36]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_36_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_36_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [36,36]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_36_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 36, i64 36>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_36_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_36_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_36_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_36_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_36_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483684,2147483684]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_36_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_36_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_36_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_36_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_36_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_36_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 36, i64 36>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_37_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_37_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_37_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_37_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_37_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_37_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [37,37]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_37_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [37,37]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_37_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [37,37]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_37_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_37_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [37,37]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_37_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 37, i64 37>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_37_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_37_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_37_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_37_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_37_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483685,2147483685]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_37_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_37_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_37_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_37_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_37_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_37_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 37, i64 37>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_38_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_38_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_38_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_38_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_38_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_38_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [38,38]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_38_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [38,38]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_38_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [38,38]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_38_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_38_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [38,38]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_38_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 38, i64 38>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_38_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_38_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_38_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_38_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_38_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483686,2147483686]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_38_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_38_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_38_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_38_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_38_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_38_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 38, i64 38>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_39_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_39_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_39_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_39_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_39_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_39_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [39,39]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_39_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [39,39]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_39_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [39,39]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_39_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_39_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [39,39]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_39_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 39, i64 39>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_39_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_39_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_39_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_39_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_39_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483687,2147483687]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_39_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_39_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_39_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_39_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_39_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_39_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 39, i64 39>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_40_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_40_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_40_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_40_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_40_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_40_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [40,40]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_40_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [40,40]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_40_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [40,40]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_40_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_40_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [40,40]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_40_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 40, i64 40>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_40_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_40_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_40_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_40_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_40_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483688,2147483688]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_40_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_40_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_40_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_40_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_40_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_40_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 40, i64 40>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_41_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_41_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_41_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_41_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_41_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_41_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [41,41]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_41_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [41,41]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_41_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [41,41]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_41_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_41_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [41,41]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_41_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 41, i64 41>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_41_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_41_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_41_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_41_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_41_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483689,2147483689]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_41_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_41_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_41_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_41_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_41_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_41_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 41, i64 41>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_42_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_42_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_42_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_42_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_42_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_42_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [42,42]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_42_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [42,42]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_42_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [42,42]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_42_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_42_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [42,42]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_42_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 42, i64 42>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_42_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_42_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_42_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_42_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_42_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483690,2147483690]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_42_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_42_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_42_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_42_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_42_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_42_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 42, i64 42>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_43_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_43_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_43_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_43_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_43_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_43_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [43,43]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_43_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [43,43]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_43_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [43,43]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_43_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_43_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [43,43]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_43_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 43, i64 43>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_43_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_43_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_43_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_43_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_43_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483691,2147483691]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_43_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_43_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_43_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_43_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_43_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_43_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 43, i64 43>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_44_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_44_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_44_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_44_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_44_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_44_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [44,44]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_44_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [44,44]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_44_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [44,44]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_44_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_44_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [44,44]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_44_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 44, i64 44>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_44_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_44_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_44_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_44_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_44_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483692,2147483692]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_44_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_44_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_44_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_44_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_44_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_44_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 44, i64 44>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_45_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_45_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_45_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_45_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_45_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_45_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [45,45]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_45_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [45,45]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_45_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [45,45]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_45_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_45_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [45,45]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_45_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 45, i64 45>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_45_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_45_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_45_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_45_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_45_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483693,2147483693]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_45_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_45_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_45_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_45_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_45_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_45_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 45, i64 45>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_46_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_46_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_46_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_46_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_46_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_46_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [46,46]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_46_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [46,46]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_46_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [46,46]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_46_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_46_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [46,46]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_46_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 46, i64 46>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_46_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_46_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_46_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_46_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_46_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483694,2147483694]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_46_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_46_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_46_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_46_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_46_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_46_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 46, i64 46>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_47_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_47_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_47_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_47_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_47_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_47_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [47,47]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_47_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [47,47]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_47_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [47,47]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_47_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_47_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [47,47]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_47_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 47, i64 47>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_47_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_47_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_47_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_47_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_47_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483695,2147483695]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_47_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_47_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_47_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_47_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_47_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_47_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 47, i64 47>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_48_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_48_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_48_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_48_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_48_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_48_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [48,48]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_48_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [48,48]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_48_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [48,48]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_48_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_48_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [48,48]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_48_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 48, i64 48>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_48_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_48_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_48_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_48_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_48_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483696,2147483696]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_48_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_48_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_48_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_48_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_48_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_48_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 48, i64 48>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_49_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_49_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_49_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_49_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_49_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_49_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [49,49]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_49_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [49,49]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_49_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [49,49]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_49_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_49_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [49,49]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_49_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 49, i64 49>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_49_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_49_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_49_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_49_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_49_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483697,2147483697]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_49_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_49_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_49_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_49_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_49_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_49_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 49, i64 49>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_50_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_50_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_50_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_50_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_50_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_50_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [50,50]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_50_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [50,50]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_50_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [50,50]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_50_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_50_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [50,50]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_50_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 50, i64 50>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_50_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_50_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_50_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_50_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_50_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483698,2147483698]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_50_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_50_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_50_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_50_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_50_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_50_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 50, i64 50>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_51_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_51_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_51_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_51_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_51_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_51_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [51,51]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_51_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [51,51]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_51_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [51,51]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_51_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_51_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [51,51]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_51_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 51, i64 51>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_51_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_51_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_51_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_51_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_51_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483699,2147483699]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_51_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_51_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_51_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_51_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_51_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_51_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 51, i64 51>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_52_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_52_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_52_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_52_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_52_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_52_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [52,52]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_52_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [52,52]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_52_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [52,52]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_52_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_52_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [52,52]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_52_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 52, i64 52>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_52_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_52_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_52_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_52_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_52_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483700,2147483700]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_52_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_52_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_52_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_52_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_52_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_52_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 52, i64 52>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_53_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_53_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_53_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_53_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_53_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_53_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [53,53]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_53_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [53,53]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_53_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [53,53]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_53_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_53_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [53,53]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_53_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 53, i64 53>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_53_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_53_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_53_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_53_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_53_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483701,2147483701]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_53_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_53_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_53_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_53_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_53_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_53_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 53, i64 53>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_54_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_54_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_54_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_54_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_54_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_54_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [54,54]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_54_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [54,54]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_54_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [54,54]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_54_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_54_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [54,54]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_54_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 54, i64 54>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_54_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_54_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_54_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_54_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_54_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483702,2147483702]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_54_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_54_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_54_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_54_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_54_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_54_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 54, i64 54>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_55_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_55_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_55_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_55_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_55_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_55_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [55,55]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_55_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [55,55]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_55_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [55,55]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_55_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_55_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [55,55]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_55_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 55, i64 55>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_55_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_55_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_55_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_55_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_55_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483703,2147483703]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_55_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_55_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_55_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_55_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_55_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_55_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 55, i64 55>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_56_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_56_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_56_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_56_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_56_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_56_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [56,56]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_56_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [56,56]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_56_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [56,56]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_56_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_56_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [56,56]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_56_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 56, i64 56>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_56_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_56_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_56_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_56_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_56_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483704,2147483704]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_56_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_56_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_56_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_56_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_56_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_56_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 56, i64 56>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_57_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_57_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_57_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_57_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_57_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_57_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [57,57]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_57_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [57,57]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_57_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [57,57]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_57_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_57_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [57,57]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_57_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 57, i64 57>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_57_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_57_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_57_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_57_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_57_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483705,2147483705]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_57_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_57_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_57_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_57_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_57_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_57_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 57, i64 57>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_58_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_58_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_58_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_58_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_58_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_58_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [58,58]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_58_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [58,58]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_58_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [58,58]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_58_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_58_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [58,58]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_58_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 58, i64 58>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_58_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_58_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_58_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_58_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_58_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483706,2147483706]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_58_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_58_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_58_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_58_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_58_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_58_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 58, i64 58>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_59_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_59_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_59_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_59_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_59_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_59_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [59,59]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_59_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [59,59]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_59_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [59,59]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_59_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_59_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [59,59]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_59_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 59, i64 59>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_59_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_59_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_59_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_59_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_59_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483707,2147483707]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_59_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_59_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_59_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_59_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_59_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_59_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 59, i64 59>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_60_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_60_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_60_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_60_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_60_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_60_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,60]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_60_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,60]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_60_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,60]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_60_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_60_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,60]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_60_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 60, i64 60>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_60_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_60_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_60_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_60_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_60_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483708,2147483708]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_60_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_60_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_60_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_60_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_60_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_60_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 60, i64 60>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_61_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_61_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_61_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_61_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_61_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_61_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [61,61]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_61_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [61,61]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_61_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [61,61]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_61_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_61_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [61,61]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_61_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 61, i64 61>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_61_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_61_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_61_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_61_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_61_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483709,2147483709]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_61_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_61_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_61_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_61_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_61_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_61_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 61, i64 61>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_62_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_62_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_62_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_62_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_62_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_62_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [62,62]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_62_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [62,62]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_62_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [62,62]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_62_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_62_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [62,62]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_62_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 62, i64 62>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_62_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_62_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_62_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_62_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_62_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483710,2147483710]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_62_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_62_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_62_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_62_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_62_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_62_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 62, i64 62>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_63_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_63_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_63_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_63_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_63_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_63_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [63,63]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_63_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [63,63]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_63_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [63,63]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_63_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_63_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [63,63]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_63_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 63, i64 63>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_63_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_63_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_63_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_63_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_63_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483711,2147483711]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_63_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_63_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_63_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_63_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_63_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_63_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 63, i64 63>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_64_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_64_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_64_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_64_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_64_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_64_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [64,64]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_64_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [64,64]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_64_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [64,64]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_64_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_64_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [64,64]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_64_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 64, i64 64>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ugt_64_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ugt_64_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ugt_64_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ugt_64_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ugt_64_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483712,2147483712]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm1, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ugt_64_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_64_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_64_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_64_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_64_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_64_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ugt <2 x i64> %2, <i64 64, i64 64>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+define <2 x i64> @ult_65_v2i64(<2 x i64> %0) {
+; SSE2-LABEL: ult_65_v2i64:
+; SSE2:       # %bb.0:
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $1, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    psubb %xmm1, %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE2-NEXT:    movdqa %xmm0, %xmm2
+; SSE2-NEXT:    pand %xmm1, %xmm2
+; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    pand %xmm1, %xmm0
+; SSE2-NEXT:    paddb %xmm2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $4, %xmm1
+; SSE2-NEXT:    paddb %xmm0, %xmm1
+; SSE2-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE2-NEXT:    pxor %xmm0, %xmm0
+; SSE2-NEXT:    psadbw %xmm1, %xmm0
+; SSE2-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [2147483713,2147483713]
+; SSE2-NEXT:    movdqa %xmm1, %xmm2
+; SSE2-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE2-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE2-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE2-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE2-NEXT:    pand %xmm3, %xmm1
+; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE2-NEXT:    por %xmm1, %xmm0
+; SSE2-NEXT:    retq
+;
+; SSE3-LABEL: ult_65_v2i64:
+; SSE3:       # %bb.0:
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $1, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    psubb %xmm1, %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
+; SSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSE3-NEXT:    pand %xmm1, %xmm2
+; SSE3-NEXT:    psrlw $2, %xmm0
+; SSE3-NEXT:    pand %xmm1, %xmm0
+; SSE3-NEXT:    paddb %xmm2, %xmm0
+; SSE3-NEXT:    movdqa %xmm0, %xmm1
+; SSE3-NEXT:    psrlw $4, %xmm1
+; SSE3-NEXT:    paddb %xmm0, %xmm1
+; SSE3-NEXT:    pand {{.*}}(%rip), %xmm1
+; SSE3-NEXT:    pxor %xmm0, %xmm0
+; SSE3-NEXT:    psadbw %xmm1, %xmm0
+; SSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483713,2147483713]
+; SSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE3-NEXT:    pand %xmm3, %xmm1
+; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE3-NEXT:    por %xmm1, %xmm0
+; SSE3-NEXT:    retq
+;
+; SSSE3-LABEL: ult_65_v2i64:
+; SSSE3:       # %bb.0:
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSSE3-NEXT:    movdqa %xmm0, %xmm2
+; SSSE3-NEXT:    pand %xmm1, %xmm2
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSSE3-NEXT:    movdqa %xmm3, %xmm4
+; SSSE3-NEXT:    pshufb %xmm2, %xmm4
+; SSSE3-NEXT:    psrlw $4, %xmm0
+; SSSE3-NEXT:    pand %xmm1, %xmm0
+; SSSE3-NEXT:    pshufb %xmm0, %xmm3
+; SSSE3-NEXT:    paddb %xmm4, %xmm3
+; SSSE3-NEXT:    pxor %xmm0, %xmm0
+; SSSE3-NEXT:    psadbw %xmm3, %xmm0
+; SSSE3-NEXT:    por {{.*}}(%rip), %xmm0
+; SSSE3-NEXT:    movdqa {{.*#+}} xmm1 = [2147483713,2147483713]
+; SSSE3-NEXT:    movdqa %xmm1, %xmm2
+; SSSE3-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSSE3-NEXT:    pand %xmm3, %xmm1
+; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSSE3-NEXT:    por %xmm1, %xmm0
+; SSSE3-NEXT:    retq
+;
+; SSE41-LABEL: ult_65_v2i64:
+; SSE41:       # %bb.0:
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; SSE41-NEXT:    movdqa %xmm0, %xmm2
+; SSE41-NEXT:    pand %xmm1, %xmm2
+; SSE41-NEXT:    movdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; SSE41-NEXT:    movdqa %xmm3, %xmm4
+; SSE41-NEXT:    pshufb %xmm2, %xmm4
+; SSE41-NEXT:    psrlw $4, %xmm0
+; SSE41-NEXT:    pand %xmm1, %xmm0
+; SSE41-NEXT:    pshufb %xmm0, %xmm3
+; SSE41-NEXT:    paddb %xmm4, %xmm3
+; SSE41-NEXT:    pxor %xmm0, %xmm0
+; SSE41-NEXT:    psadbw %xmm3, %xmm0
+; SSE41-NEXT:    por {{.*}}(%rip), %xmm0
+; SSE41-NEXT:    movdqa {{.*#+}} xmm1 = [2147483713,2147483713]
+; SSE41-NEXT:    movdqa %xmm1, %xmm2
+; SSE41-NEXT:    pcmpgtd %xmm0, %xmm2
+; SSE41-NEXT:    pshufd {{.*#+}} xmm3 = xmm2[0,0,2,2]
+; SSE41-NEXT:    pcmpeqd %xmm1, %xmm0
+; SSE41-NEXT:    pshufd {{.*#+}} xmm1 = xmm0[1,1,3,3]
+; SSE41-NEXT:    pand %xmm3, %xmm1
+; SSE41-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,1,3,3]
+; SSE41-NEXT:    por %xmm1, %xmm0
+; SSE41-NEXT:    retq
+;
+; AVX1-LABEL: ult_65_v2i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [65,65]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_65_v2i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX2-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX2-NEXT:    vpaddb %xmm2, %xmm0, %xmm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} xmm1 = [65,65]
+; AVX2-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_65_v2i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [65,65]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX512VPOPCNTDQ-NEXT:    vzeroupper
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_65_v2i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_65_v2i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [65,65]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; BITALG_NOVLX-NEXT:    vzeroupper
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_65_v2i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %xmm1, %xmm0, %xmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
+; BITALG-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
+; BITALG-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
+  %3 = icmp ult <2 x i64> %2, <i64 65, i64 65>
+  %4 = sext <2 x i1> %3 to <2 x i64>
+  ret <2 x i64> %4
+}
+
+declare <16 x i8> @llvm.ctpop.v16i8(<16 x i8>)
+declare <8 x i16> @llvm.ctpop.v8i16(<8 x i16>)
+declare <4 x i32> @llvm.ctpop.v4i32(<4 x i32>)
+declare <2 x i64> @llvm.ctpop.v2i64(<2 x i64>)
diff --git a/llvm/test/CodeGen/X86/vector-popcnt-128.ll b/llvm/test/CodeGen/X86/vector-popcnt-128.ll
index e05b1490ac056..89e789010b254 100644
--- a/llvm/test/CodeGen/X86/vector-popcnt-128.ll
+++ b/llvm/test/CodeGen/X86/vector-popcnt-128.ll
@@ -681,222 +681,6 @@ define <16 x i8> @foldv16i8() nounwind {
   ret <16 x i8> %out
 }
 
-define <2 x i64> @ult_2_v2i64(<2 x i64> %0) {
-; SSE2-LABEL: ult_2_v2i64:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT:    paddq %xmm0, %xmm1
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    pxor %xmm2, %xmm2
-; SSE2-NEXT:    pcmpeqd %xmm1, %xmm2
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
-; SSE2-NEXT:    pand %xmm2, %xmm0
-; SSE2-NEXT:    retq
-;
-; SSE3-LABEL: ult_2_v2i64:
-; SSE3:       # %bb.0:
-; SSE3-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE3-NEXT:    paddq %xmm0, %xmm1
-; SSE3-NEXT:    pand %xmm0, %xmm1
-; SSE3-NEXT:    pxor %xmm2, %xmm2
-; SSE3-NEXT:    pcmpeqd %xmm1, %xmm2
-; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
-; SSE3-NEXT:    pand %xmm2, %xmm0
-; SSE3-NEXT:    retq
-;
-; SSSE3-LABEL: ult_2_v2i64:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSSE3-NEXT:    paddq %xmm0, %xmm1
-; SSSE3-NEXT:    pand %xmm0, %xmm1
-; SSSE3-NEXT:    pxor %xmm2, %xmm2
-; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm2
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm2[1,0,3,2]
-; SSSE3-NEXT:    pand %xmm2, %xmm0
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: ult_2_v2i64:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE41-NEXT:    paddq %xmm0, %xmm1
-; SSE41-NEXT:    pand %xmm1, %xmm0
-; SSE41-NEXT:    pxor %xmm1, %xmm1
-; SSE41-NEXT:    pcmpeqq %xmm1, %xmm0
-; SSE41-NEXT:    retq
-;
-; AVX1-LABEL: ult_2_v2i64:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ult_2_v2i64:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v2i64:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vzeroupper
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ult_2_v2i64:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip), %xmm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v2i64:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
-; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v2i64:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
-; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
-  %3 = icmp ult <2 x i64> %2, <i64 2, i64 2>
-  %4 = sext <2 x i1> %3 to <2 x i64>
-  ret <2 x i64> %4
-}
-
-define <2 x i64> @ugt_1_v2i64(<2 x i64> %0) {
-; SSE2-LABEL: ugt_1_v2i64:
-; SSE2:       # %bb.0:
-; SSE2-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE2-NEXT:    movdqa %xmm0, %xmm2
-; SSE2-NEXT:    paddq %xmm1, %xmm2
-; SSE2-NEXT:    pand %xmm0, %xmm2
-; SSE2-NEXT:    pxor %xmm3, %xmm3
-; SSE2-NEXT:    pcmpeqd %xmm2, %xmm3
-; SSE2-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
-; SSE2-NEXT:    pand %xmm3, %xmm0
-; SSE2-NEXT:    pxor %xmm1, %xmm0
-; SSE2-NEXT:    retq
-;
-; SSE3-LABEL: ugt_1_v2i64:
-; SSE3:       # %bb.0:
-; SSE3-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE3-NEXT:    movdqa %xmm0, %xmm2
-; SSE3-NEXT:    paddq %xmm1, %xmm2
-; SSE3-NEXT:    pand %xmm0, %xmm2
-; SSE3-NEXT:    pxor %xmm3, %xmm3
-; SSE3-NEXT:    pcmpeqd %xmm2, %xmm3
-; SSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
-; SSE3-NEXT:    pand %xmm3, %xmm0
-; SSE3-NEXT:    pxor %xmm1, %xmm0
-; SSE3-NEXT:    retq
-;
-; SSSE3-LABEL: ugt_1_v2i64:
-; SSSE3:       # %bb.0:
-; SSSE3-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSSE3-NEXT:    movdqa %xmm0, %xmm2
-; SSSE3-NEXT:    paddq %xmm1, %xmm2
-; SSSE3-NEXT:    pand %xmm0, %xmm2
-; SSSE3-NEXT:    pxor %xmm3, %xmm3
-; SSSE3-NEXT:    pcmpeqd %xmm2, %xmm3
-; SSSE3-NEXT:    pshufd {{.*#+}} xmm0 = xmm3[1,0,3,2]
-; SSSE3-NEXT:    pand %xmm3, %xmm0
-; SSSE3-NEXT:    pxor %xmm1, %xmm0
-; SSSE3-NEXT:    retq
-;
-; SSE41-LABEL: ugt_1_v2i64:
-; SSE41:       # %bb.0:
-; SSE41-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE41-NEXT:    movdqa %xmm0, %xmm1
-; SSE41-NEXT:    paddq %xmm2, %xmm1
-; SSE41-NEXT:    pand %xmm0, %xmm1
-; SSE41-NEXT:    pxor %xmm0, %xmm0
-; SSE41-NEXT:    pcmpeqq %xmm0, %xmm1
-; SSE41-NEXT:    pxor %xmm2, %xmm1
-; SSE41-NEXT:    movdqa %xmm1, %xmm0
-; SSE41-NEXT:    retq
-;
-; AVX1-LABEL: ugt_1_v2i64:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddq %xmm1, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpeqq %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v2i64:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddq %xmm1, %xmm0, %xmm2
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqq %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v2i64:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq {{.*}}(%rip), %xmm0, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vzeroupper
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v2i64:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip), %xmm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %xmm0, %xmm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v2i64:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
-; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
-; BITALG_NOVLX-NEXT:    vzeroupper
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v2i64:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpaddq %xmm1, %xmm0, %xmm1
-; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqq %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <2 x i64> @llvm.ctpop.v2i64(<2 x i64> %0)
-  %3 = icmp ugt <2 x i64> %2, <i64 1, i64 1>
-  %4 = sext <2 x i1> %3 to <2 x i64>
-  ret <2 x i64> %4
-}
-
 define <2 x i64> @eq_1_v2i64(<2 x i64> %0) {
 ; SSE2-LABEL: eq_1_v2i64:
 ; SSE2:       # %bb.0:
@@ -1153,151 +937,6 @@ define <2 x i64> @ne_1_v2i64(<2 x i64> %0) {
   ret <2 x i64> %4
 }
 
-define <4 x i32> @ult_2_v4i32(<4 x i32> %0) {
-; SSE-LABEL: ult_2_v4i32:
-; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE-NEXT:    paddd %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pcmpeqd %xmm1, %xmm0
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: ult_2_v4i32:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
-; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ult_2_v4i32:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
-; AVX2-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v4i32:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [2,2,2,2]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vzeroupper
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ult_2_v4i32:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to4}, %xmm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v4i32:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
-; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v4i32:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
-; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
-  %3 = icmp ult <4 x i32> %2, <i32 2, i32 2, i32 2, i32 2>
-  %4 = sext <4 x i1> %3 to <4 x i32>
-  ret <4 x i32> %4
-}
-
-define <4 x i32> @ugt_1_v4i32(<4 x i32> %0) {
-; SSE-LABEL: ugt_1_v4i32:
-; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    paddd %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    pcmpeqd %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: ugt_1_v4i32:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v4i32:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddd %xmm1, %xmm0, %xmm2
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqd %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v4i32:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} xmm1 = [1,1,1,1]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vzeroupper
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v4i32:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to4}, %xmm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm0, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %xmm0, %xmm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v4i32:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
-; BITALG_NOVLX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
-; BITALG_NOVLX-NEXT:    vzeroupper
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v4i32:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpaddd %xmm1, %xmm0, %xmm1
-; BITALG-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqd %xmm1, %xmm0, %xmm0
-; BITALG-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <4 x i32> @llvm.ctpop.v4i32(<4 x i32> %0)
-  %3 = icmp ugt <4 x i32> %2, <i32 1, i32 1, i32 1, i32 1>
-  %4 = sext <4 x i1> %3 to <4 x i32>
-  ret <4 x i32> %4
-}
-
 define <4 x i32> @eq_1_v4i32(<4 x i32> %0) {
 ; SSE-LABEL: eq_1_v4i32:
 ; SSE:       # %bb.0:
@@ -1462,121 +1101,6 @@ define <4 x i32> @ne_1_v4i32(<4 x i32> %0) {
   ret <4 x i32> %4
 }
 
-define <8 x i16> @ult_2_v8i16(<8 x i16> %0) {
-; SSE-LABEL: ult_2_v8i16:
-; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE-NEXT:    paddw %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pcmpeqw %xmm1, %xmm0
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: ult_2_v8i16:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpaddw %xmm1, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v8i16:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2]
-; BITALG_NOVLX-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
-; BITALG_NOVLX-NEXT:    vzeroupper
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v8i16:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
-; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %xmm0, %k0
-; BITALG-NEXT:    vpmovm2w %k0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
-  %3 = icmp ult <8 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
-  %4 = sext <8 x i1> %3 to <8 x i16>
-  ret <8 x i16> %4
-}
-
-define <8 x i16> @ugt_1_v8i16(<8 x i16> %0) {
-; SSE-LABEL: ugt_1_v8i16:
-; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    paddw %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    pcmpeqw %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: ugt_1_v8i16:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddw %xmm1, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpeqw %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v8i16:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddw %xmm1, %xmm0, %xmm2
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqw %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v8i16:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpaddw %xmm1, %xmm0, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vzeroupper
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v8i16:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpaddw %xmm1, %xmm0, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v8i16:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vzeroupper
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v8i16:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntw %xmm0, %xmm0
-; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %xmm0, %k0
-; BITALG-NEXT:    vpmovm2w %k0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <8 x i16> @llvm.ctpop.v8i16(<8 x i16> %0)
-  %3 = icmp ugt <8 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %4 = sext <8 x i1> %3 to <8 x i16>
-  ret <8 x i16> %4
-}
-
 define <8 x i16> @eq_1_v8i16(<8 x i16> %0) {
 ; SSE-LABEL: eq_1_v8i16:
 ; SSE:       # %bb.0:
@@ -1704,121 +1228,6 @@ define <8 x i16> @ne_1_v8i16(<8 x i16> %0) {
   ret <8 x i16> %4
 }
 
-define <16 x i8> @ult_2_v16i8(<16 x i8> %0) {
-; SSE-LABEL: ult_2_v16i8:
-; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqd %xmm1, %xmm1
-; SSE-NEXT:    paddb %xmm0, %xmm1
-; SSE-NEXT:    pand %xmm1, %xmm0
-; SSE-NEXT:    pxor %xmm1, %xmm1
-; SSE-NEXT:    pcmpeqb %xmm1, %xmm0
-; SSE-NEXT:    retq
-;
-; AVX-LABEL: ult_2_v16i8:
-; AVX:       # %bb.0:
-; AVX-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
-; AVX-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
-; AVX-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v16i8:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
-; BITALG_NOVLX-NEXT:    vpcmpgtb %xmm0, %xmm1, %xmm0
-; BITALG_NOVLX-NEXT:    vzeroupper
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v16i8:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
-; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %xmm0, %k0
-; BITALG-NEXT:    vpmovm2b %k0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
-  %3 = icmp ult <16 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
-  %4 = sext <16 x i1> %3 to <16 x i8>
-  ret <16 x i8> %4
-}
-
-define <16 x i8> @ugt_1_v16i8(<16 x i8> %0) {
-; SSE-LABEL: ugt_1_v16i8:
-; SSE:       # %bb.0:
-; SSE-NEXT:    pcmpeqd %xmm2, %xmm2
-; SSE-NEXT:    movdqa %xmm0, %xmm1
-; SSE-NEXT:    paddb %xmm2, %xmm1
-; SSE-NEXT:    pand %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm0, %xmm0
-; SSE-NEXT:    pcmpeqb %xmm0, %xmm1
-; SSE-NEXT:    pxor %xmm2, %xmm1
-; SSE-NEXT:    movdqa %xmm1, %xmm0
-; SSE-NEXT:    retq
-;
-; AVX1-LABEL: ugt_1_v16i8:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddb %xmm1, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v16i8:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpaddb %xmm1, %xmm0, %xmm2
-; AVX2-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqb %xmm2, %xmm0, %xmm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm0, %xmm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v16i8:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $xmm0 killed $xmm0 killed $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vzeroupper
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v16i8:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpaddb %xmm1, %xmm0, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpand %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %xmm0, %xmm0, %xmm0
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v16i8:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $xmm0 killed $xmm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %xmm0, %xmm0
-; BITALG_NOVLX-NEXT:    vzeroupper
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v16i8:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntb %xmm0, %xmm0
-; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %xmm0, %k0
-; BITALG-NEXT:    vpmovm2b %k0, %xmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <16 x i8> @llvm.ctpop.v16i8(<16 x i8> %0)
-  %3 = icmp ugt <16 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
-  %4 = sext <16 x i1> %3 to <16 x i8>
-  ret <16 x i8> %4
-}
-
 define <16 x i8> @eq_1_v16i8(<16 x i8> %0) {
 ; SSE-LABEL: eq_1_v16i8:
 ; SSE:       # %bb.0:
diff --git a/llvm/test/CodeGen/X86/vector-popcnt-256-ult-ugt.ll b/llvm/test/CodeGen/X86/vector-popcnt-256-ult-ugt.ll
new file mode 100644
index 0000000000000..809c1c52f3ede
--- /dev/null
+++ b/llvm/test/CodeGen/X86/vector-popcnt-256-ult-ugt.ll
@@ -0,0 +1,22064 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX1
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx2 | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX2
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vpopcntdq | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX512VPOPCNTDQ
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vpopcntdq,+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=AVX512VPOPCNTDQVL
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bitalg | FileCheck %s --check-prefix=ALL --check-prefix=AVX --check-prefix=BITALG_NOVLX
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bitalg,+avx512vl | FileCheck %s --check-prefix=ALL --check-prefix=BITALG
+
+
+define <32 x i8> @ult_0_v32i8(<32 x i8> %0) {
+; ALL-LABEL: ult_0_v32i8:
+; ALL:       # %bb.0:
+; ALL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; ALL-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_0_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_0_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpxor %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_1_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_1_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_1_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_1_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddb %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddb %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_2_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_2_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddb %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_2_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_2_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_2_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_2_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_3_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_3_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_3_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_3_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_4_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_4_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_4_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_4_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_5_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_5_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_5_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_5_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_6_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_6_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_6_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_6_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_7_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_7_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_7_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_7_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_8_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_8_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ugt_8_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ugt_8_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpmaxub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmaxub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ugt <32 x i8> %2, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <32 x i8> @ult_9_v32i8(<32 x i8> %0) {
+; AVX1-LABEL: ult_9_v32i8:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpminub %xmm1, %xmm0, %xmm3
+; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpminub %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpcmpeqb %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v32i8:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v32i8:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v32i8:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQVL-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpminub {{.*}}(%rip), %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v32i8:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v32i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
+  %3 = icmp ult <32 x i8> %2, <i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9>
+  %4 = sext <32 x i1> %3 to <32 x i8>
+  ret <32 x i8> %4
+}
+
+define <16 x i16> @ult_0_v16i16(<16 x i16> %0) {
+; ALL-LABEL: ult_0_v16i16:
+; ALL:       # %bb.0:
+; ALL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; ALL-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_0_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_0_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpxor %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_1_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_1_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_1_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_1_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddw %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddw %xmm2, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddw %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqw %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_2_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_2_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddw %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddw %xmm2, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_2_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_2_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
+; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_2_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_2_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2,2,2,2,2]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_3_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_3_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_3_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_3_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3,3,3,3,3]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_4_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_4_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_4_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_4_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_5_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_5_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_5_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_5_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5,5,5,5,5]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_6_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_6_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_6_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_6_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6,6,6,6,6]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_7_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_7_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_7_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_7_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7,7,7,7,7]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_8_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_8_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_8_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_8_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8,8,8,8,8]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_9_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_9_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_9_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_9_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9,9,9,9,9]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_9_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_9_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_9_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_10_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_10_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_10_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_10_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_10_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_10_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_10_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10,10,10,10,10]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_10_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_10_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_10_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_11_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_11_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_11_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_11_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_11_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_11_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_11_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11,11,11,11,11]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_11_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_11_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_11_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_12_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_12_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_12_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_12_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_12_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_12_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_12_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12,12,12,12,12]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_12_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_12_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_12_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_13_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_13_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_13_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_13_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_13_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_13_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_13_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13,13,13,13,13]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_13_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_13_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_13_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_14_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_14_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_14_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_14_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_14_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_14_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_14_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14,14,14,14,14]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_14_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_14_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_14_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_15_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_15_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_15_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_15_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_15_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_15_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_15_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_15_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_15_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_15_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_16_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_16_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_16_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_16_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_16_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ugt_16_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ugt_16_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16,16,16,16,16]
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_16_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_16_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_16_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ugt <16 x i16> %2, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <16 x i16> @ult_17_v16i16(<16 x i16> %0) {
+; AVX1-LABEL: ult_17_v16i16:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsllw $8, %xmm2, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpsrlw $8, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpsllw $8, %xmm0, %xmm1
+; AVX1-NEXT:    vpaddb %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpsrlw $8, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17,17,17,17,17]
+; AVX1-NEXT:    vpcmpgtw %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtw %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_17_v16i16:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX2-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17]
+; AVX2-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v16i16:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vmovdqa {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_17_v16i16:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQVL-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_17_v16i16:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v16i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
+  %3 = icmp ult <16 x i16> %2, <i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17>
+  %4 = sext <16 x i1> %3 to <16 x i16>
+  ret <16 x i16> %4
+}
+
+define <8 x i32> @ult_0_v8i32(<8 x i32> %0) {
+; ALL-LABEL: ult_0_v8i32:
+; ALL:       # %bb.0:
+; ALL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; ALL-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_0_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_0_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpxor %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_1_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_1_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm4, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_1_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_1_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddd %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddd %xmm2, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG_NOVLX-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
+; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
+; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_2_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_2_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddd %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddd %xmm2, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_2_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_2_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG_NOVLX-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
+; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
+; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_2_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_2_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2,2,2]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_3_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_3_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_3_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_3_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3,3,3]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_4_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_4_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_4_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_4_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4,4,4]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_5_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_5_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_5_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_5_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5,5,5]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_6_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_6_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_6_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_6_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6,6,6]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_7_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_7_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_7_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_7_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7,7,7]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_8_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_8_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_8_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_8_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8,8,8]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_9_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_9_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_9_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_9_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9,9,9]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_9_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_9_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_9_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_10_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_10_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_10_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_10_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_10_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_10_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_10_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10,10,10]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_10_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_10_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_10_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_11_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_11_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_11_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_11_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_11_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_11_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_11_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11,11,11]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_11_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_11_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_11_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_12_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_12_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_12_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_12_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_12_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_12_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_12_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12,12,12]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_12_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_12_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_12_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_13_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_13_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_13_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_13_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_13_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_13_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_13_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13,13,13]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_13_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_13_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_13_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_14_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_14_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_14_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_14_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_14_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_14_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_14_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14,14,14]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_14_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_14_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_14_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_15_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_15_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_15_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_15_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_15_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_15_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_15_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_15_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_15_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_15_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_16_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_16_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_16_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_16_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_16_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_16_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_16_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16,16,16]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_16_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_16_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_16_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_17_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_17_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_17_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_17_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_17_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_17_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_17_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17,17,17]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_17_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_17_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_17_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_17_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_17_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_18_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_18_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18,18,18]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_18_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [18,18,18,18,18,18,18,18]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_18_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [18,18,18,18,18,18,18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_18_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_18_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [18,18,18,18,18,18,18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_18_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_18_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_18_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18,18,18]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_18_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [18,18,18,18,18,18,18,18]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_18_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [18,18,18,18,18,18,18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_18_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_18_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [18,18,18,18,18,18,18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_18_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_19_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_19_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19,19,19]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_19_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [19,19,19,19,19,19,19,19]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_19_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [19,19,19,19,19,19,19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_19_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_19_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [19,19,19,19,19,19,19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_19_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_19_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_19_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19,19,19]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_19_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [19,19,19,19,19,19,19,19]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_19_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [19,19,19,19,19,19,19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_19_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_19_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [19,19,19,19,19,19,19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_19_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_20_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_20_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20,20,20]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_20_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [20,20,20,20,20,20,20,20]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_20_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [20,20,20,20,20,20,20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_20_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_20_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [20,20,20,20,20,20,20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_20_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_20_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_20_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20,20,20]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_20_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [20,20,20,20,20,20,20,20]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_20_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [20,20,20,20,20,20,20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_20_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_20_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [20,20,20,20,20,20,20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_20_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_21_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_21_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21,21,21]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_21_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [21,21,21,21,21,21,21,21]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_21_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [21,21,21,21,21,21,21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_21_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_21_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [21,21,21,21,21,21,21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_21_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_21_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_21_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21,21,21]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_21_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [21,21,21,21,21,21,21,21]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_21_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [21,21,21,21,21,21,21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_21_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_21_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [21,21,21,21,21,21,21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_21_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_22_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_22_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22,22,22]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_22_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [22,22,22,22,22,22,22,22]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_22_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [22,22,22,22,22,22,22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_22_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_22_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [22,22,22,22,22,22,22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_22_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_22_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_22_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22,22,22]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_22_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [22,22,22,22,22,22,22,22]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_22_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [22,22,22,22,22,22,22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_22_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_22_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [22,22,22,22,22,22,22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_22_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_23_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_23_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23,23,23]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_23_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [23,23,23,23,23,23,23,23]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_23_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [23,23,23,23,23,23,23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_23_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_23_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [23,23,23,23,23,23,23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_23_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_23_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_23_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23,23,23]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_23_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [23,23,23,23,23,23,23,23]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_23_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [23,23,23,23,23,23,23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_23_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_23_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [23,23,23,23,23,23,23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_23_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_24_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_24_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24,24,24]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_24_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [24,24,24,24,24,24,24,24]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_24_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [24,24,24,24,24,24,24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_24_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_24_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [24,24,24,24,24,24,24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_24_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_24_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_24_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24,24,24]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_24_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [24,24,24,24,24,24,24,24]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_24_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [24,24,24,24,24,24,24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_24_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_24_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [24,24,24,24,24,24,24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_24_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_25_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_25_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25,25,25]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_25_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [25,25,25,25,25,25,25,25]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_25_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [25,25,25,25,25,25,25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_25_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_25_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [25,25,25,25,25,25,25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_25_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_25_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_25_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25,25,25]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_25_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [25,25,25,25,25,25,25,25]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_25_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [25,25,25,25,25,25,25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_25_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_25_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [25,25,25,25,25,25,25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_25_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_26_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_26_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26,26,26]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_26_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [26,26,26,26,26,26,26,26]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_26_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [26,26,26,26,26,26,26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_26_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_26_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [26,26,26,26,26,26,26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_26_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_26_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_26_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26,26,26]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_26_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [26,26,26,26,26,26,26,26]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_26_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [26,26,26,26,26,26,26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_26_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_26_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [26,26,26,26,26,26,26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_26_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_27_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_27_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27,27,27]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_27_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [27,27,27,27,27,27,27,27]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_27_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [27,27,27,27,27,27,27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_27_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_27_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [27,27,27,27,27,27,27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_27_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_27_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_27_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27,27,27]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_27_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [27,27,27,27,27,27,27,27]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_27_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [27,27,27,27,27,27,27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_27_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_27_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [27,27,27,27,27,27,27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_27_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_28_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_28_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28,28,28]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_28_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_28_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_28_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_28_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_28_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_28_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_28_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28,28,28]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_28_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_28_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_28_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_28_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [28,28,28,28,28,28,28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_28_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_29_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_29_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29,29,29]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_29_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [29,29,29,29,29,29,29,29]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_29_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [29,29,29,29,29,29,29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_29_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_29_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [29,29,29,29,29,29,29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_29_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_29_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_29_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29,29,29]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_29_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [29,29,29,29,29,29,29,29]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_29_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [29,29,29,29,29,29,29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_29_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_29_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [29,29,29,29,29,29,29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_29_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_30_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_30_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30,30,30]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_30_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [30,30,30,30,30,30,30,30]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_30_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [30,30,30,30,30,30,30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_30_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_30_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [30,30,30,30,30,30,30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_30_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_30_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_30_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30,30,30]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_30_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [30,30,30,30,30,30,30,30]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_30_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [30,30,30,30,30,30,30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_30_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_30_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [30,30,30,30,30,30,30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_30_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_31_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_31_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31,31,31]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_31_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [31,31,31,31,31,31,31,31]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_31_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [31,31,31,31,31,31,31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_31_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_31_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [31,31,31,31,31,31,31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_31_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_31_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_31_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31,31,31]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_31_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [31,31,31,31,31,31,31,31]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_31_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [31,31,31,31,31,31,31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_31_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_31_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [31,31,31,31,31,31,31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_31_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_32_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_32_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32,32,32]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_32_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_32_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_32_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_32_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_32_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ugt_32_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ugt_32_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32,32,32]
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_32_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32]
+; AVX2-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_32_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_32_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_32_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [32,32,32,32,32,32,32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_32_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ugt <8 x i32> %2, <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <8 x i32> @ult_33_v8i32(<8 x i32> %0) {
+; AVX1-LABEL: ult_33_v8i32:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm5 = xmm2[2],xmm4[2],xmm2[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm5, %xmm5
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm2 = xmm2[0],zero,xmm2[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpackuswb %xmm5, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpunpckhdq {{.*#+}} xmm1 = xmm0[2],xmm4[2],xmm0[3],xmm4[3]
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm1, %xmm1
+; AVX1-NEXT:    vpmovzxdq {{.*#+}} xmm0 = xmm0[0],zero,xmm0[1],zero
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpackuswb %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33,33,33]
+; AVX1-NEXT:    vpcmpgtd %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtd %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_33_v8i32:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; AVX2-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [33,33,33,33,33,33,33,33]
+; AVX2-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_33_v8i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [33,33,33,33,33,33,33,33]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_33_v8i32:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_33_v8i32:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG_NOVLX-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [33,33,33,33,33,33,33,33]
+; BITALG_NOVLX-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_33_v8i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm1[2],ymm0[3],ymm1[3],ymm0[6],ymm1[6],ymm0[7],ymm1[7]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm2, %ymm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm1[0],ymm0[1],ymm1[1],ymm0[4],ymm1[4],ymm0[5],ymm1[5]
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
+  %3 = icmp ult <8 x i32> %2, <i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33>
+  %4 = sext <8 x i1> %3 to <8 x i32>
+  ret <8 x i32> %4
+}
+
+define <4 x i64> @ult_0_v4i64(<4 x i64> %0) {
+; ALL-LABEL: ult_0_v4i64:
+; ALL:       # %bb.0:
+; ALL-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; ALL-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 0, i64 0, i64 0, i64 0>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_0_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_0_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqd %xmm1, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vpxor %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_0_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_0_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_0_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 0, i64 0, i64 0, i64 0>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_1_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_1_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm4, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_1_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_1_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_1_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 1, i64 1, i64 1, i64 1>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_1_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_1_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddq %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_1_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddq %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX2-NEXT:    vpcmpeqq %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_1_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_1_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG_NOVLX-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
+; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
+; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 1, i64 1, i64 1, i64 1>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_2_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_2_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
+; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
+; AVX1-NEXT:    vpaddq %xmm2, %xmm1, %xmm3
+; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm1, %xmm1
+; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm2
+; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm0, %xmm0
+; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_2_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; AVX2-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2,2,2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_2_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_2_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG_NOVLX-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
+; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
+; BITALG-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
+; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 2, i64 2, i64 2, i64 2>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_2_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_2_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [2,2]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_2_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2,2,2,2]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2,2,2,2]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_2_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_2_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2,2,2,2]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 2, i64 2, i64 2, i64 2>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_3_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_3_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_3_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [3,3,3,3]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_3_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_3_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 3, i64 3, i64 3, i64 3>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_3_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_3_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [3,3]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_3_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [3,3,3,3]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [3,3,3,3]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_3_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_3_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [3,3,3,3]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 3, i64 3, i64 3, i64 3>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_4_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_4_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_4_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4,4,4,4]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_4_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_4_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 4, i64 4, i64 4, i64 4>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_4_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_4_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [4,4]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_4_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4,4,4,4]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4,4,4,4]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_4_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_4_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [4,4,4,4]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 4, i64 4, i64 4, i64 4>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_5_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_5_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_5_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [5,5,5,5]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_5_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_5_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 5, i64 5, i64 5, i64 5>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_5_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_5_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [5,5]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_5_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [5,5,5,5]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [5,5,5,5]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_5_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_5_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [5,5,5,5]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 5, i64 5, i64 5, i64 5>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_6_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_6_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_6_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_6_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_6_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 6, i64 6, i64 6, i64 6>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_6_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_6_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [6,6]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_6_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_6_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_6_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [6,6,6,6]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 6, i64 6, i64 6, i64 6>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_7_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_7_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_7_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7,7,7,7]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_7_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_7_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 7, i64 7, i64 7, i64 7>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_7_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_7_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [7,7]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_7_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7,7,7,7]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7,7,7,7]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_7_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_7_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [7,7,7,7]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 7, i64 7, i64 7, i64 7>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_8_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_8_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_8_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8,8,8,8]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_8_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_8_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 8, i64 8, i64 8, i64 8>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_8_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_8_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [8,8]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_8_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8,8,8,8]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8,8,8,8]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_8_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_8_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [8,8,8,8]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 8, i64 8, i64 8, i64 8>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_9_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_9_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_9_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9,9,9,9]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_9_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_9_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 9, i64 9, i64 9, i64 9>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_9_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_9_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [9,9]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_9_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9,9,9,9]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9,9,9,9]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_9_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_9_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [9,9,9,9]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 9, i64 9, i64 9, i64 9>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_10_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_10_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_10_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [10,10,10,10]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_10_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_10_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 10, i64 10, i64 10, i64 10>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_10_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_10_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [10,10]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_10_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [10,10,10,10]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [10,10,10,10]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_10_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_10_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [10,10,10,10]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 10, i64 10, i64 10, i64 10>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_11_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_11_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_11_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [11,11,11,11]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_11_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_11_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 11, i64 11, i64 11, i64 11>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_11_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_11_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [11,11]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_11_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [11,11,11,11]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [11,11,11,11]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_11_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_11_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [11,11,11,11]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 11, i64 11, i64 11, i64 11>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_12_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_12_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_12_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [12,12,12,12]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_12_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_12_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 12, i64 12, i64 12, i64 12>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_12_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_12_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [12,12]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_12_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [12,12,12,12]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [12,12,12,12]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_12_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_12_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [12,12,12,12]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 12, i64 12, i64 12, i64 12>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_13_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_13_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_13_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [13,13,13,13]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_13_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_13_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 13, i64 13, i64 13, i64 13>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_13_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_13_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [13,13]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_13_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [13,13,13,13]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [13,13,13,13]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_13_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_13_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [13,13,13,13]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 13, i64 13, i64 13, i64 13>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_14_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_14_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_14_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [14,14,14,14]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_14_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_14_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 14, i64 14, i64 14, i64 14>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_14_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_14_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [14,14]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_14_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [14,14,14,14]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [14,14,14,14]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_14_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_14_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [14,14,14,14]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 14, i64 14, i64 14, i64 14>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_15_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_15_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_15_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [15,15,15,15]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_15_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_15_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 15, i64 15, i64 15, i64 15>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_15_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_15_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_15_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [15,15,15,15]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [15,15,15,15]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_15_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_15_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [15,15,15,15]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 15, i64 15, i64 15, i64 15>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_16_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_16_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_16_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [16,16,16,16]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_16_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_16_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 16, i64 16, i64 16, i64 16>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_16_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_16_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [16,16]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_16_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [16,16,16,16]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [16,16,16,16]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_16_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_16_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [16,16,16,16]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 16, i64 16, i64 16, i64 16>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_17_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_17_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_17_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [17,17,17,17]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_17_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_17_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 17, i64 17, i64 17, i64 17>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_17_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_17_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [17,17]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_17_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [17,17,17,17]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_17_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [17,17,17,17]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_17_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_17_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [17,17,17,17]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_17_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 17, i64 17, i64 17, i64 17>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_18_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_18_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_18_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18,18,18,18]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_18_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18,18,18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_18_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_18_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18,18,18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_18_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 18, i64 18, i64 18, i64 18>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_18_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_18_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [18,18]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_18_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18,18,18,18]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_18_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18,18,18,18]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_18_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_18_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [18,18,18,18]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_18_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 18, i64 18, i64 18, i64 18>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_19_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_19_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_19_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [19,19,19,19]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_19_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [19,19,19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_19_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_19_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [19,19,19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_19_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 19, i64 19, i64 19, i64 19>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_19_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_19_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [19,19]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_19_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [19,19,19,19]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_19_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [19,19,19,19]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_19_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_19_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [19,19,19,19]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_19_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 19, i64 19, i64 19, i64 19>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_20_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_20_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_20_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [20,20,20,20]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_20_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [20,20,20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_20_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_20_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [20,20,20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_20_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 20, i64 20, i64 20, i64 20>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_20_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_20_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [20,20]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_20_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [20,20,20,20]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_20_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [20,20,20,20]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_20_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_20_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [20,20,20,20]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_20_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 20, i64 20, i64 20, i64 20>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_21_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_21_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_21_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [21,21,21,21]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_21_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [21,21,21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_21_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_21_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [21,21,21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_21_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 21, i64 21, i64 21, i64 21>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_21_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_21_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [21,21]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_21_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [21,21,21,21]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_21_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [21,21,21,21]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_21_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_21_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [21,21,21,21]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_21_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 21, i64 21, i64 21, i64 21>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_22_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_22_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_22_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [22,22,22,22]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_22_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [22,22,22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_22_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_22_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [22,22,22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_22_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 22, i64 22, i64 22, i64 22>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_22_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_22_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [22,22]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_22_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [22,22,22,22]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_22_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [22,22,22,22]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_22_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_22_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [22,22,22,22]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_22_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 22, i64 22, i64 22, i64 22>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_23_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_23_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_23_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [23,23,23,23]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_23_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [23,23,23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_23_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_23_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [23,23,23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_23_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 23, i64 23, i64 23, i64 23>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_23_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_23_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [23,23]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_23_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [23,23,23,23]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_23_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [23,23,23,23]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_23_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_23_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [23,23,23,23]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_23_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 23, i64 23, i64 23, i64 23>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_24_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_24_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_24_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [24,24,24,24]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_24_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [24,24,24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_24_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_24_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [24,24,24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_24_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 24, i64 24, i64 24, i64 24>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_24_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_24_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [24,24]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_24_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [24,24,24,24]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_24_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [24,24,24,24]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_24_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_24_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [24,24,24,24]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_24_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 24, i64 24, i64 24, i64 24>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_25_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_25_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_25_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25,25,25,25]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_25_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25,25,25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_25_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_25_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25,25,25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_25_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 25, i64 25, i64 25, i64 25>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_25_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_25_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [25,25]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_25_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25,25,25,25]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_25_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25,25,25,25]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_25_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_25_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [25,25,25,25]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_25_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 25, i64 25, i64 25, i64 25>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_26_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_26_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_26_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [26,26,26,26]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_26_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [26,26,26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_26_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_26_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [26,26,26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_26_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 26, i64 26, i64 26, i64 26>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_26_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_26_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [26,26]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_26_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [26,26,26,26]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_26_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [26,26,26,26]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_26_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_26_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [26,26,26,26]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_26_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 26, i64 26, i64 26, i64 26>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_27_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_27_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_27_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [27,27,27,27]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_27_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [27,27,27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_27_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_27_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [27,27,27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_27_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 27, i64 27, i64 27, i64 27>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_27_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_27_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [27,27]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_27_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [27,27,27,27]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_27_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [27,27,27,27]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_27_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_27_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [27,27,27,27]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_27_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 27, i64 27, i64 27, i64 27>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_28_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_28_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_28_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [28,28,28,28]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_28_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [28,28,28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_28_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_28_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [28,28,28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_28_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 28, i64 28, i64 28, i64 28>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_28_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_28_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [28,28]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_28_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [28,28,28,28]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_28_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [28,28,28,28]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_28_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_28_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [28,28,28,28]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_28_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 28, i64 28, i64 28, i64 28>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_29_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_29_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_29_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [29,29,29,29]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_29_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [29,29,29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_29_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_29_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [29,29,29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_29_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 29, i64 29, i64 29, i64 29>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_29_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_29_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [29,29]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_29_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [29,29,29,29]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_29_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [29,29,29,29]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_29_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_29_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [29,29,29,29]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_29_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 29, i64 29, i64 29, i64 29>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_30_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_30_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_30_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [30,30,30,30]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_30_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [30,30,30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_30_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_30_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [30,30,30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_30_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 30, i64 30, i64 30, i64 30>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_30_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_30_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [30,30]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_30_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [30,30,30,30]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_30_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [30,30,30,30]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_30_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_30_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [30,30,30,30]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_30_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 30, i64 30, i64 30, i64 30>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_31_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_31_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_31_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [31,31,31,31]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_31_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [31,31,31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_31_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_31_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [31,31,31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_31_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 31, i64 31, i64 31, i64 31>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_31_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_31_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [31,31]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_31_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [31,31,31,31]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_31_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [31,31,31,31]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_31_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_31_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [31,31,31,31]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_31_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 31, i64 31, i64 31, i64 31>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_32_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_32_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_32_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [32,32,32,32]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_32_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [32,32,32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_32_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_32_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [32,32,32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_32_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 32, i64 32, i64 32, i64 32>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_32_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_32_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [32,32]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_32_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [32,32,32,32]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_32_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [32,32,32,32]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_32_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_32_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [32,32,32,32]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_32_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 32, i64 32, i64 32, i64 32>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_33_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_33_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_33_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [33,33,33,33]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_33_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [33,33,33,33]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_33_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_33_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [33,33,33,33]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_33_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 33, i64 33, i64 33, i64 33>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_33_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_33_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [33,33]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_33_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [33,33,33,33]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_33_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [33,33,33,33]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_33_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_33_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [33,33,33,33]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_33_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 33, i64 33, i64 33, i64 33>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_34_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_34_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [34,34]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_34_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [34,34,34,34]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_34_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [34,34,34,34]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_34_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_34_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [34,34,34,34]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_34_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 34, i64 34, i64 34, i64 34>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_34_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_34_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [34,34]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_34_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [34,34,34,34]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_34_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [34,34,34,34]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_34_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_34_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [34,34,34,34]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_34_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 34, i64 34, i64 34, i64 34>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_35_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_35_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [35,35]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_35_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [35,35,35,35]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_35_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [35,35,35,35]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_35_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_35_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [35,35,35,35]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_35_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 35, i64 35, i64 35, i64 35>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_35_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_35_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [35,35]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_35_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [35,35,35,35]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_35_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [35,35,35,35]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_35_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_35_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [35,35,35,35]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_35_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 35, i64 35, i64 35, i64 35>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_36_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_36_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [36,36]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_36_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [36,36,36,36]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_36_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [36,36,36,36]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_36_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_36_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [36,36,36,36]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_36_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 36, i64 36, i64 36, i64 36>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_36_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_36_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [36,36]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_36_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [36,36,36,36]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_36_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [36,36,36,36]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_36_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_36_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [36,36,36,36]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_36_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 36, i64 36, i64 36, i64 36>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_37_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_37_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [37,37]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_37_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [37,37,37,37]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_37_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [37,37,37,37]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_37_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_37_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [37,37,37,37]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_37_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 37, i64 37, i64 37, i64 37>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_37_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_37_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [37,37]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_37_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [37,37,37,37]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_37_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [37,37,37,37]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_37_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_37_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [37,37,37,37]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_37_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 37, i64 37, i64 37, i64 37>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_38_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_38_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [38,38]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_38_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [38,38,38,38]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_38_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [38,38,38,38]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_38_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_38_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [38,38,38,38]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_38_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 38, i64 38, i64 38, i64 38>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_38_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_38_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [38,38]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_38_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [38,38,38,38]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_38_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [38,38,38,38]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_38_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_38_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [38,38,38,38]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_38_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 38, i64 38, i64 38, i64 38>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_39_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_39_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [39,39]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_39_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [39,39,39,39]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_39_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [39,39,39,39]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_39_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_39_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [39,39,39,39]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_39_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 39, i64 39, i64 39, i64 39>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_39_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_39_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [39,39]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_39_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [39,39,39,39]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_39_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [39,39,39,39]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_39_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_39_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [39,39,39,39]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_39_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 39, i64 39, i64 39, i64 39>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_40_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_40_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [40,40]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_40_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [40,40,40,40]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_40_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [40,40,40,40]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_40_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_40_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [40,40,40,40]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_40_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 40, i64 40, i64 40, i64 40>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_40_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_40_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [40,40]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_40_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [40,40,40,40]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_40_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [40,40,40,40]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_40_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_40_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [40,40,40,40]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_40_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 40, i64 40, i64 40, i64 40>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_41_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_41_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [41,41]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_41_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [41,41,41,41]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_41_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [41,41,41,41]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_41_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_41_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [41,41,41,41]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_41_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 41, i64 41, i64 41, i64 41>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_41_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_41_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [41,41]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_41_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [41,41,41,41]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_41_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [41,41,41,41]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_41_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_41_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [41,41,41,41]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_41_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 41, i64 41, i64 41, i64 41>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_42_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_42_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [42,42]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_42_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [42,42,42,42]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_42_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [42,42,42,42]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_42_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_42_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [42,42,42,42]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_42_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 42, i64 42, i64 42, i64 42>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_42_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_42_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [42,42]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_42_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [42,42,42,42]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_42_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [42,42,42,42]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_42_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_42_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [42,42,42,42]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_42_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 42, i64 42, i64 42, i64 42>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_43_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_43_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [43,43]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_43_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [43,43,43,43]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_43_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [43,43,43,43]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_43_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_43_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [43,43,43,43]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_43_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 43, i64 43, i64 43, i64 43>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_43_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_43_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [43,43]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_43_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [43,43,43,43]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_43_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [43,43,43,43]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_43_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_43_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [43,43,43,43]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_43_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 43, i64 43, i64 43, i64 43>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_44_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_44_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [44,44]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_44_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [44,44,44,44]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_44_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [44,44,44,44]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_44_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_44_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [44,44,44,44]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_44_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 44, i64 44, i64 44, i64 44>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_44_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_44_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [44,44]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_44_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [44,44,44,44]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_44_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [44,44,44,44]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_44_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_44_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [44,44,44,44]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_44_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 44, i64 44, i64 44, i64 44>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_45_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_45_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [45,45]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_45_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [45,45,45,45]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_45_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [45,45,45,45]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_45_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_45_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [45,45,45,45]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_45_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 45, i64 45, i64 45, i64 45>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_45_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_45_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [45,45]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_45_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [45,45,45,45]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_45_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [45,45,45,45]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_45_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_45_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [45,45,45,45]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_45_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 45, i64 45, i64 45, i64 45>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_46_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_46_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [46,46]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_46_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [46,46,46,46]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_46_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [46,46,46,46]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_46_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_46_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [46,46,46,46]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_46_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 46, i64 46, i64 46, i64 46>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_46_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_46_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [46,46]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_46_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [46,46,46,46]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_46_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [46,46,46,46]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_46_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_46_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [46,46,46,46]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_46_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 46, i64 46, i64 46, i64 46>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_47_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_47_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [47,47]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_47_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [47,47,47,47]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_47_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [47,47,47,47]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_47_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_47_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [47,47,47,47]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_47_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 47, i64 47, i64 47, i64 47>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_47_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_47_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [47,47]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_47_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [47,47,47,47]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_47_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [47,47,47,47]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_47_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_47_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [47,47,47,47]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_47_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 47, i64 47, i64 47, i64 47>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_48_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_48_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [48,48]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_48_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [48,48,48,48]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_48_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [48,48,48,48]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_48_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_48_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [48,48,48,48]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_48_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 48, i64 48, i64 48, i64 48>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_48_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_48_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [48,48]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_48_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [48,48,48,48]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_48_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [48,48,48,48]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_48_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_48_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [48,48,48,48]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_48_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 48, i64 48, i64 48, i64 48>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_49_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_49_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [49,49]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_49_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [49,49,49,49]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_49_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [49,49,49,49]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_49_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_49_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [49,49,49,49]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_49_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 49, i64 49, i64 49, i64 49>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_49_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_49_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [49,49]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_49_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [49,49,49,49]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_49_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [49,49,49,49]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_49_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_49_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [49,49,49,49]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_49_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 49, i64 49, i64 49, i64 49>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_50_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_50_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [50,50]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_50_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [50,50,50,50]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_50_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [50,50,50,50]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_50_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_50_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [50,50,50,50]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_50_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 50, i64 50, i64 50, i64 50>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_50_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_50_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [50,50]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_50_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [50,50,50,50]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_50_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [50,50,50,50]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_50_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_50_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [50,50,50,50]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_50_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 50, i64 50, i64 50, i64 50>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_51_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_51_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [51,51]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_51_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [51,51,51,51]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_51_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [51,51,51,51]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_51_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_51_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [51,51,51,51]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_51_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 51, i64 51, i64 51, i64 51>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_51_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_51_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [51,51]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_51_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [51,51,51,51]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_51_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [51,51,51,51]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_51_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_51_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [51,51,51,51]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_51_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 51, i64 51, i64 51, i64 51>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_52_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_52_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [52,52]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_52_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [52,52,52,52]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_52_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [52,52,52,52]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_52_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_52_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [52,52,52,52]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_52_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 52, i64 52, i64 52, i64 52>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_52_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_52_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [52,52]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_52_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [52,52,52,52]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_52_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [52,52,52,52]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_52_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_52_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [52,52,52,52]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_52_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 52, i64 52, i64 52, i64 52>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_53_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_53_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [53,53]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_53_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [53,53,53,53]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_53_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [53,53,53,53]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_53_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_53_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [53,53,53,53]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_53_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 53, i64 53, i64 53, i64 53>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_53_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_53_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [53,53]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_53_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [53,53,53,53]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_53_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [53,53,53,53]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_53_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_53_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [53,53,53,53]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_53_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 53, i64 53, i64 53, i64 53>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_54_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_54_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [54,54]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_54_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [54,54,54,54]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_54_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [54,54,54,54]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_54_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_54_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [54,54,54,54]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_54_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 54, i64 54, i64 54, i64 54>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_54_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_54_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [54,54]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_54_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [54,54,54,54]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_54_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [54,54,54,54]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_54_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_54_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [54,54,54,54]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_54_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 54, i64 54, i64 54, i64 54>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_55_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_55_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [55,55]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_55_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [55,55,55,55]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_55_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [55,55,55,55]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_55_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_55_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [55,55,55,55]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_55_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 55, i64 55, i64 55, i64 55>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_55_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_55_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [55,55]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_55_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [55,55,55,55]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_55_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [55,55,55,55]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_55_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_55_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [55,55,55,55]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_55_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 55, i64 55, i64 55, i64 55>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_56_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_56_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [56,56]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_56_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [56,56,56,56]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_56_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [56,56,56,56]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_56_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_56_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [56,56,56,56]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_56_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 56, i64 56, i64 56, i64 56>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_56_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_56_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [56,56]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_56_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [56,56,56,56]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_56_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [56,56,56,56]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_56_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_56_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [56,56,56,56]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_56_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 56, i64 56, i64 56, i64 56>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_57_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_57_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [57,57]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_57_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [57,57,57,57]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_57_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [57,57,57,57]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_57_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_57_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [57,57,57,57]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_57_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 57, i64 57, i64 57, i64 57>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_57_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_57_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [57,57]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_57_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [57,57,57,57]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_57_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [57,57,57,57]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_57_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_57_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [57,57,57,57]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_57_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 57, i64 57, i64 57, i64 57>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_58_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_58_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [58,58]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_58_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [58,58,58,58]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_58_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [58,58,58,58]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_58_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_58_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [58,58,58,58]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_58_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 58, i64 58, i64 58, i64 58>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_58_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_58_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [58,58]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_58_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [58,58,58,58]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_58_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [58,58,58,58]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_58_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_58_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [58,58,58,58]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_58_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 58, i64 58, i64 58, i64 58>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_59_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_59_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [59,59]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_59_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [59,59,59,59]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_59_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [59,59,59,59]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_59_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_59_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [59,59,59,59]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_59_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 59, i64 59, i64 59, i64 59>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_59_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_59_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [59,59]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_59_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [59,59,59,59]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_59_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [59,59,59,59]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_59_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_59_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [59,59,59,59]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_59_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 59, i64 59, i64 59, i64 59>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_60_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_60_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,60]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_60_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [60,60,60,60]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_60_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [60,60,60,60]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_60_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_60_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [60,60,60,60]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_60_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 60, i64 60, i64 60, i64 60>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_60_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_60_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [60,60]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_60_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [60,60,60,60]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_60_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [60,60,60,60]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_60_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_60_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [60,60,60,60]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_60_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 60, i64 60, i64 60, i64 60>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_61_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_61_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [61,61]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_61_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [61,61,61,61]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_61_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [61,61,61,61]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_61_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_61_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [61,61,61,61]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_61_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 61, i64 61, i64 61, i64 61>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_61_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_61_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [61,61]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_61_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [61,61,61,61]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_61_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [61,61,61,61]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_61_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_61_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [61,61,61,61]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_61_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 61, i64 61, i64 61, i64 61>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_62_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_62_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [62,62]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_62_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [62,62,62,62]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_62_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [62,62,62,62]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_62_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_62_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [62,62,62,62]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_62_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 62, i64 62, i64 62, i64 62>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_62_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_62_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [62,62]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_62_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [62,62,62,62]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_62_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [62,62,62,62]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_62_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_62_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [62,62,62,62]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_62_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 62, i64 62, i64 62, i64 62>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_63_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_63_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [63,63]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_63_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [63,63,63,63]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_63_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [63,63,63,63]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_63_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_63_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [63,63,63,63]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_63_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 63, i64 63, i64 63, i64 63>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_63_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_63_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [63,63]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_63_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [63,63,63,63]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_63_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [63,63,63,63]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_63_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_63_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [63,63,63,63]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_63_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 63, i64 63, i64 63, i64 63>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_64_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_64_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [64,64]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_64_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [64,64,64,64]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_64_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [64,64,64,64]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_64_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_64_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [64,64,64,64]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_64_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 64, i64 64, i64 64, i64 64>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ugt_64_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ugt_64_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [64,64]
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm1, %xmm2, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ugt_64_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [64,64,64,64]
+; AVX2-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_64_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [64,64,64,64]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ugt_64_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ugt_64_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [64,64,64,64]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ugt_64_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ugt <4 x i64> %2, <i64 64, i64 64, i64 64, i64 64>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+define <4 x i64> @ult_65_v4i64(<4 x i64> %0) {
+; AVX1-LABEL: ult_65_v4i64:
+; AVX1:       # %bb.0:
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm2
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX1-NEXT:    vpshufb %xmm2, %xmm3, %xmm2
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm4
+; AVX1-NEXT:    vpand %xmm1, %xmm4, %xmm4
+; AVX1-NEXT:    vpshufb %xmm4, %xmm3, %xmm4
+; AVX1-NEXT:    vpaddb %xmm2, %xmm4, %xmm2
+; AVX1-NEXT:    vpxor %xmm4, %xmm4, %xmm4
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm2, %xmm2
+; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm5
+; AVX1-NEXT:    vpshufb %xmm5, %xmm3, %xmm5
+; AVX1-NEXT:    vpsrlw $4, %xmm0, %xmm0
+; AVX1-NEXT:    vpand %xmm1, %xmm0, %xmm0
+; AVX1-NEXT:    vpshufb %xmm0, %xmm3, %xmm0
+; AVX1-NEXT:    vpaddb %xmm5, %xmm0, %xmm0
+; AVX1-NEXT:    vpsadbw %xmm4, %xmm0, %xmm0
+; AVX1-NEXT:    vmovdqa {{.*#+}} xmm1 = [65,65]
+; AVX1-NEXT:    vpcmpgtq %xmm0, %xmm1, %xmm0
+; AVX1-NEXT:    vpcmpgtq %xmm2, %xmm1, %xmm1
+; AVX1-NEXT:    vinsertf128 $1, %xmm0, %ymm1, %ymm0
+; AVX1-NEXT:    retq
+;
+; AVX2-LABEL: ult_65_v4i64:
+; AVX2:       # %bb.0:
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX2-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX2-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX2-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX2-NEXT:    vpaddb %ymm2, %ymm0, %ymm0
+; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX2-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; AVX2-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [65,65,65,65]
+; AVX2-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX2-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_65_v4i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [65,65,65,65]
+; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; AVX512VPOPCNTDQVL-LABEL: ult_65_v4i64:
+; AVX512VPOPCNTDQVL:       # %bb.0:
+; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; AVX512VPOPCNTDQVL-NEXT:    retq
+;
+; BITALG_NOVLX-LABEL: ult_65_v4i64:
+; BITALG_NOVLX:       # %bb.0:
+; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
+; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG_NOVLX-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG_NOVLX-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [65,65,65,65]
+; BITALG_NOVLX-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
+; BITALG_NOVLX-NEXT:    retq
+;
+; BITALG-LABEL: ult_65_v4i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %ymm1, %ymm0, %ymm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
+; BITALG-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
+; BITALG-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
+  %3 = icmp ult <4 x i64> %2, <i64 65, i64 65, i64 65, i64 65>
+  %4 = sext <4 x i1> %3 to <4 x i64>
+  ret <4 x i64> %4
+}
+
+declare <32 x i8> @llvm.ctpop.v32i8(<32 x i8>)
+declare <16 x i16> @llvm.ctpop.v16i16(<16 x i16>)
+declare <8 x i32> @llvm.ctpop.v8i32(<8 x i32>)
+declare <4 x i64> @llvm.ctpop.v4i64(<4 x i64>)
diff --git a/llvm/test/CodeGen/X86/vector-popcnt-256.ll b/llvm/test/CodeGen/X86/vector-popcnt-256.ll
index 268c054195a67..640d6ab56efc6 100644
--- a/llvm/test/CodeGen/X86/vector-popcnt-256.ll
+++ b/llvm/test/CodeGen/X86/vector-popcnt-256.ll
@@ -341,138 +341,6 @@ define <32 x i8> @foldv32i8() nounwind {
   ret <32 x i8> %out
 }
 
-define <4 x i64> @ult_2_v4i64(<4 x i64> %0) {
-; AVX1-LABEL: ult_2_v4i64:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddq %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ult_2_v4i64:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v4i64:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [2,2,2,2]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm0, %ymm1, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ult_2_v4i64:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpltuq {{.*}}(%rip){1to4}, %ymm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v4i64:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG_NOVLX-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
-; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v4i64:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
-; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
-  %3 = icmp ult <4 x i64> %2, <i64 2, i64 2, i64 2, i64 2>
-  %4 = sext <4 x i1> %3 to <4 x i64>
-  ret <4 x i64> %4
-}
-
-define <4 x i64> @ugt_1_v4i64(<4 x i64> %0) {
-; AVX1-LABEL: ugt_1_v4i64:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddq %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddq %xmm2, %xmm0, %xmm4
-; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqq %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v4i64:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddq %ymm1, %ymm0, %ymm2
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqq %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v4i64:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpbroadcastq {{.*#+}} ymm1 = [1,1,1,1]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtq %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v4i64:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntq %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleuq {{.*}}(%rip){1to4}, %ymm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa64 %ymm0, %ymm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v4i64:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG_NOVLX-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
-; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v4i64:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG-NEXT:    vpaddq %ymm1, %ymm0, %ymm1
-; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqq %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <4 x i64> @llvm.ctpop.v4i64(<4 x i64> %0)
-  %3 = icmp ugt <4 x i64> %2, <i64 1, i64 1, i64 1, i64 1>
-  %4 = sext <4 x i1> %3 to <4 x i64>
-  ret <4 x i64> %4
-}
-
 define <4 x i64> @eq_1_v4i64(<4 x i64> %0) {
 ; AVX1-LABEL: eq_1_v4i64:
 ; AVX1:       # %bb.0:
@@ -624,138 +492,6 @@ define <4 x i64> @ne_1_v4i64(<4 x i64> %0) {
   ret <4 x i64> %4
 }
 
-define <8 x i32> @ult_2_v8i32(<8 x i32> %0) {
-; AVX1-LABEL: ult_2_v8i32:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddd %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddd %xmm2, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ult_2_v8i32:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v8i32:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm0, %ymm1, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ult_2_v8i32:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpltud {{.*}}(%rip){1to8}, %ymm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v8i32:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG_NOVLX-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
-; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v8i32:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
-; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
-  %3 = icmp ult <8 x i32> %2, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
-  %4 = sext <8 x i1> %3 to <8 x i32>
-  ret <8 x i32> %4
-}
-
-define <8 x i32> @ugt_1_v8i32(<8 x i32> %0) {
-; AVX1-LABEL: ugt_1_v8i32:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddd %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddd %xmm2, %xmm0, %xmm4
-; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqd %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v8i32:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddd %ymm1, %ymm0, %ymm2
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqd %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v8i32:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpbroadcastd {{.*#+}} ymm1 = [1,1,1,1,1,1,1,1]
-; AVX512VPOPCNTDQ-NEXT:    vpcmpgtd %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v8i32:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpopcntd %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpnleud {{.*}}(%rip){1to8}, %ymm0, %k1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm0, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vmovdqa32 %ymm0, %ymm0 {%k1} {z}
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v8i32:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG_NOVLX-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
-; BITALG_NOVLX-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG_NOVLX-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v8i32:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; BITALG-NEXT:    vpaddd %ymm1, %ymm0, %ymm1
-; BITALG-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; BITALG-NEXT:    vpcmpeqd %ymm1, %ymm0, %ymm0
-; BITALG-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <8 x i32> @llvm.ctpop.v8i32(<8 x i32> %0)
-  %3 = icmp ugt <8 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %4 = sext <8 x i1> %3 to <8 x i32>
-  ret <8 x i32> %4
-}
-
 define <8 x i32> @eq_1_v8i32(<8 x i32> %0) {
 ; AVX1-LABEL: eq_1_v8i32:
 ; AVX1:       # %bb.0:
@@ -907,135 +643,6 @@ define <8 x i32> @ne_1_v8i32(<8 x i32> %0) {
   ret <8 x i32> %4
 }
 
-define <16 x i16> @ult_2_v16i16(<16 x i16> %0) {
-; AVX1-LABEL: ult_2_v16i16:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddw %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddw %xmm2, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ult_2_v16i16:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v16i16:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ult_2_v16i16:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v16i16:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
-; BITALG_NOVLX-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v16i16:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
-; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %ymm0, %k0
-; BITALG-NEXT:    vpmovm2w %k0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
-  %3 = icmp ult <16 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
-  %4 = sext <16 x i1> %3 to <16 x i16>
-  ret <16 x i16> %4
-}
-
-define <16 x i16> @ugt_1_v16i16(<16 x i16> %0) {
-; AVX1-LABEL: ugt_1_v16i16:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddw %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddw %xmm2, %xmm0, %xmm4
-; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqw %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v16i16:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddw %ymm1, %ymm0, %ymm2
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqw %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v16i16:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v16i16:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpaddw %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v16i16:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntw %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vpcmpgtw {{.*}}(%rip), %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v16i16:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntw %ymm0, %ymm0
-; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %ymm0, %k0
-; BITALG-NEXT:    vpmovm2w %k0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <16 x i16> @llvm.ctpop.v16i16(<16 x i16> %0)
-  %3 = icmp ugt <16 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %4 = sext <16 x i1> %3 to <16 x i16>
-  ret <16 x i16> %4
-}
-
 define <16 x i16> @eq_1_v16i16(<16 x i16> %0) {
 ; AVX1-LABEL: eq_1_v16i16:
 ; AVX1:       # %bb.0:
@@ -1183,135 +790,6 @@ define <16 x i16> @ne_1_v16i16(<16 x i16> %0) {
   ret <16 x i16> %4
 }
 
-define <32 x i8> @ult_2_v32i8(<32 x i8> %0) {
-; AVX1-LABEL: ult_2_v32i8:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddb %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm2
-; AVX1-NEXT:    vpand %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ult_2_v32i8:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
-; AVX2-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX2-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v32i8:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ult_2_v32i8:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ult_2_v32i8:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
-; BITALG_NOVLX-NEXT:    vpcmpgtb %ymm0, %ymm1, %ymm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v32i8:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
-; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %ymm0, %k0
-; BITALG-NEXT:    vpmovm2b %k0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
-  %3 = icmp ult <32 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
-  %4 = sext <32 x i1> %3 to <32 x i8>
-  ret <32 x i8> %4
-}
-
-define <32 x i8> @ugt_1_v32i8(<32 x i8> %0) {
-; AVX1-LABEL: ugt_1_v32i8:
-; AVX1:       # %bb.0:
-; AVX1-NEXT:    vextractf128 $1, %ymm0, %xmm1
-; AVX1-NEXT:    vpcmpeqd %xmm2, %xmm2, %xmm2
-; AVX1-NEXT:    vpaddb %xmm2, %xmm1, %xmm3
-; AVX1-NEXT:    vpand %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm1, %xmm1
-; AVX1-NEXT:    vpxor %xmm2, %xmm1, %xmm1
-; AVX1-NEXT:    vpaddb %xmm2, %xmm0, %xmm4
-; AVX1-NEXT:    vpand %xmm4, %xmm0, %xmm0
-; AVX1-NEXT:    vpcmpeqb %xmm3, %xmm0, %xmm0
-; AVX1-NEXT:    vpxor %xmm2, %xmm0, %xmm0
-; AVX1-NEXT:    vinsertf128 $1, %xmm1, %ymm0, %ymm0
-; AVX1-NEXT:    retq
-;
-; AVX2-LABEL: ugt_1_v32i8:
-; AVX2:       # %bb.0:
-; AVX2-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX2-NEXT:    vpaddb %ymm1, %ymm0, %ymm2
-; AVX2-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %xmm2, %xmm2, %xmm2
-; AVX2-NEXT:    vpcmpeqb %ymm2, %ymm0, %ymm0
-; AVX2-NEXT:    vpxor %ymm1, %ymm0, %ymm0
-; AVX2-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v32i8:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQ-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQ-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    # kill: def $ymm0 killed $ymm0 killed $zmm0
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; AVX512VPOPCNTDQVL-LABEL: ugt_1_v32i8:
-; AVX512VPOPCNTDQVL:       # %bb.0:
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqd %ymm1, %ymm1, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpaddb %ymm1, %ymm0, %ymm1
-; AVX512VPOPCNTDQVL-NEXT:    vpand %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpxor %xmm1, %xmm1, %xmm1
-; AVX512VPOPCNTDQVL-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    vpternlogq $15, %ymm0, %ymm0, %ymm0
-; AVX512VPOPCNTDQVL-NEXT:    retq
-;
-; BITALG_NOVLX-LABEL: ugt_1_v32i8:
-; BITALG_NOVLX:       # %bb.0:
-; BITALG_NOVLX-NEXT:    # kill: def $ymm0 killed $ymm0 def $zmm0
-; BITALG_NOVLX-NEXT:    vpopcntb %zmm0, %zmm0
-; BITALG_NOVLX-NEXT:    vpcmpgtb {{.*}}(%rip), %ymm0, %ymm0
-; BITALG_NOVLX-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v32i8:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntb %ymm0, %ymm0
-; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %ymm0, %k0
-; BITALG-NEXT:    vpmovm2b %k0, %ymm0
-; BITALG-NEXT:    retq
-  %2 = tail call <32 x i8> @llvm.ctpop.v32i8(<32 x i8> %0)
-  %3 = icmp ugt <32 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
-  %4 = sext <32 x i1> %3 to <32 x i8>
-  ret <32 x i8> %4
-}
-
 define <32 x i8> @eq_1_v32i8(<32 x i8> %0) {
 ; AVX1-LABEL: eq_1_v32i8:
 ; AVX1:       # %bb.0:
diff --git a/llvm/test/CodeGen/X86/vector-popcnt-512-ult-ugt.ll b/llvm/test/CodeGen/X86/vector-popcnt-512-ult-ugt.ll
new file mode 100644
index 0000000000000..939dabd5c2179
--- /dev/null
+++ b/llvm/test/CodeGen/X86/vector-popcnt-512-ult-ugt.ll
@@ -0,0 +1,17920 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512f | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512F
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bw | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512BW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vpopcntdq | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512VPOPCNTDQ --check-prefix=AVX512VPOPCNTDQ-NOBW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512vpopcntdq,+avx512bw | FileCheck %s --check-prefix=AVX512 --check-prefix=AVX512VPOPCNTDQ --check-prefix=AVX512VPOPCNTDQ-BW
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=+avx512bitalg | FileCheck %s --check-prefix=AVX512 --check-prefix=BITALG
+
+
+define <64 x i8> @ult_0_v64i8(<64 x i8> %0) {
+; AVX512-LABEL: ult_0_v64i8:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX512-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_0_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_0_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_0_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestmb %zmm0, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_0_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_0_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestmb %zmm0, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vptestmb %zmm0, %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_1_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_1_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_1_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_1_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_1_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vptestnmb %zmm0, %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_1_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_1_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_1_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestmb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_1_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_1_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestmb %zmm1, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_2_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_2_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_2_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestnmb %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_2_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_2_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestnmb %zmm1, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_2_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_2_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_2_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_2_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_2_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_3_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_3_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_3_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_3_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_3_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_3_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_3_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_3_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_3_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_3_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3, i8 3>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_4_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_4_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_4_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_4_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_4_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_4_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_4_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_4_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_4_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_4_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_5_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_5_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_5_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_5_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_5_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_5_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_5_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_5_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_5_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_5_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5, i8 5>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_6_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_6_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_6_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_6_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_6_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_6_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_6_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_6_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_6_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_6_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6, i8 6>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_7_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_7_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_7_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_7_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_7_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_7_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_7_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_7_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_7_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_7_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7, i8 7>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_8_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_8_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_8_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_8_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_8_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ugt_8_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ugt_8_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_8_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_8_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmaxub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_8_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ugt <64 x i8> %2, <i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8, i8 8>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <64 x i8> @ult_9_v64i8(<64 x i8> %0) {
+; AVX512F-LABEL: ult_9_v64i8:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_9_v64i8:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_9_v64i8:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm0, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpminub %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_9_v64i8:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v64i8:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2b %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
+  %3 = icmp ult <64 x i8> %2, <i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9, i8 9>
+  %4 = sext <64 x i1> %3 to <64 x i8>
+  ret <64 x i8> %4
+}
+
+define <32 x i16> @ult_0_v32i16(<32 x i16> %0) {
+; AVX512-LABEL: ult_0_v32i16:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX512-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_0_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_0_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512F-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_0_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestmw %zmm0, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_0_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_0_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestmw %zmm0, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vptestmw %zmm0, %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_1_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_1_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512F-NEXT:    vpcmpeqw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_1_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_1_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm2, %xmm2, %xmm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_1_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vptestnmw %zmm0, %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_1_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_1_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512F-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_1_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestmw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_1_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_1_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestmw %zmm1, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_2_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_2_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512F-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_2_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestnmw %zmm1, %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_2_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_2_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vptestnmw %zmm1, %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_2_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_2_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_2_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_2_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_2_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_3_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_3_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_3_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_3_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_3_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_3_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_3_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_3_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_3_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [3,3,3,3,3,3,3,3,3,3,3,3,3,3,3,3]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_3_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3, i16 3>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_4_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_4_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_4_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_4_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_4_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_4_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_4_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_4_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_4_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [4,4,4,4,4,4,4,4,4,4,4,4,4,4,4,4]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_4_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4, i16 4>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_5_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_5_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_5_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_5_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_5_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_5_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_5_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_5_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_5_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [5,5,5,5,5,5,5,5,5,5,5,5,5,5,5,5]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_5_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5, i16 5>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_6_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_6_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_6_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_6_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_6_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_6_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_6_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_6_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_6_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [6,6,6,6,6,6,6,6,6,6,6,6,6,6,6,6]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_6_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6, i16 6>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_7_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_7_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_7_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_7_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_7_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_7_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_7_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_7_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_7_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_7_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7, i16 7>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_8_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_8_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_8_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_8_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_8_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_8_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_8_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_8_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_8_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [8,8,8,8,8,8,8,8,8,8,8,8,8,8,8,8]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_8_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8, i16 8>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_9_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_9_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_9_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_9_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_9_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_9_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_9_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_9_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_9_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_9_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9, i16 9>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_10_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_10_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_10_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_10_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_10_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_10_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_10_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_10_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_10_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [10,10,10,10,10,10,10,10,10,10,10,10,10,10,10,10]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_10_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10, i16 10>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_11_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_11_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_11_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_11_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_11_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_11_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_11_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_11_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_11_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [11,11,11,11,11,11,11,11,11,11,11,11,11,11,11,11]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_11_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11, i16 11>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_12_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_12_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_12_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_12_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_12_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_12_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_12_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_12_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_12_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [12,12,12,12,12,12,12,12,12,12,12,12,12,12,12,12]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_12_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12, i16 12>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_13_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_13_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_13_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_13_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_13_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_13_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_13_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_13_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_13_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [13,13,13,13,13,13,13,13,13,13,13,13,13,13,13,13]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_13_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13, i16 13>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_14_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_14_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_14_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_14_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_14_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_14_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_14_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_14_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_14_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [14,14,14,14,14,14,14,14,14,14,14,14,14,14,14,14]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_14_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14, i16 14>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_15_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_15_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_15_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_15_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_15_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_15_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_15_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_15_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_15_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_15_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15, i16 15>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_16_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_16_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_16_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_16_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_16_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ugt_16_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ugt_16_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_16_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_16_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [16,16,16,16,16,16,16,16,16,16,16,16,16,16,16,16]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ugt_16_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ugt <32 x i16> %2, <i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16, i16 16>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <32 x i16> @ult_17_v32i16(<32 x i16> %0) {
+; AVX512F-LABEL: ult_17_v32i16:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm2
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm2, %ymm3, %ymm2
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm4
+; AVX512F-NEXT:    vpand %ymm1, %ymm4, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsllw $8, %ymm2, %ymm4
+; AVX512F-NEXT:    vpaddb %ymm2, %ymm4, %ymm2
+; AVX512F-NEXT:    vpsrlw $8, %ymm2, %ymm2
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm4
+; AVX512F-NEXT:    vpshufb %ymm4, %ymm3, %ymm4
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm1, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm3, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsllw $8, %ymm0, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpsrlw $8, %ymm0, %ymm0
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm1 = [17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17]
+; AVX512F-NEXT:    vpcmpgtw %ymm0, %ymm1, %ymm0
+; AVX512F-NEXT:    vpcmpgtw %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_17_v32i16:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-NOBW-LABEL: ult_17_v32i16:
+; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm1 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm1, %zmm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm1, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovzxwd {{.*#+}} zmm0 = ymm0[0],zero,ymm0[1],zero,ymm0[2],zero,ymm0[3],zero,ymm0[4],zero,ymm0[5],zero,ymm0[6],zero,ymm0[7],zero,ymm0[8],zero,ymm0[9],zero,ymm0[10],zero,ymm0[11],zero,ymm0[12],zero,ymm0[13],zero,ymm0[14],zero,ymm0[15],zero
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpmovdw %zmm0, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vmovdqa {{.*#+}} ymm2 = [17,17,17,17,17,17,17,17,17,17,17,17,17,17,17,17]
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm0, %ymm2, %ymm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpgtw %ymm1, %ymm2, %ymm1
+; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-BW-LABEL: ult_17_v32i16:
+; AVX512VPOPCNTDQ-BW:       # %bb.0:
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsllw $8, %zmm0, %zmm1
+; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm0, %zmm1, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpsrlw $8, %zmm0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
+; AVX512VPOPCNTDQ-BW-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v32i16:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
+; BITALG-NEXT:    vpmovm2w %k0, %zmm0
+; BITALG-NEXT:    retq
+  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
+  %3 = icmp ult <32 x i16> %2, <i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17, i16 17>
+  %4 = sext <32 x i1> %3 to <32 x i16>
+  ret <32 x i16> %4
+}
+
+define <16 x i32> @ult_0_v16i32(<16 x i32> %0) {
+; AVX512-LABEL: ult_0_v16i32:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX512-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_0_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_0_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_0_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vptestmd %zmm0, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_1_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_1_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestnmd %zmm0, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_1_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestnmd %zmm0, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vptestnmd %zmm0, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vptestnmd %zmm0, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_1_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_1_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512F-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
+; AVX512F-NEXT:    vptestmd %zmm1, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_1_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestmd %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; BITALG-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
+; BITALG-NEXT:    vptestmd %zmm1, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_2_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_2_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512F-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
+; AVX512F-NEXT:    vptestnmd %zmm1, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_2_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestnmd %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; BITALG-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
+; BITALG-NEXT:    vptestnmd %zmm1, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_2_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_2_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_2_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_3_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_3_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_3_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_3_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_3_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_3_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_4_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_4_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_4_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_4_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_4_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_4_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4, i32 4>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_5_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_5_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_5_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_5_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_5_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_5_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5, i32 5>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_6_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_6_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_6_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_6_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_6_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_6_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6, i32 6>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_7_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_7_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_7_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_7_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_7_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_7_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7, i32 7>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_8_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_8_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_8_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_8_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_8_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_8_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_9_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_9_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_9_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_9_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_9_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_9_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9, i32 9>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_10_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_10_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_10_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_10_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_10_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_10_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10, i32 10>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_11_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_11_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_11_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_11_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_11_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_11_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_12_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_12_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_12_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_12_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_12_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_12_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12, i32 12>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_13_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_13_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_13_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_13_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_13_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_13_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13, i32 13>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_14_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_14_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_14_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_14_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_14_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_14_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14, i32 14>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_15_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_15_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_15_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_15_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_15_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_15_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15, i32 15>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_16_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_16_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_16_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_16_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_16_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_16_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_17_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_17_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_17_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_17_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_17_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_17_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_17_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_17_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17, i32 17>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_18_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_18_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_18_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_18_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_18_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_18_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_18_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_18_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_18_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_18_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18, i32 18>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_19_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_19_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_19_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_19_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_19_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_19_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_19_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_19_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_19_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_19_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19, i32 19>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_20_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_20_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_20_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_20_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_20_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_20_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_20_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_20_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_20_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_20_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20, i32 20>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_21_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_21_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_21_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_21_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_21_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_21_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_21_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_21_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_21_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_21_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21, i32 21>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_22_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_22_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_22_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_22_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_22_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_22_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_22_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_22_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_22_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_22_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22, i32 22>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_23_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_23_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_23_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_23_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_23_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_23_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_23_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_23_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_23_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_23_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23, i32 23>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_24_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_24_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_24_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_24_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_24_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_24_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_24_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_24_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_24_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_24_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24, i32 24>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_25_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_25_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_25_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_25_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_25_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_25_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_25_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_25_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_25_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_25_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25, i32 25>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_26_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_26_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_26_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_26_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_26_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_26_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_26_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_26_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_26_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_26_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26, i32 26>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_27_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_27_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_27_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_27_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_27_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_27_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_27_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_27_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_27_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_27_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27, i32 27>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_28_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_28_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_28_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_28_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_28_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_28_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_28_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_28_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_28_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_28_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28, i32 28>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_29_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_29_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_29_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_29_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_29_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_29_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_29_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_29_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_29_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_29_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29, i32 29>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_30_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_30_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_30_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_30_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_30_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_30_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_30_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_30_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_30_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_30_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30, i32 30>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_31_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_31_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_31_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_31_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_31_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_31_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_31_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_31_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_31_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_31_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31, i32 31>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_32_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_32_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_32_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_32_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_32_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ugt_32_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ugt_32_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_32_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_32_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_32_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ugt <16 x i32> %2, <i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32, i32 32>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <16 x i32> @ult_33_v16i32(<16 x i32> %0) {
+; AVX512F-LABEL: ult_33_v16i32:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm5 = ymm1[2],ymm3[2],ymm1[3],ymm3[3],ymm1[6],ymm3[6],ymm1[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm5, %ymm5
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm1 = ymm1[0],ymm3[0],ymm1[1],ymm3[1],ymm1[4],ymm3[4],ymm1[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpackuswb %ymm5, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpunpckhdq {{.*#+}} ymm2 = ymm0[2],ymm3[2],ymm0[3],ymm3[3],ymm0[6],ymm3[6],ymm0[7],ymm3[7]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm2, %ymm2
+; AVX512F-NEXT:    vpunpckldq {{.*#+}} ymm0 = ymm0[0],ymm3[0],ymm0[1],ymm3[1],ymm0[4],ymm3[4],ymm0[5],ymm3[5]
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vpackuswb %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_33_v16i32:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; AVX512BW-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_33_v16i32:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_33_v16i32:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpunpckhdq {{.*#+}} zmm2 = zmm0[2],zmm1[2],zmm0[3],zmm1[3],zmm0[6],zmm1[6],zmm0[7],zmm1[7],zmm0[10],zmm1[10],zmm0[11],zmm1[11],zmm0[14],zmm1[14],zmm0[15],zmm1[15]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm2, %zmm2
+; BITALG-NEXT:    vpunpckldq {{.*#+}} zmm0 = zmm0[0],zmm1[0],zmm0[1],zmm1[1],zmm0[4],zmm1[4],zmm0[5],zmm1[5],zmm0[8],zmm1[8],zmm0[9],zmm1[9],zmm0[12],zmm1[12],zmm0[13],zmm1[13]
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpackuswb %zmm2, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
+  %3 = icmp ult <16 x i32> %2, <i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33, i32 33>
+  %4 = sext <16 x i1> %3 to <16 x i32>
+  ret <16 x i32> %4
+}
+
+define <8 x i64> @ult_0_v8i64(<8 x i64> %0) {
+; AVX512-LABEL: ult_0_v8i64:
+; AVX512:       # %bb.0:
+; AVX512-NEXT:    vxorps %xmm0, %xmm0, %xmm0
+; AVX512-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_0_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_0_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_0_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_0_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_0_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vptestmq %zmm0, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_1_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_1_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_1_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_1_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_1_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vptestnmq %zmm0, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_1_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_1_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512F-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
+; AVX512F-NEXT:    vptestmq %zmm1, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_1_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestmq %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_1_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_1_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; BITALG-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
+; BITALG-NEXT:    vptestmq %zmm1, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_2_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_2_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512F-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
+; AVX512F-NEXT:    vptestnmq %zmm1, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_2_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; AVX512BW-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
+; AVX512BW-NEXT:    vptestnmq %zmm1, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_2_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_2_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
+; BITALG-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
+; BITALG-NEXT:    vptestnmq %zmm1, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_2_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_2_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_2_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_2_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_2_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_3_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_3_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_3_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_3_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_3_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 3, i64 3, i64 3, i64 3, i64 3, i64 3, i64 3, i64 3>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_3_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_3_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_3_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_3_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_3_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 3, i64 3, i64 3, i64 3, i64 3, i64 3, i64 3, i64 3>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_4_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_4_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_4_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_4_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_4_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_4_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_4_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_4_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_4_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_4_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4, i64 4>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_5_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_5_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_5_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_5_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_5_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_5_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_5_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_5_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_5_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_5_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5, i64 5>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_6_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_6_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_6_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_6_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_6_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_6_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_6_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_6_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_6_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_6_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6, i64 6>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_7_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_7_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_7_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_7_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_7_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_7_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_7_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_7_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_7_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_7_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7, i64 7>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_8_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_8_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_8_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_8_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_8_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_8_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_8_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_8_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_8_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_8_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_9_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_9_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_9_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_9_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_9_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 9, i64 9, i64 9, i64 9, i64 9, i64 9, i64 9, i64 9>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_9_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_9_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_9_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_9_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_9_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 9, i64 9, i64 9, i64 9, i64 9, i64 9, i64 9, i64 9>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_10_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_10_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_10_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_10_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_10_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 10, i64 10, i64 10, i64 10, i64 10, i64 10, i64 10, i64 10>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_10_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_10_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_10_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_10_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_10_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 10, i64 10, i64 10, i64 10, i64 10, i64 10, i64 10, i64 10>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_11_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_11_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_11_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_11_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_11_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 11, i64 11, i64 11, i64 11, i64 11, i64 11, i64 11, i64 11>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_11_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_11_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_11_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_11_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_11_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 11, i64 11, i64 11, i64 11, i64 11, i64 11, i64 11, i64 11>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_12_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_12_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_12_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_12_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_12_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 12, i64 12, i64 12, i64 12, i64 12, i64 12, i64 12, i64 12>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_12_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_12_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_12_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_12_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_12_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 12, i64 12, i64 12, i64 12, i64 12, i64 12, i64 12, i64 12>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_13_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_13_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_13_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_13_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_13_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 13, i64 13, i64 13, i64 13, i64 13, i64 13, i64 13, i64 13>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_13_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_13_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_13_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_13_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_13_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 13, i64 13, i64 13, i64 13, i64 13, i64 13, i64 13, i64 13>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_14_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_14_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_14_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_14_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_14_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 14, i64 14, i64 14, i64 14, i64 14, i64 14, i64 14, i64 14>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_14_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_14_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_14_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_14_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_14_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 14, i64 14, i64 14, i64 14, i64 14, i64 14, i64 14, i64 14>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_15_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_15_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_15_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_15_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_15_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 15, i64 15, i64 15, i64 15, i64 15, i64 15, i64 15, i64 15>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_15_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_15_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_15_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_15_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_15_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 15, i64 15, i64 15, i64 15, i64 15, i64 15, i64 15, i64 15>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_16_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_16_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_16_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_16_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_16_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_16_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_16_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_16_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_16_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_16_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16, i64 16>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_17_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_17_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_17_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_17_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_17_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 17, i64 17, i64 17, i64 17, i64 17, i64 17, i64 17, i64 17>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_17_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_17_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_17_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_17_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_17_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 17, i64 17, i64 17, i64 17, i64 17, i64 17, i64 17, i64 17>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_18_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_18_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_18_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_18_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_18_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 18, i64 18, i64 18, i64 18, i64 18, i64 18, i64 18, i64 18>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_18_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_18_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_18_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_18_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_18_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 18, i64 18, i64 18, i64 18, i64 18, i64 18, i64 18, i64 18>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_19_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_19_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_19_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_19_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_19_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 19, i64 19, i64 19, i64 19, i64 19, i64 19, i64 19, i64 19>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_19_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_19_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_19_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_19_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_19_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 19, i64 19, i64 19, i64 19, i64 19, i64 19, i64 19, i64 19>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_20_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_20_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_20_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_20_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_20_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 20, i64 20, i64 20, i64 20, i64 20, i64 20, i64 20, i64 20>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_20_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_20_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_20_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_20_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_20_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 20, i64 20, i64 20, i64 20, i64 20, i64 20, i64 20, i64 20>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_21_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_21_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_21_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_21_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_21_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 21, i64 21, i64 21, i64 21, i64 21, i64 21, i64 21, i64 21>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_21_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_21_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_21_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_21_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_21_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 21, i64 21, i64 21, i64 21, i64 21, i64 21, i64 21, i64 21>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_22_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_22_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_22_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_22_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_22_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 22, i64 22, i64 22, i64 22, i64 22, i64 22, i64 22, i64 22>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_22_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_22_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_22_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_22_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_22_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 22, i64 22, i64 22, i64 22, i64 22, i64 22, i64 22, i64 22>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_23_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_23_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_23_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_23_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_23_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 23, i64 23, i64 23, i64 23, i64 23, i64 23, i64 23, i64 23>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_23_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_23_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_23_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_23_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_23_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 23, i64 23, i64 23, i64 23, i64 23, i64 23, i64 23, i64 23>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_24_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_24_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_24_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_24_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_24_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 24, i64 24, i64 24, i64 24, i64 24, i64 24, i64 24, i64 24>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_24_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_24_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_24_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_24_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_24_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 24, i64 24, i64 24, i64 24, i64 24, i64 24, i64 24, i64 24>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_25_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_25_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_25_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_25_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_25_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 25, i64 25, i64 25, i64 25, i64 25, i64 25, i64 25, i64 25>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_25_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_25_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_25_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_25_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_25_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 25, i64 25, i64 25, i64 25, i64 25, i64 25, i64 25, i64 25>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_26_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_26_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_26_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_26_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_26_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 26, i64 26, i64 26, i64 26, i64 26, i64 26, i64 26, i64 26>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_26_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_26_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_26_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_26_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_26_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 26, i64 26, i64 26, i64 26, i64 26, i64 26, i64 26, i64 26>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_27_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_27_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_27_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_27_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_27_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 27, i64 27, i64 27, i64 27, i64 27, i64 27, i64 27, i64 27>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_27_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_27_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_27_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_27_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_27_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 27, i64 27, i64 27, i64 27, i64 27, i64 27, i64 27, i64 27>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_28_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_28_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_28_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_28_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_28_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 28, i64 28, i64 28, i64 28, i64 28, i64 28, i64 28, i64 28>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_28_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_28_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_28_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_28_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_28_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 28, i64 28, i64 28, i64 28, i64 28, i64 28, i64 28, i64 28>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_29_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_29_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_29_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_29_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_29_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 29, i64 29, i64 29, i64 29, i64 29, i64 29, i64 29, i64 29>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_29_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_29_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_29_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_29_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_29_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 29, i64 29, i64 29, i64 29, i64 29, i64 29, i64 29, i64 29>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_30_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_30_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_30_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_30_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_30_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 30, i64 30, i64 30, i64 30, i64 30, i64 30, i64 30, i64 30>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_30_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_30_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_30_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_30_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_30_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 30, i64 30, i64 30, i64 30, i64 30, i64 30, i64 30, i64 30>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_31_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_31_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_31_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_31_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_31_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_31_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_31_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_31_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_31_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_31_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31, i64 31>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_32_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_32_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_32_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_32_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_32_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_32_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_32_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_32_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_32_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_32_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32, i64 32>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_33_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_33_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_33_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_33_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_33_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 33, i64 33, i64 33, i64 33, i64 33, i64 33, i64 33, i64 33>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_33_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_33_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_33_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_33_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_33_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 33, i64 33, i64 33, i64 33, i64 33, i64 33, i64 33, i64 33>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_34_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_34_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_34_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_34_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_34_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 34, i64 34, i64 34, i64 34, i64 34, i64 34, i64 34, i64 34>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_34_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_34_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_34_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_34_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_34_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 34, i64 34, i64 34, i64 34, i64 34, i64 34, i64 34, i64 34>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_35_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_35_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_35_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_35_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_35_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 35, i64 35, i64 35, i64 35, i64 35, i64 35, i64 35, i64 35>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_35_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_35_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_35_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_35_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_35_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 35, i64 35, i64 35, i64 35, i64 35, i64 35, i64 35, i64 35>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_36_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_36_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_36_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_36_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_36_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 36, i64 36, i64 36, i64 36, i64 36, i64 36, i64 36, i64 36>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_36_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_36_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_36_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_36_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_36_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 36, i64 36, i64 36, i64 36, i64 36, i64 36, i64 36, i64 36>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_37_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_37_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_37_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_37_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_37_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 37, i64 37, i64 37, i64 37, i64 37, i64 37, i64 37, i64 37>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_37_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_37_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_37_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_37_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_37_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 37, i64 37, i64 37, i64 37, i64 37, i64 37, i64 37, i64 37>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_38_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_38_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_38_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_38_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_38_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 38, i64 38, i64 38, i64 38, i64 38, i64 38, i64 38, i64 38>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_38_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_38_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_38_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_38_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_38_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 38, i64 38, i64 38, i64 38, i64 38, i64 38, i64 38, i64 38>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_39_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_39_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_39_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_39_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_39_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 39, i64 39, i64 39, i64 39, i64 39, i64 39, i64 39, i64 39>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_39_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_39_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_39_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_39_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_39_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 39, i64 39, i64 39, i64 39, i64 39, i64 39, i64 39, i64 39>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_40_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_40_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_40_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_40_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_40_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 40, i64 40, i64 40, i64 40, i64 40, i64 40, i64 40, i64 40>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_40_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_40_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_40_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_40_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_40_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 40, i64 40, i64 40, i64 40, i64 40, i64 40, i64 40, i64 40>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_41_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_41_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_41_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_41_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_41_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 41, i64 41, i64 41, i64 41, i64 41, i64 41, i64 41, i64 41>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_41_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_41_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_41_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_41_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_41_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 41, i64 41, i64 41, i64 41, i64 41, i64 41, i64 41, i64 41>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_42_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_42_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_42_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_42_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_42_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 42, i64 42, i64 42, i64 42, i64 42, i64 42, i64 42, i64 42>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_42_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_42_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_42_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_42_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_42_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 42, i64 42, i64 42, i64 42, i64 42, i64 42, i64 42, i64 42>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_43_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_43_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_43_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_43_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_43_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 43, i64 43, i64 43, i64 43, i64 43, i64 43, i64 43, i64 43>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_43_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_43_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_43_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_43_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_43_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 43, i64 43, i64 43, i64 43, i64 43, i64 43, i64 43, i64 43>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_44_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_44_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_44_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_44_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_44_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 44, i64 44, i64 44, i64 44, i64 44, i64 44, i64 44, i64 44>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_44_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_44_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_44_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_44_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_44_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 44, i64 44, i64 44, i64 44, i64 44, i64 44, i64 44, i64 44>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_45_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_45_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_45_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_45_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_45_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 45, i64 45, i64 45, i64 45, i64 45, i64 45, i64 45, i64 45>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_45_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_45_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_45_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_45_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_45_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 45, i64 45, i64 45, i64 45, i64 45, i64 45, i64 45, i64 45>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_46_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_46_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_46_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_46_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_46_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 46, i64 46, i64 46, i64 46, i64 46, i64 46, i64 46, i64 46>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_46_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_46_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_46_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_46_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_46_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 46, i64 46, i64 46, i64 46, i64 46, i64 46, i64 46, i64 46>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_47_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_47_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_47_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_47_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_47_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 47, i64 47, i64 47, i64 47, i64 47, i64 47, i64 47, i64 47>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_47_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_47_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_47_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_47_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_47_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 47, i64 47, i64 47, i64 47, i64 47, i64 47, i64 47, i64 47>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_48_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_48_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_48_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_48_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_48_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 48, i64 48, i64 48, i64 48, i64 48, i64 48, i64 48, i64 48>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_48_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_48_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_48_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_48_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_48_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 48, i64 48, i64 48, i64 48, i64 48, i64 48, i64 48, i64 48>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_49_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_49_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_49_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_49_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_49_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 49, i64 49, i64 49, i64 49, i64 49, i64 49, i64 49, i64 49>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_49_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_49_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_49_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_49_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_49_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 49, i64 49, i64 49, i64 49, i64 49, i64 49, i64 49, i64 49>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_50_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_50_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_50_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_50_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_50_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 50, i64 50, i64 50, i64 50, i64 50, i64 50, i64 50, i64 50>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_50_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_50_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_50_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_50_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_50_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 50, i64 50, i64 50, i64 50, i64 50, i64 50, i64 50, i64 50>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_51_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_51_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_51_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_51_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_51_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 51, i64 51, i64 51, i64 51, i64 51, i64 51, i64 51, i64 51>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_51_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_51_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_51_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_51_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_51_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 51, i64 51, i64 51, i64 51, i64 51, i64 51, i64 51, i64 51>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_52_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_52_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_52_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_52_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_52_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 52, i64 52, i64 52, i64 52, i64 52, i64 52, i64 52, i64 52>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_52_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_52_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_52_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_52_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_52_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 52, i64 52, i64 52, i64 52, i64 52, i64 52, i64 52, i64 52>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_53_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_53_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_53_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_53_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_53_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 53, i64 53, i64 53, i64 53, i64 53, i64 53, i64 53, i64 53>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_53_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_53_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_53_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_53_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_53_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 53, i64 53, i64 53, i64 53, i64 53, i64 53, i64 53, i64 53>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_54_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_54_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_54_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_54_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_54_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 54, i64 54, i64 54, i64 54, i64 54, i64 54, i64 54, i64 54>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_54_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_54_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_54_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_54_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_54_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 54, i64 54, i64 54, i64 54, i64 54, i64 54, i64 54, i64 54>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_55_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_55_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_55_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_55_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_55_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 55, i64 55, i64 55, i64 55, i64 55, i64 55, i64 55, i64 55>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_55_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_55_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_55_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_55_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_55_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 55, i64 55, i64 55, i64 55, i64 55, i64 55, i64 55, i64 55>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_56_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_56_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_56_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_56_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_56_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 56, i64 56, i64 56, i64 56, i64 56, i64 56, i64 56, i64 56>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_56_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_56_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_56_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_56_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_56_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 56, i64 56, i64 56, i64 56, i64 56, i64 56, i64 56, i64 56>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_57_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_57_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_57_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_57_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_57_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 57, i64 57, i64 57, i64 57, i64 57, i64 57, i64 57, i64 57>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_57_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_57_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_57_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_57_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_57_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 57, i64 57, i64 57, i64 57, i64 57, i64 57, i64 57, i64 57>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_58_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_58_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_58_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_58_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_58_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 58, i64 58, i64 58, i64 58, i64 58, i64 58, i64 58, i64 58>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_58_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_58_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_58_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_58_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_58_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 58, i64 58, i64 58, i64 58, i64 58, i64 58, i64 58, i64 58>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_59_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_59_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_59_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_59_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_59_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 59, i64 59, i64 59, i64 59, i64 59, i64 59, i64 59, i64 59>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_59_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_59_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_59_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_59_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_59_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 59, i64 59, i64 59, i64 59, i64 59, i64 59, i64 59, i64 59>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_60_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_60_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_60_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_60_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_60_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 60, i64 60, i64 60, i64 60, i64 60, i64 60, i64 60, i64 60>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_60_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_60_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_60_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_60_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_60_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 60, i64 60, i64 60, i64 60, i64 60, i64 60, i64 60, i64 60>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_61_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_61_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_61_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_61_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_61_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 61, i64 61, i64 61, i64 61, i64 61, i64 61, i64 61, i64 61>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_61_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_61_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_61_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_61_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_61_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 61, i64 61, i64 61, i64 61, i64 61, i64 61, i64 61, i64 61>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_62_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_62_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_62_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_62_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_62_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 62, i64 62, i64 62, i64 62, i64 62, i64 62, i64 62, i64 62>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_62_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_62_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_62_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_62_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_62_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 62, i64 62, i64 62, i64 62, i64 62, i64 62, i64 62, i64 62>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_63_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_63_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_63_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_63_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_63_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 63, i64 63, i64 63, i64 63, i64 63, i64 63, i64 63, i64 63>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_63_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_63_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_63_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_63_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_63_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 63, i64 63, i64 63, i64 63, i64 63, i64 63, i64 63, i64 63>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_64_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_64_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_64_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_64_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_64_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 64, i64 64, i64 64, i64 64, i64 64, i64 64, i64 64, i64 64>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ugt_64_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ugt_64_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ugt_64_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ugt_64_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ugt_64_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ugt <8 x i64> %2, <i64 64, i64 64, i64 64, i64 64, i64 64, i64 64, i64 64, i64 64>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+define <8 x i64> @ult_65_v8i64(<8 x i64> %0) {
+; AVX512F-LABEL: ult_65_v8i64:
+; AVX512F:       # %bb.0:
+; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm2 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm3
+; AVX512F-NEXT:    vmovdqa {{.*#+}} ymm4 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512F-NEXT:    vpshufb %ymm3, %ymm4, %ymm3
+; AVX512F-NEXT:    vpsrlw $4, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm1, %ymm1
+; AVX512F-NEXT:    vpshufb %ymm1, %ymm4, %ymm1
+; AVX512F-NEXT:    vpaddb %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm1, %ymm1
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm5
+; AVX512F-NEXT:    vpshufb %ymm5, %ymm4, %ymm5
+; AVX512F-NEXT:    vpsrlw $4, %ymm0, %ymm0
+; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
+; AVX512F-NEXT:    vpshufb %ymm0, %ymm4, %ymm0
+; AVX512F-NEXT:    vpaddb %ymm5, %ymm0, %ymm0
+; AVX512F-NEXT:    vpsadbw %ymm3, %ymm0, %ymm0
+; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
+; AVX512F-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512F-NEXT:    retq
+;
+; AVX512BW-LABEL: ult_65_v8i64:
+; AVX512BW:       # %bb.0:
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm1 = [15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15,15]
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm2
+; AVX512BW-NEXT:    vmovdqa64 {{.*#+}} zmm3 = [0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4,0,1,1,2,1,2,2,3,1,2,2,3,2,3,3,4]
+; AVX512BW-NEXT:    vpshufb %zmm2, %zmm3, %zmm2
+; AVX512BW-NEXT:    vpsrlw $4, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpandq %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpshufb %zmm0, %zmm3, %zmm0
+; AVX512BW-NEXT:    vpaddb %zmm2, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; AVX512BW-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; AVX512BW-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512BW-NEXT:    retq
+;
+; AVX512VPOPCNTDQ-LABEL: ult_65_v8i64:
+; AVX512VPOPCNTDQ:       # %bb.0:
+; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
+; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; AVX512VPOPCNTDQ-NEXT:    retq
+;
+; BITALG-LABEL: ult_65_v8i64:
+; BITALG:       # %bb.0:
+; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
+; BITALG-NEXT:    vpxor %xmm1, %xmm1, %xmm1
+; BITALG-NEXT:    vpsadbw %zmm1, %zmm0, %zmm0
+; BITALG-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
+; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
+; BITALG-NEXT:    retq
+  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
+  %3 = icmp ult <8 x i64> %2, <i64 65, i64 65, i64 65, i64 65, i64 65, i64 65, i64 65, i64 65>
+  %4 = sext <8 x i1> %3 to <8 x i64>
+  ret <8 x i64> %4
+}
+
+declare <64 x i8> @llvm.ctpop.v64i8(<64 x i8>)
+declare <32 x i16> @llvm.ctpop.v32i16(<32 x i16>)
+declare <16 x i32> @llvm.ctpop.v16i32(<16 x i32>)
+declare <8 x i64> @llvm.ctpop.v8i64(<8 x i64>)
diff --git a/llvm/test/CodeGen/X86/vector-popcnt-512.ll b/llvm/test/CodeGen/X86/vector-popcnt-512.ll
index dfca538206039..ddcd9eeb3659c 100644
--- a/llvm/test/CodeGen/X86/vector-popcnt-512.ll
+++ b/llvm/test/CodeGen/X86/vector-popcnt-512.ll
@@ -277,80 +277,6 @@ define <64 x i8> @testv64i8(<64 x i8> %in) nounwind {
   ret <64 x i8> %out
 }
 
-define <8 x i64> @ult_2_v8i64(<8 x i64> %0) {
-; AVX512F-LABEL: ult_2_v8i64:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
-; AVX512F-NEXT:    vptestnmq %zmm1, %zmm0, %k1
-; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ult_2_v8i64:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestnmq %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v8i64:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpcmpltuq {{.*}}(%rip){1to8}, %zmm0, %k1
-; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v8i64:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; BITALG-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
-; BITALG-NEXT:    vptestnmq %zmm1, %zmm0, %k1
-; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; BITALG-NEXT:    retq
-  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
-  %3 = icmp ult <8 x i64> %2, <i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2, i64 2>
-  %4 = sext <8 x i1> %3 to <8 x i64>
-  ret <8 x i64> %4
-}
-
-define <8 x i64> @ugt_1_v8i64(<8 x i64> %0) {
-; AVX512F-LABEL: ugt_1_v8i64:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
-; AVX512F-NEXT:    vptestmq %zmm1, %zmm0, %k1
-; AVX512F-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ugt_1_v8i64:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestmq %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v8i64:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpopcntq %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpcmpnleuq {{.*}}(%rip){1to8}, %zmm0, %k1
-; AVX512VPOPCNTDQ-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v8i64:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; BITALG-NEXT:    vpaddq %zmm1, %zmm0, %zmm1
-; BITALG-NEXT:    vptestmq %zmm1, %zmm0, %k1
-; BITALG-NEXT:    vpternlogq $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; BITALG-NEXT:    retq
-  %2 = tail call <8 x i64> @llvm.ctpop.v8i64(<8 x i64> %0)
-  %3 = icmp ugt <8 x i64> %2, <i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1, i64 1>
-  %4 = sext <8 x i1> %3 to <8 x i64>
-  ret <8 x i64> %4
-}
-
 define <8 x i64> @eq_1_v8i64(<8 x i64> %0) {
 ; AVX512F-LABEL: eq_1_v8i64:
 ; AVX512F:       # %bb.0:
@@ -434,80 +360,6 @@ define <8 x i64> @ne_1_v8i64(<8 x i64> %0) {
   ret <8 x i64> %4
 }
 
-define <16 x i32> @ult_2_v16i32(<16 x i32> %0) {
-; AVX512F-LABEL: ult_2_v16i32:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
-; AVX512F-NEXT:    vptestnmd %zmm1, %zmm0, %k1
-; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ult_2_v16i32:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestnmd %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ult_2_v16i32:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpcmpltud {{.*}}(%rip){1to16}, %zmm0, %k1
-; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v16i32:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; BITALG-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
-; BITALG-NEXT:    vptestnmd %zmm1, %zmm0, %k1
-; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; BITALG-NEXT:    retq
-  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
-  %3 = icmp ult <16 x i32> %2, <i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2, i32 2>
-  %4 = sext <16 x i1> %3 to <16 x i32>
-  ret <16 x i32> %4
-}
-
-define <16 x i32> @ugt_1_v16i32(<16 x i32> %0) {
-; AVX512F-LABEL: ugt_1_v16i32:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512F-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
-; AVX512F-NEXT:    vptestmd %zmm1, %zmm0, %k1
-; AVX512F-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ugt_1_v16i32:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestmd %zmm1, %zmm0, %k1
-; AVX512BW-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-LABEL: ugt_1_v16i32:
-; AVX512VPOPCNTDQ:       # %bb.0:
-; AVX512VPOPCNTDQ-NEXT:    vpopcntd %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NEXT:    vpcmpnleud {{.*}}(%rip){1to16}, %zmm0, %k1
-; AVX512VPOPCNTDQ-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; AVX512VPOPCNTDQ-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v16i32:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; BITALG-NEXT:    vpaddd %zmm1, %zmm0, %zmm1
-; BITALG-NEXT:    vptestmd %zmm1, %zmm0, %k1
-; BITALG-NEXT:    vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
-; BITALG-NEXT:    retq
-  %2 = tail call <16 x i32> @llvm.ctpop.v16i32(<16 x i32> %0)
-  %3 = icmp ugt <16 x i32> %2, <i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1, i32 1>
-  %4 = sext <16 x i1> %3 to <16 x i32>
-  ret <16 x i32> %4
-}
-
 define <16 x i32> @eq_1_v16i32(<16 x i32> %0) {
 ; AVX512F-LABEL: eq_1_v16i32:
 ; AVX512F:       # %bb.0:
@@ -591,122 +443,6 @@ define <16 x i32> @ne_1_v16i32(<16 x i32> %0) {
   ret <16 x i32> %4
 }
 
-define <32 x i16> @ult_2_v32i16(<32 x i16> %0) {
-; AVX512F-LABEL: ult_2_v32i16:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
-; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ult_2_v32i16:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestnmw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-NOBW-LABEL: ult_2_v32i16:
-; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-BW-LABEL: ult_2_v32i16:
-; AVX512VPOPCNTDQ-BW:       # %bb.0:
-; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vptestnmw %zmm1, %zmm0, %k0
-; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
-; AVX512VPOPCNTDQ-BW-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v32i16:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
-; BITALG-NEXT:    vpcmpltuw {{.*}}(%rip), %zmm0, %k0
-; BITALG-NEXT:    vpmovm2w %k0, %zmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
-  %3 = icmp ult <32 x i16> %2, <i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2, i16 2>
-  %4 = sext <32 x i1> %3 to <32 x i16>
-  ret <32 x i16> %4
-}
-
-define <32 x i16> @ugt_1_v32i16(<32 x i16> %0) {
-; AVX512F-LABEL: ugt_1_v32i16:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
-; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ugt_1_v32i16:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestmw %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    vpmovm2w %k0, %zmm0
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_1_v32i16:
-; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm1, %ymm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddw %ymm2, %ymm0, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqw %ymm3, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-BW-LABEL: ugt_1_v32i16:
-; AVX512VPOPCNTDQ-BW:       # %bb.0:
-; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vpaddw %zmm1, %zmm0, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vptestmw %zmm1, %zmm0, %k0
-; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2w %k0, %zmm0
-; AVX512VPOPCNTDQ-BW-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v32i16:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntw %zmm0, %zmm0
-; BITALG-NEXT:    vpcmpnleuw {{.*}}(%rip), %zmm0, %k0
-; BITALG-NEXT:    vpmovm2w %k0, %zmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <32 x i16> @llvm.ctpop.v32i16(<32 x i16> %0)
-  %3 = icmp ugt <32 x i16> %2, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1>
-  %4 = sext <32 x i1> %3 to <32 x i16>
-  ret <32 x i16> %4
-}
-
 define <32 x i16> @eq_1_v32i16(<32 x i16> %0) {
 ; AVX512F-LABEL: eq_1_v32i16:
 ; AVX512F:       # %bb.0:
@@ -845,122 +581,6 @@ define <32 x i16> @ne_1_v32i16(<32 x i16> %0) {
   ret <32 x i16> %4
 }
 
-define <64 x i8> @ult_2_v64i8(<64 x i8> %0) {
-; AVX512F-LABEL: ult_2_v64i8:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
-; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ult_2_v64i8:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestnmb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-NOBW-LABEL: ult_2_v64i8:
-; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-BW-LABEL: ult_2_v64i8:
-; AVX512VPOPCNTDQ-BW:       # %bb.0:
-; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vptestnmb %zmm1, %zmm0, %k0
-; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
-; AVX512VPOPCNTDQ-BW-NEXT:    retq
-;
-; BITALG-LABEL: ult_2_v64i8:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
-; BITALG-NEXT:    vpcmpltub {{.*}}(%rip), %zmm0, %k0
-; BITALG-NEXT:    vpmovm2b %k0, %zmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
-  %3 = icmp ult <64 x i8> %2, <i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2, i8 2>
-  %4 = sext <64 x i1> %3 to <64 x i8>
-  ret <64 x i8> %4
-}
-
-define <64 x i8> @ugt_1_v64i8(<64 x i8> %0) {
-; AVX512F-LABEL: ugt_1_v64i8:
-; AVX512F:       # %bb.0:
-; AVX512F-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512F-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
-; AVX512F-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
-; AVX512F-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
-; AVX512F-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512F-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
-; AVX512F-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512F-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512F-NEXT:    retq
-;
-; AVX512BW-LABEL: ugt_1_v64i8:
-; AVX512BW:       # %bb.0:
-; AVX512BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
-; AVX512BW-NEXT:    vptestmb %zmm1, %zmm0, %k0
-; AVX512BW-NEXT:    vpmovm2b %k0, %zmm0
-; AVX512BW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-NOBW-LABEL: ugt_1_v64i8:
-; AVX512VPOPCNTDQ-NOBW:       # %bb.0:
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vextracti64x4 $1, %zmm0, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqd %ymm2, %ymm2, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm1, %ymm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpxor %xmm3, %xmm3, %xmm3
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm1, %ymm1
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpaddb %ymm2, %ymm0, %ymm2
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpand %ymm2, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpcmpeqb %ymm3, %ymm0, %ymm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vinserti64x4 $1, %ymm1, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    vpternlogq $15, %zmm0, %zmm0, %zmm0
-; AVX512VPOPCNTDQ-NOBW-NEXT:    retq
-;
-; AVX512VPOPCNTDQ-BW-LABEL: ugt_1_v64i8:
-; AVX512VPOPCNTDQ-BW:       # %bb.0:
-; AVX512VPOPCNTDQ-BW-NEXT:    vpternlogd $255, %zmm1, %zmm1, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vpaddb %zmm1, %zmm0, %zmm1
-; AVX512VPOPCNTDQ-BW-NEXT:    vptestmb %zmm1, %zmm0, %k0
-; AVX512VPOPCNTDQ-BW-NEXT:    vpmovm2b %k0, %zmm0
-; AVX512VPOPCNTDQ-BW-NEXT:    retq
-;
-; BITALG-LABEL: ugt_1_v64i8:
-; BITALG:       # %bb.0:
-; BITALG-NEXT:    vpopcntb %zmm0, %zmm0
-; BITALG-NEXT:    vpcmpnleub {{.*}}(%rip), %zmm0, %k0
-; BITALG-NEXT:    vpmovm2b %k0, %zmm0
-; BITALG-NEXT:    retq
-  %2 = tail call <64 x i8> @llvm.ctpop.v64i8(<64 x i8> %0)
-  %3 = icmp ugt <64 x i8> %2, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1>
-  %4 = sext <64 x i1> %3 to <64 x i8>
-  ret <64 x i8> %4
-}
-
 define <64 x i8> @eq_1_v64i8(<64 x i8> %0) {
 ; AVX512F-LABEL: eq_1_v64i8:
 ; AVX512F:       # %bb.0:
diff --git a/llvm/test/DebugInfo/MIR/InstrRef/lit.local.cfg b/llvm/test/DebugInfo/MIR/InstrRef/lit.local.cfg
new file mode 100644
index 0000000000000..c8625f4d9d248
--- /dev/null
+++ b/llvm/test/DebugInfo/MIR/InstrRef/lit.local.cfg
@@ -0,0 +1,2 @@
+if not 'X86' in config.root.targets:
+    config.unsupported = True
diff --git a/llvm/test/DebugInfo/MIR/InstrRef/survives-livedebugvars.mir b/llvm/test/DebugInfo/MIR/InstrRef/survives-livedebugvars.mir
new file mode 100644
index 0000000000000..80ae2e78fea07
--- /dev/null
+++ b/llvm/test/DebugInfo/MIR/InstrRef/survives-livedebugvars.mir
@@ -0,0 +1,142 @@
+# RUN: llc -start-after=phi-node-elimination -stop-after=virtregrewriter %s -mtriple=x86_64-unknown-unknown -o - -experimental-debug-variable-locations | FileCheck %s
+# RUN: llc -O0 -start-after=phi-node-elimination -stop-after=regallocfast %s -mtriple=x86_64-unknown-unknown -o - -experimental-debug-variable-locations | FileCheck %s --check-prefix=FASTREG
+#
+# Test that DBG_INSTR_REFs can pass through livedebugvariables to the end of
+# regalloc without problem. Program body copied from
+# livedebugvars-crossbb-interval.mir.
+#
+# CHECK-LABEL: bb.0:
+# CHECK:       DBG_INSTR_REF 1, 0
+# CHECK-NEXT:  JMP_1
+# CHECK-LABEL: bb.1:
+# CHECK:       DBG_INSTR_REF 2, 0
+# CHECK-NEXT:  JMP_1
+# CHECK-LABEL: bb.2:
+# CHECK:       DBG_INSTR_REF 3, 0
+# CHECK-NEXT:  CALL64pcrel32
+# CHECK-LABEL: bb.3:
+# CHECK:       DBG_INSTR_REF 4, 0
+# CHECK-NEXT:  JMP_1
+#
+#
+# The fast register allocator puts some spills in -- these are no-ops as far
+# as the slot indexes are concerned. It doesn't matter which side of spills
+# the DBG_INSTR_REF lands on.
+#
+# FASTREG-LABEL: bb.0:
+# FASTREG-DAG:   DBG_INSTR_REF 1, 0
+# FASTREG-DAG:   MOV64mr
+# FASTREG-DAG:   MOV32mr
+# FASTREG-NEXT:  JMP_1
+# FASTREG-LABEL: bb.1:
+# FASTREG:       DBG_INSTR_REF 2, 0
+# FASTREG-NEXT:  JMP_1
+# FASTREG-LABEL: bb.2:
+# FASTREG:       DBG_INSTR_REF 3, 0
+# FASTREG-NEXT:  CALL64pcrel32
+# FASTREG-LABEL: bb.3:
+# FASTREG-DAG:   MOV32rm
+# FASTREG-DAG:   DBG_INSTR_REF 4, 0
+# FASTREG-DAG:   MOV32mr
+# FASTREG-NEXT:  JMP_1
+# FASTREG-LABEL: bb.4:
+# FASTREG:       DBG_INSTR_REF 5, 0
+# FASTREG-NEXT:  RETQ
+
+--- |
+  ; ModuleID = 'tmp.ll'
+  source_filename = "tmp.ll"
+  target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
+
+  %struct.a = type { i32 }
+
+  ; Function Attrs: nounwind ssp
+  define i32 @bar() !dbg !4 {
+    ret i32 0, !dbg !18
+  }
+
+  declare i32 @foo();
+
+  ; Function Attrs: nounwind readnone speculatable willreturn
+  declare void @llvm.dbg.value(metadata, metadata, metadata)
+
+  !llvm.dbg.cu = !{!0}
+  !llvm.module.flags = !{!3}
+  
+  !0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "asdf", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, retainedTypes: !2)
+  !1 = !DIFile(filename: "bar.c", directory: "asdf")
+  !2 = !{}
+  !3 = !{i32 1, !"Debug Info Version", i32 3}
+  !4 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 5, type: !5, virtualIndex: 6, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !8)
+  !5 = !DISubroutineType(types: !6)
+  !6 = !{!7}
+  !7 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
+  !8 = !{!9, !14}
+  !9 = !DILocalVariable(name: "b", arg: 1, scope: !4, file: !1, line: 5, type: !10)
+  !10 = !DIDerivedType(tag: DW_TAG_pointer_type, scope: !0, baseType: !11, size: 64, align: 64)
+  !11 = !DICompositeType(tag: DW_TAG_structure_type, name: "a", scope: !0, file: !1, line: 1, size: 32, align: 32, elements: !12)
+  !12 = !{!13}
+  !13 = !DIDerivedType(tag: DW_TAG_member, name: "c", scope: !1, file: !1, line: 2, baseType: !7, size: 32, align: 32)
+  !14 = !DILocalVariable(name: "x", scope: !15, file: !1, line: 6, type: !7)
+  !15 = distinct !DILexicalBlock(scope: !4, file: !1, line: 5, column: 22)
+  !16 = !DILocation(line: 5, column: 19, scope: !4)
+  !17 = !DILocation(line: 6, column: 14, scope: !15)
+  !18 = !DILocation(line: 8, column: 2, scope: !15)
+  !19 = !DILocation(line: 7, column: 2, scope: !15)
+
+...
+---
+name:            bar
+alignment:       16
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: gr64 }
+  - { id: 1, class: gr32 }
+  - { id: 2, class: gr64 }
+  - { id: 3, class: gr64 }
+  - { id: 4, class: gr32 }
+  - { id: 5, class: gr32 }
+  - { id: 6, class: gr32 }
+  - { id: 7, class: gr32 }
+liveins:
+  - { reg: '$rdi', virtual-reg: '%2' }
+  - { reg: '$esi', virtual-reg: '%4' }
+frameInfo:
+  hasCalls:        true
+machineFunctionInfo: {}
+body:             |
+  bb.0:
+    liveins: $rdi, $esi
+
+    %4:gr32 = COPY $esi
+    %2:gr64 = COPY $rdi
+    %3:gr64 = COPY killed %2
+    %5:gr32 = COPY killed %4
+    DBG_INSTR_REF 1, 0, !9, !DIExpression(), debug-location !16
+    JMP_1 %bb.3
+
+  bb.1:
+    DBG_INSTR_REF 2, 0, !9, !DIExpression(), debug-location !16
+    JMP_1 %bb.4
+
+  bb.2:
+    ADJCALLSTACKDOWN64 0, 0, 0, implicit-def $rsp, implicit-def $eflags, implicit-def $ssp, implicit $rsp, implicit $ssp, debug-location !19
+    $edi = COPY %6, debug-location !19
+    $al = MOV8ri 0, debug-location !19
+    DBG_INSTR_REF 3, 0, !9, !DIExpression(), debug-location !16
+    CALL64pcrel32 @foo, csr_64, implicit $rsp, implicit $ssp, implicit $al, implicit $edi, implicit-def $eax, debug-location !19
+    ADJCALLSTACKUP64 0, 0, implicit-def $rsp, implicit-def $eflags, implicit-def $ssp, implicit $rsp, implicit $ssp, debug-location !19
+    %7:gr32 = COPY $eax, debug-location !19
+    JMP_1 %bb.1
+
+  bb.3:
+    %6:gr32 = MOV32rm %3, 1, $noreg, 0, $noreg, debug-location !17
+    DBG_INSTR_REF 4, 0, !9, !DIExpression(), debug-location !16
+    JMP_1 %bb.2
+
+  bb.4:
+    $eax = COPY %5, debug-location !18
+    DBG_INSTR_REF 5, 0, !9, !DIExpression(), debug-location !16
+    RETQ implicit $eax, debug-location !18
+
+...
diff --git a/llvm/test/DebugInfo/MIR/InstrRef/twoaddr-to-threeaddr-sub.mir b/llvm/test/DebugInfo/MIR/InstrRef/twoaddr-to-threeaddr-sub.mir
new file mode 100644
index 0000000000000..55a8d6be2f0f9
--- /dev/null
+++ b/llvm/test/DebugInfo/MIR/InstrRef/twoaddr-to-threeaddr-sub.mir
@@ -0,0 +1,42 @@
+# RUN: llc -run-pass=twoaddressinstruction -mtriple=x86_64-- -o - %s -experimental-debug-variable-locations | FileCheck %s
+#
+# REQUIRES: x86-registered-target
+#
+# Test that a new instruction (LEA) is created when the two-addr add below is
+# converted to three address; and that an appropriate substitution is created.
+# Maybe at some point we'll normalise DBG_INSTR_REFs on output, but until then,
+# lets not.
+#
+# CHECK:      debugValueSubstitutions:
+# CHECK-NEXT:  - { srcinst: 1, srcop: 0, dstinst: 2, dstop: 0 }
+#
+# CHECK:      LEA64_32r
+# CHECK-SAME: debug-instr-number 2
+#
+# CHECK:      DBG_INSTR_REF 1, 0
+---
+name:            test1
+alignment:       16
+tracksRegLiveness: true
+registers:
+  - { id: 0, class: gr32 }
+  - { id: 1, class: gr32 }
+  - { id: 2, class: gr32 }
+liveins:
+  - { reg: '$edi', virtual-reg: '%0' }
+frameInfo:
+  maxAlignment:    1
+machineFunctionInfo: {}
+body:             |
+  bb.0:
+    liveins: $edi
+
+    %0:gr32 = COPY killed $edi
+    %1:gr32 = SHL32ri killed %0, 5, implicit-def dead $eflags
+    %2:gr32 = ADD32ri8_DB killed %1, 3, implicit-def dead $eflags, debug-instr-number 1
+    DBG_INSTR_REF 1, 0
+    $eax = COPY killed %2
+    RET 0, killed $eax
+
+...
+
diff --git a/llvm/test/DebugInfo/symbolize-macho-universal-unknown-arch.test b/llvm/test/DebugInfo/symbolize-macho-universal-unknown-arch.test
index b6b1ee77b8b7a..744a0d3e5492a 100644
--- a/llvm/test/DebugInfo/symbolize-macho-universal-unknown-arch.test
+++ b/llvm/test/DebugInfo/symbolize-macho-universal-unknown-arch.test
@@ -1,5 +1,5 @@
-RUN: not llvm-symbolizer --obj=%p/Inputs/macho-universal 0x1f84 2>&1 | FileCheck --implicit-check-not=main %s
-
-CHECK: LLVMSymbolizer: error reading file: Unknown architecture named:
+RUN: llvm-symbolizer --obj=%p/Inputs/macho-universal 0x1f84 | FileCheck %s
 
+CHECK-NOT: main
 CHECK: ??
+CHECK-NOT: main
diff --git a/llvm/test/DebugInfo/symbolize-missing-file.test b/llvm/test/DebugInfo/symbolize-missing-file.test
index bf148517f3c48..2762af4ff5dad 100644
--- a/llvm/test/DebugInfo/symbolize-missing-file.test
+++ b/llvm/test/DebugInfo/symbolize-missing-file.test
@@ -1,3 +1,3 @@
-RUN: not llvm-symbolizer --obj=unexisting-file 0x1234 2>&1 | FileCheck %s
+RUN: llvm-symbolizer --obj=unexisting-file 0x1234 2>&1 | FileCheck %s
 
 CHECK: LLVMSymbolizer: error reading file: {{[Nn]}}o such file or directory
diff --git a/llvm/test/Instrumentation/AddressSanitizer/alloca-offset-lifetime.ll b/llvm/test/Instrumentation/AddressSanitizer/alloca-offset-lifetime.ll
index d4bacbcae4d97..1a44f4585b7a7 100644
--- a/llvm/test/Instrumentation/AddressSanitizer/alloca-offset-lifetime.ll
+++ b/llvm/test/Instrumentation/AddressSanitizer/alloca-offset-lifetime.ll
@@ -1,6 +1,7 @@
 ; Test that ASAN will not instrument lifetime markers on alloca offsets.
 ;
-; RUN: opt < %s --asan --asan-use-after-scope -S | FileCheck %s
+; RUN: opt < %s --asan --asan-use-after-scope -S -enable-new-pm=0 | FileCheck %s
+; RUN: opt < %s -passes=asan-function-pipeline --asan-use-after-scope -S | FileCheck %s
 
 target datalayout = "e-m:o-p270:32:32-p271:32:32-p272:64:64-i64:64-f80:128-n8:16:32:64-S128"
 target triple = "x86_64-apple-macosx10.15.0"
diff --git a/llvm/test/MC/Disassembler/X86/x86-64.txt b/llvm/test/MC/Disassembler/X86/x86-64.txt
index b56d78297eea2..23c5a6ea59b26 100644
--- a/llvm/test/MC/Disassembler/X86/x86-64.txt
+++ b/llvm/test/MC/Disassembler/X86/x86-64.txt
@@ -715,3 +715,27 @@
 
 # CHECK: hreset $1
 0xf3 0x0f 0x3a 0xf0 0xc0 0x01
+
+# CHECK: uiret
+0xf3,0x0f,0x01,0xec
+
+# CHECK: clui
+0xf3,0x0f,0x01,0xee
+
+# CHECK: stui
+0xf3,0x0f,0x01,0xef
+
+# CHECK: testui
+0xf3,0x0f,0x01,0xed
+
+# CHECK: senduipi %rax
+0xf3,0x0f,0xc7,0xf0
+
+# CHECK: senduipi %rdx
+0xf3,0x0f,0xc7,0xf2
+
+# CHECK: senduipi %r8
+0xf3,0x41,0x0f,0xc7,0xf0
+
+# CHECK: senduipi %r13
+0xf3,0x41,0x0f,0xc7,0xf5
diff --git a/llvm/test/MC/X86/encoder-fail.s b/llvm/test/MC/X86/encoder-fail.s
index d8d321fa8a1a7..a8b9f48c8fb70 100644
--- a/llvm/test/MC/X86/encoder-fail.s
+++ b/llvm/test/MC/X86/encoder-fail.s
@@ -1,3 +1,16 @@
-// RUN: not --crash llvm-mc -triple x86_64-unknown-unknown --show-encoding %s 2>&1 | FileCheck %s
-// CHECK: LLVM ERROR: Cannot encode high byte register in REX-prefixed instruction
- movzx %dh, %rsi
+// RUN: not llvm-mc -triple x86_64-unknown-unknown --show-encoding %s 2>&1 | FileCheck %s
+
+// CHECK: error: can't encode 'dh' in an instruction requiring REX prefix
+movzx %dh, %rsi
+
+// CHECK: error: can't encode 'ah' in an instruction requiring REX prefix
+movzx %ah, %r8d
+
+// CHECK: error: can't encode 'bh' in an instruction requiring REX prefix
+add %bh, %sil
+
+// CHECK: error: can't encode 'ch' in an instruction requiring REX prefix
+mov %ch, (%r8)
+
+// CHECK: error: can't encode 'dh' in an instruction requiring REX prefix
+mov %dh, (%rax,%r8)
diff --git a/llvm/test/MC/X86/x86-64.s b/llvm/test/MC/X86/x86-64.s
index 74777dffab235..77c45d3f80e83 100644
--- a/llvm/test/MC/X86/x86-64.s
+++ b/llvm/test/MC/X86/x86-64.s
@@ -2018,3 +2018,35 @@ tdcall
 // CHECK: hreset
 // CHECK: encoding: [0xf3,0x0f,0x3a,0xf0,0xc0,0x01]
 hreset $1
+
+// CHECK: uiret
+// CHECK: encoding: [0xf3,0x0f,0x01,0xec]
+uiret
+
+// CHECK: clui
+// CHECK: encoding: [0xf3,0x0f,0x01,0xee]
+clui
+
+// CHECK: stui
+// CHECK: encoding: [0xf3,0x0f,0x01,0xef]
+stui
+
+// CHECK: testui
+// CHECK: encoding: [0xf3,0x0f,0x01,0xed]
+testui
+
+// CHECK: senduipi %rax
+// CHECK: encoding: [0xf3,0x0f,0xc7,0xf0]
+senduipi %rax
+
+// CHECK: senduipi %rdx
+// CHECK: encoding: [0xf3,0x0f,0xc7,0xf2]
+senduipi %rdx
+
+// CHECK: senduipi %r8
+// CHECK: encoding: [0xf3,0x41,0x0f,0xc7,0xf0]
+senduipi %r8
+
+// CHECK: senduipi %r13
+// CHECK: encoding: [0xf3,0x41,0x0f,0xc7,0xf5]
+senduipi %r13
diff --git a/llvm/test/Other/2002-01-31-CallGraph.ll b/llvm/test/Other/2002-01-31-CallGraph.ll
index d4819357ac67e..52d88688987d3 100644
--- a/llvm/test/Other/2002-01-31-CallGraph.ll
+++ b/llvm/test/Other/2002-01-31-CallGraph.ll
@@ -1,6 +1,7 @@
 ;  Call graph construction crash: Not handling indirect calls right
 ;
-; RUN: opt < %s -analyze -print-callgraph > /dev/null 2>&1
+; RUN: opt < %s -analyze -print-callgraph -enable-new-pm=0 > /dev/null 2>&1
+; RUN: opt < %s -passes=print-callgraph > /dev/null 2>&1
 ;
 
         %FunTy = type i32 (i32)
diff --git a/llvm/test/Other/2007-06-05-PassID.ll b/llvm/test/Other/2007-06-05-PassID.ll
index 6964f2824d68b..b166f0575d5cb 100644
--- a/llvm/test/Other/2007-06-05-PassID.ll
+++ b/llvm/test/Other/2007-06-05-PassID.ll
@@ -1,13 +1,13 @@
-;RUN: opt < %s -analyze -dot-cfg-only 2>/dev/null
-;RUN: opt < %s -analyze -passes=dot-cfg-only 2>/dev/null
+;RUN: opt < %s -analyze -dot-cfg-only -enable-new-pm=0 2>/dev/null
+;RUN: opt < %s -passes=dot-cfg-only 2>/dev/null
 ;RUN: opt < %s -analyze -dot-cfg-only \
-;RUN:          -cfg-heat-colors=true -cfg-weights=true 2>/dev/null
+;RUN:          -cfg-heat-colors=true -cfg-weights=true -enable-new-pm=0 2>/dev/null
 ;RUN: opt < %s -analyze -dot-cfg-only \
-;RUN:          -cfg-heat-colors=false -cfg-weights=false 2>/dev/null
+;RUN:          -cfg-heat-colors=false -cfg-weights=false -enable-new-pm=0 2>/dev/null
 ;RUN: opt < %s -analyze -dot-cfg \
-;RUN:          -cfg-heat-colors=true -cfg-weights=true 2>/dev/null
+;RUN:          -cfg-heat-colors=true -cfg-weights=true -enable-new-pm=0 2>/dev/null
 ;RUN: opt < %s -analyze -dot-cfg \
-;RUN:          -cfg-heat-colors=false -cfg-weights=false 2>/dev/null
+;RUN:          -cfg-heat-colors=false -cfg-weights=false -enable-new-pm=0 2>/dev/null
 ;PR 1497
 
 define void @foo() {
diff --git a/llvm/test/Other/2007-06-28-PassManager.ll b/llvm/test/Other/2007-06-28-PassManager.ll
index 0ed275948dba3..c6a69a210bc2b 100644
--- a/llvm/test/Other/2007-06-28-PassManager.ll
+++ b/llvm/test/Other/2007-06-28-PassManager.ll
@@ -1,6 +1,6 @@
-; RUN: opt < %s -analyze -inline
+; RUN: opt < %s -analyze -inline -enable-new-pm=0
 ; PR1526
-; RUN: opt < %s -analyze -indvars
+; RUN: opt < %s -analyze -indvars -enable-new-pm=0
 ; PR1539
 define i32 @test1() {
        ret i32 0
diff --git a/llvm/test/Other/cfg-printer-branch-weights-percent.ll b/llvm/test/Other/cfg-printer-branch-weights-percent.ll
index bfbb73160444f..d1a647d4fc434 100644
--- a/llvm/test/Other/cfg-printer-branch-weights-percent.ll
+++ b/llvm/test/Other/cfg-printer-branch-weights-percent.ll
@@ -1,4 +1,4 @@
-;RUN: opt < %s -analyze -dot-cfg -cfg-weights -cfg-dot-filename-prefix=%t 2>/dev/null
+;RUN: opt < %s -dot-cfg -cfg-weights -cfg-dot-filename-prefix=%t 2>/dev/null
 ;RUN: FileCheck %s -input-file=%t.f.dot
 
 define void @f(i32) {
diff --git a/llvm/test/Other/cfg-printer-branch-weights.ll b/llvm/test/Other/cfg-printer-branch-weights.ll
index 96bd2120ef1bc..63616b699dd6e 100644
--- a/llvm/test/Other/cfg-printer-branch-weights.ll
+++ b/llvm/test/Other/cfg-printer-branch-weights.ll
@@ -1,4 +1,4 @@
-;RUN: opt < %s -analyze -dot-cfg -cfg-weights -cfg-raw-weights -cfg-dot-filename-prefix=%t 2>/dev/null
+;RUN: opt < %s -dot-cfg -cfg-weights -cfg-raw-weights -cfg-dot-filename-prefix=%t 2>/dev/null
 ;RUN: FileCheck %s -input-file=%t.f.dot
 
 ;TODO: Investigate why this test doesn't work with dynamically linked libraries
diff --git a/llvm/test/Other/cfg_deopt_unreach.ll b/llvm/test/Other/cfg_deopt_unreach.ll
index 826e4abb42bb4..ec6f872c6054c 100644
--- a/llvm/test/Other/cfg_deopt_unreach.ll
+++ b/llvm/test/Other/cfg_deopt_unreach.ll
@@ -1,12 +1,12 @@
 ; RUN: rm -rf %t
 ; RUN: mkdir -p %t
-; RUN: opt < %s -analyze -dot-cfg -cfg-hide-unreachable-paths -cfg-dot-filename-prefix=%t/unreach 2>/dev/null
+; RUN: opt < %s -dot-cfg -cfg-hide-unreachable-paths -cfg-dot-filename-prefix=%t/unreach 2>/dev/null
 ; RUN: FileCheck %s -input-file=%t/unreach.callee.dot -check-prefix=UNREACH
-; RUN: opt < %s -analyze -dot-cfg -cfg-hide-deoptimize-paths -cfg-dot-filename-prefix=%t/deopt 2>/dev/null
+; RUN: opt < %s -dot-cfg -cfg-hide-deoptimize-paths -cfg-dot-filename-prefix=%t/deopt 2>/dev/null
 ; RUN: FileCheck %s -input-file=%t/deopt.callee.dot -check-prefix=DEOPT
-; RUN: opt < %s -analyze -dot-cfg -cfg-dot-filename-prefix=%t/no-flags 2>/dev/null
+; RUN: opt < %s -dot-cfg -cfg-dot-filename-prefix=%t/no-flags 2>/dev/null
 ; RUN: FileCheck %s -input-file=%t/no-flags.callee.dot -check-prefix=NO-FLAGS
-; RUN: opt < %s -analyze -dot-cfg -cfg-hide-unreachable-paths -cfg-hide-deoptimize-paths -cfg-dot-filename-prefix=%t/both-flags 2>/dev/null
+; RUN: opt < %s -dot-cfg -cfg-hide-unreachable-paths -cfg-hide-deoptimize-paths -cfg-dot-filename-prefix=%t/both-flags 2>/dev/null
 ; RUN: FileCheck %s -input-file=%t/both-flags.callee.dot -check-prefix=BOTH-FLAGS
 
 ; TODO: Investigate why this test doesn't work with dynamically linked libraries
diff --git a/llvm/test/Other/debugcounter-predicateinfo.ll b/llvm/test/Other/debugcounter-predicateinfo.ll
index 6c82148f9e399..bbc7a0f712711 100644
--- a/llvm/test/Other/debugcounter-predicateinfo.ll
+++ b/llvm/test/Other/debugcounter-predicateinfo.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; REQUIRES: asserts
-; RUN: opt -debug-counter=predicateinfo-rename-skip=1,predicateinfo-rename-count=1 -print-predicateinfo -analyze  < %s 2>&1 | FileCheck %s
+; RUN: opt -debug-counter=predicateinfo-rename-skip=1,predicateinfo-rename-count=1 -print-predicateinfo < %s 2>&1 | FileCheck %s
 ;; Test that, with debug counters on, we don't rename the first info, only the second
 define fastcc void @barney() {
 ; CHECK-LABEL: @barney(
diff --git a/llvm/test/Other/heat-colors-graphs.ll b/llvm/test/Other/heat-colors-graphs.ll
index 3af384fbcf5db..4cc58a640cd5f 100644
--- a/llvm/test/Other/heat-colors-graphs.ll
+++ b/llvm/test/Other/heat-colors-graphs.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s -analyze -dot-cfg -cfg-heat-colors -cfg-dot-filename-prefix=%t 2>/dev/null
+; RUN: opt < %s -dot-cfg -cfg-heat-colors -cfg-dot-filename-prefix=%t 2>/dev/null
 ; RUN: FileCheck %s -input-file=%t.f.dot --check-prefixes=CHECK-CFG,CHECK-BOTH
 ; RUN: opt %s -dot-callgraph -callgraph-heat-colors -callgraph-dot-filename-prefix=%t 2>/dev/null
 ; RUN: FileCheck %s -input-file=%t.callgraph.dot --check-prefix=CHECK-BOTH
diff --git a/llvm/test/Other/new-pm-defaults.ll b/llvm/test/Other/new-pm-defaults.ll
index 2576f939c28ea..af1537581713c 100644
--- a/llvm/test/Other/new-pm-defaults.ll
+++ b/llvm/test/Other/new-pm-defaults.ll
@@ -202,16 +202,15 @@
 ; CHECK-O23SZ-NEXT: Running analysis: LazyValueAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: CorrelatedValuePropagationPass
 ; CHECK-O23SZ-NEXT: Invalidating analysis: LazyValueAnalysis
+; CHECK-O-NEXT: Running pass: ADCEPass
+; CHECK-O-NEXT: Running analysis: PostDominatorTreeAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: DSEPass
-; CHECK-O23SZ-NEXT: Running analysis: PostDominatorTreeAnalysis
 ; CHECK-O23SZ-NEXT: Starting llvm::Function pass manager run.
 ; CHECK-O23SZ-NEXT: Running pass: LoopSimplifyPass
 ; CHECK-O23SZ-NEXT: Running pass: LCSSAPass
 ; CHECK-O23SZ-NEXT: Finished llvm::Function pass manager run.
 ; CHECK-O23SZ-NEXT: Running pass: LICMPass
 ; CHECK-EP-SCALAR-LATE-NEXT: Running pass: NoOpFunctionPass
-; CHECK-O-NEXT: Running pass: ADCEPass
-; CHECK-O1-NEXT: Running analysis: PostDominatorTreeAnalysis
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-EP-PEEPHOLE-NEXT: Running pass: NoOpFunctionPass
diff --git a/llvm/test/Other/new-pm-thinlto-defaults.ll b/llvm/test/Other/new-pm-thinlto-defaults.ll
index bc1e6577f73a6..ad16091280cfc 100644
--- a/llvm/test/Other/new-pm-thinlto-defaults.ll
+++ b/llvm/test/Other/new-pm-thinlto-defaults.ll
@@ -175,15 +175,14 @@
 ; CHECK-O23SZ-NEXT: Running analysis: LazyValueAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: CorrelatedValuePropagationPass
 ; CHECK-O23SZ-NEXT: Invalidating analysis: LazyValueAnalysis
+; CHECK-O-NEXT: Running pass: ADCEPass
+; CHECK-O-NEXT: Running analysis: PostDominatorTreeAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: DSEPass
-; CHECK-O23SZ-NEXT: Running analysis: PostDominatorTreeAnalysis on foo
 ; CHECK-O23SZ-NEXT: Starting llvm::Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LoopSimplifyPass
 ; CHECK-O23SZ-NEXT: Running pass: LCSSAPass
 ; CHECK-O23SZ-NEXT: Finished llvm::Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LICMPass on Loop at depth 1 containing: %loop
-; CHECK-O-NEXT: Running pass: ADCEPass
-; CHECK-O1-NEXT: Running analysis: PostDominatorTreeAnalysis
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O-NEXT: Finished llvm::Function pass manager run.
diff --git a/llvm/test/Other/new-pm-thinlto-postlink-pgo-defaults.ll b/llvm/test/Other/new-pm-thinlto-postlink-pgo-defaults.ll
index 0e287cc156b8b..84a51d8daa096 100644
--- a/llvm/test/Other/new-pm-thinlto-postlink-pgo-defaults.ll
+++ b/llvm/test/Other/new-pm-thinlto-postlink-pgo-defaults.ll
@@ -148,13 +148,13 @@
 ; CHECK-O23SZ-NEXT: Running analysis: LazyValueAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: CorrelatedValuePropagationPass
 ; CHECK-O23SZ-NEXT: Invalidating analysis: LazyValueAnalysis
+; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O23SZ-NEXT: Running pass: DSEPass
 ; CHECK-O23SZ-NEXT: Starting {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LoopSimplifyPass
 ; CHECK-O23SZ-NEXT: Running pass: LCSSAPass
 ; CHECK-O23SZ-NEXT: Finished {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LICMPass
-; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O-NEXT: Finished {{.*}}Function pass manager run.
diff --git a/llvm/test/Other/new-pm-thinlto-postlink-samplepgo-defaults.ll b/llvm/test/Other/new-pm-thinlto-postlink-samplepgo-defaults.ll
index da63b9580d767..4616fc39d15c4 100644
--- a/llvm/test/Other/new-pm-thinlto-postlink-samplepgo-defaults.ll
+++ b/llvm/test/Other/new-pm-thinlto-postlink-samplepgo-defaults.ll
@@ -156,13 +156,13 @@
 ; CHECK-O23SZ-NEXT: Running analysis: LazyValueAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: CorrelatedValuePropagationPass
 ; CHECK-O23SZ-NEXT: Invalidating analysis: LazyValueAnalysis
+; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O23SZ-NEXT: Running pass: DSEPass
 ; CHECK-O23SZ-NEXT: Starting {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LoopSimplifyPass
 ; CHECK-O23SZ-NEXT: Running pass: LCSSAPass
 ; CHECK-O23SZ-NEXT: Finished {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LICMPass
-; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O3-NEXT: Running pass: ControlHeightReductionPass on foo
diff --git a/llvm/test/Other/new-pm-thinlto-prelink-pgo-defaults.ll b/llvm/test/Other/new-pm-thinlto-prelink-pgo-defaults.ll
index 11bd207781d86..ffcc6ecc789ad 100644
--- a/llvm/test/Other/new-pm-thinlto-prelink-pgo-defaults.ll
+++ b/llvm/test/Other/new-pm-thinlto-prelink-pgo-defaults.ll
@@ -198,13 +198,13 @@
 ; CHECK-O23SZ-NEXT: Running analysis: LazyValueAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: CorrelatedValuePropagationPass
 ; CHECK-O23SZ-NEXT: Invalidating analysis: LazyValueAnalysis
+; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O23SZ-NEXT: Running pass: DSEPass
 ; CHECK-O23SZ-NEXT: Starting {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LoopSimplifyPass
 ; CHECK-O23SZ-NEXT: Running pass: LCSSAPass
 ; CHECK-O23SZ-NEXT: Finished {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LICMPass
-; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O3-NEXT: Running pass: ControlHeightReductionPass on foo
diff --git a/llvm/test/Other/new-pm-thinlto-prelink-samplepgo-defaults.ll b/llvm/test/Other/new-pm-thinlto-prelink-samplepgo-defaults.ll
index 19a1fd551bf17..1bb1bc0c4aa1d 100644
--- a/llvm/test/Other/new-pm-thinlto-prelink-samplepgo-defaults.ll
+++ b/llvm/test/Other/new-pm-thinlto-prelink-samplepgo-defaults.ll
@@ -154,13 +154,13 @@
 ; CHECK-O23SZ-NEXT: Running analysis: LazyValueAnalysis
 ; CHECK-O23SZ-NEXT: Running pass: CorrelatedValuePropagationPass
 ; CHECK-O23SZ-NEXT: Invalidating analysis: LazyValueAnalysis
+; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O23SZ-NEXT: Running pass: DSEPass
 ; CHECK-O23SZ-NEXT: Starting {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LoopSimplifyPass
 ; CHECK-O23SZ-NEXT: Running pass: LCSSAPass
 ; CHECK-O23SZ-NEXT: Finished {{.*}}Function pass manager run
 ; CHECK-O23SZ-NEXT: Running pass: LICMPass
-; CHECK-O-NEXT: Running pass: ADCEPass
 ; CHECK-O-NEXT: Running pass: SimplifyCFGPass
 ; CHECK-O-NEXT: Running pass: InstCombinePass
 ; CHECK-O3-NEXT: Running pass: ControlHeightReductionPass on foo
diff --git a/llvm/test/Other/opt-O2-pipeline.ll b/llvm/test/Other/opt-O2-pipeline.ll
index 4d944208df7a8..e5f43715530da 100644
--- a/llvm/test/Other/opt-O2-pipeline.ll
+++ b/llvm/test/Other/opt-O2-pipeline.ll
@@ -160,9 +160,10 @@
 ; CHECK-NEXT:         Lazy Value Information Analysis
 ; CHECK-NEXT:         Jump Threading
 ; CHECK-NEXT:         Value Propagation
+; CHECK-NEXT:         Post-Dominator Tree Construction
+; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
 ; CHECK-NEXT:         Function Alias Analysis Results
-; CHECK-NEXT:         Post-Dominator Tree Construction
 ; CHECK-NEXT:         Memory SSA
 ; CHECK-NEXT:         Dead Store Elimination
 ; CHECK-NEXT:         Natural Loop Information
@@ -175,8 +176,6 @@
 ; CHECK-NEXT:         Lazy Block Frequency Analysis
 ; CHECK-NEXT:         Loop Pass Manager
 ; CHECK-NEXT:           Loop Invariant Code Motion
-; CHECK-NEXT:         Post-Dominator Tree Construction
-; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Simplify the CFG
 ; CHECK-NEXT:         Dominator Tree Construction
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
diff --git a/llvm/test/Other/opt-O3-pipeline-enable-matrix.ll b/llvm/test/Other/opt-O3-pipeline-enable-matrix.ll
index 5c3f2a57e7e4b..1f3d763be601e 100644
--- a/llvm/test/Other/opt-O3-pipeline-enable-matrix.ll
+++ b/llvm/test/Other/opt-O3-pipeline-enable-matrix.ll
@@ -165,9 +165,10 @@
 ; CHECK-NEXT:         Lazy Value Information Analysis
 ; CHECK-NEXT:         Jump Threading
 ; CHECK-NEXT:         Value Propagation
+; CHECK-NEXT:         Post-Dominator Tree Construction
+; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
 ; CHECK-NEXT:         Function Alias Analysis Results
-; CHECK-NEXT:         Post-Dominator Tree Construction
 ; CHECK-NEXT:         Memory SSA
 ; CHECK-NEXT:         Dead Store Elimination
 ; CHECK-NEXT:         Natural Loop Information
@@ -180,8 +181,6 @@
 ; CHECK-NEXT:         Lazy Block Frequency Analysis
 ; CHECK-NEXT:         Loop Pass Manager
 ; CHECK-NEXT:           Loop Invariant Code Motion
-; CHECK-NEXT:         Post-Dominator Tree Construction
-; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Simplify the CFG
 ; CHECK-NEXT:         Dominator Tree Construction
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
diff --git a/llvm/test/Other/opt-O3-pipeline.ll b/llvm/test/Other/opt-O3-pipeline.ll
index d910ad52ade6b..99f401973699b 100644
--- a/llvm/test/Other/opt-O3-pipeline.ll
+++ b/llvm/test/Other/opt-O3-pipeline.ll
@@ -165,9 +165,10 @@
 ; CHECK-NEXT:         Lazy Value Information Analysis
 ; CHECK-NEXT:         Jump Threading
 ; CHECK-NEXT:         Value Propagation
+; CHECK-NEXT:         Post-Dominator Tree Construction
+; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
 ; CHECK-NEXT:         Function Alias Analysis Results
-; CHECK-NEXT:         Post-Dominator Tree Construction
 ; CHECK-NEXT:         Memory SSA
 ; CHECK-NEXT:         Dead Store Elimination
 ; CHECK-NEXT:         Natural Loop Information
@@ -180,8 +181,6 @@
 ; CHECK-NEXT:         Lazy Block Frequency Analysis
 ; CHECK-NEXT:         Loop Pass Manager
 ; CHECK-NEXT:           Loop Invariant Code Motion
-; CHECK-NEXT:         Post-Dominator Tree Construction
-; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Simplify the CFG
 ; CHECK-NEXT:         Dominator Tree Construction
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
diff --git a/llvm/test/Other/opt-Os-pipeline.ll b/llvm/test/Other/opt-Os-pipeline.ll
index d89f9071470d4..a38062ac5a5e5 100644
--- a/llvm/test/Other/opt-Os-pipeline.ll
+++ b/llvm/test/Other/opt-Os-pipeline.ll
@@ -146,9 +146,10 @@
 ; CHECK-NEXT:         Lazy Value Information Analysis
 ; CHECK-NEXT:         Jump Threading
 ; CHECK-NEXT:         Value Propagation
+; CHECK-NEXT:         Post-Dominator Tree Construction
+; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
 ; CHECK-NEXT:         Function Alias Analysis Results
-; CHECK-NEXT:         Post-Dominator Tree Construction
 ; CHECK-NEXT:         Memory SSA
 ; CHECK-NEXT:         Dead Store Elimination
 ; CHECK-NEXT:         Natural Loop Information
@@ -161,8 +162,6 @@
 ; CHECK-NEXT:         Lazy Block Frequency Analysis
 ; CHECK-NEXT:         Loop Pass Manager
 ; CHECK-NEXT:           Loop Invariant Code Motion
-; CHECK-NEXT:         Post-Dominator Tree Construction
-; CHECK-NEXT:         Aggressive Dead Code Elimination
 ; CHECK-NEXT:         Simplify the CFG
 ; CHECK-NEXT:         Dominator Tree Construction
 ; CHECK-NEXT:         Basic Alias Analysis (stateless AA impl)
diff --git a/llvm/test/Transforms/BlockExtractor/extract-blocks.ll b/llvm/test/Transforms/BlockExtractor/extract-blocks.ll
index e720953a1e7a6..beee6356decdf 100644
--- a/llvm/test/Transforms/BlockExtractor/extract-blocks.ll
+++ b/llvm/test/Transforms/BlockExtractor/extract-blocks.ll
@@ -2,6 +2,8 @@
 ; RUN: echo 'foo bb20' >> %t
 ; RUN: opt -S -extract-blocks -extract-blocks-file=%t %s | FileCheck %s --check-prefix=CHECK-NO-ERASE
 ; RUN: opt -S -extract-blocks -extract-blocks-file=%t -extract-blocks-erase-funcs %s | FileCheck %s --check-prefix=CHECK-ERASE
+; RUN: opt -S -passes=extract-blocks -extract-blocks-file=%t %s | FileCheck %s --check-prefix=CHECK-NO-ERASE
+; RUN: opt -S -passes=extract-blocks -extract-blocks-file=%t -extract-blocks-erase-funcs %s | FileCheck %s --check-prefix=CHECK-ERASE
 
 ; CHECK-NO-ERASE: @foo(
 ; CHECK-NO-ERASE: @foo.bb9(
diff --git a/llvm/test/Transforms/IndVarSimplify/X86/eliminate-trunc.ll b/llvm/test/Transforms/IndVarSimplify/X86/eliminate-trunc.ll
index 4738369c0aca9..9fb7977c207f5 100644
--- a/llvm/test/Transforms/IndVarSimplify/X86/eliminate-trunc.ll
+++ b/llvm/test/Transforms/IndVarSimplify/X86/eliminate-trunc.ll
@@ -474,7 +474,7 @@ define void @test_10(i32 %n) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = zext i32 [[TMP0]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult i64 [[TMP1]], 90
 ; CHECK-NEXT:    [[UMIN:%.*]] = select i1 [[TMP2]], i64 [[TMP1]], i64 90
-; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw i64 [[UMIN]], -99
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[UMIN]], -99
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ -100, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/IndVarSimplify/loop_evaluate_2.ll b/llvm/test/Transforms/IndVarSimplify/loop_evaluate_2.ll
index 930819e77d339..3d416d33e130e 100644
--- a/llvm/test/Transforms/IndVarSimplify/loop_evaluate_2.ll
+++ b/llvm/test/Transforms/IndVarSimplify/loop_evaluate_2.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s -indvars -loop-deletion -simplifycfg | opt -analyze -loops | FileCheck %s
+; RUN: opt < %s -indvars -loop-deletion -simplifycfg | opt -analyze -loops -enable-new-pm=0 | FileCheck %s
 ; PR1179
 
 ; CHECK-NOT: Loop Containing
diff --git a/llvm/test/Transforms/IndVarSimplify/promote-iv-to-eliminate-casts.ll b/llvm/test/Transforms/IndVarSimplify/promote-iv-to-eliminate-casts.ll
index 090efa44d9c04..5cc288c58e68c 100644
--- a/llvm/test/Transforms/IndVarSimplify/promote-iv-to-eliminate-casts.ll
+++ b/llvm/test/Transforms/IndVarSimplify/promote-iv-to-eliminate-casts.ll
@@ -196,7 +196,7 @@ define void @promote_latch_condition_decrementing_loop_01(i32* %p, i32* %a) {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[TMP0]], [[PREHEADER]] ]
 ; CHECK-NEXT:    [[EL:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store atomic i32 0, i32* [[EL]] unordered, align 4
-; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp ult i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp slt i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
 ; CHECK-NEXT:    br i1 [[LOOPCOND]], label [[LOOPEXIT_LOOPEXIT:%.*]], label [[LOOP]]
 ;
@@ -241,7 +241,7 @@ define void @promote_latch_condition_decrementing_loop_02(i32* %p, i32* %a) {
 ; CHECK-NEXT:    [[EL:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store atomic i32 0, i32* [[EL]] unordered, align 4
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp ult i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp slt i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    br i1 [[LOOPCOND]], label [[LOOPEXIT_LOOPEXIT:%.*]], label [[LOOP]]
 ;
 
@@ -285,7 +285,7 @@ define void @promote_latch_condition_decrementing_loop_03(i32* %p, i32* %a) {
 ; CHECK-NEXT:    [[EL:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store atomic i32 0, i32* [[EL]] unordered, align 4
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
-; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp ult i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp slt i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    br i1 [[LOOPCOND]], label [[LOOPEXIT_LOOPEXIT:%.*]], label [[LOOP]]
 ;
 
@@ -336,7 +336,7 @@ define void @promote_latch_condition_decrementing_loop_04(i32* %p, i32* %a, i1 %
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ [[TMP0]], [[PREHEADER]] ]
 ; CHECK-NEXT:    [[EL:%.*]] = getelementptr inbounds i32, i32* [[A:%.*]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store atomic i32 0, i32* [[EL]] unordered, align 4
-; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp ult i64 [[INDVARS_IV]], 1
+; CHECK-NEXT:    [[LOOPCOND:%.*]] = icmp slt i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
 ; CHECK-NEXT:    br i1 [[LOOPCOND]], label [[LOOPEXIT_LOOPEXIT:%.*]], label [[LOOP]]
 ;
diff --git a/llvm/test/Transforms/IndVarSimplify/shrunk-constant.ll b/llvm/test/Transforms/IndVarSimplify/shrunk-constant.ll
index ba20b2c6a512e..cd8720cb4b38b 100644
--- a/llvm/test/Transforms/IndVarSimplify/shrunk-constant.ll
+++ b/llvm/test/Transforms/IndVarSimplify/shrunk-constant.ll
@@ -1,4 +1,5 @@
-; RUN: opt < %s -scalar-evolution -analyze | FileCheck %s
+; RUN: opt < %s -scalar-evolution -analyze -enable-new-pm=0 | FileCheck %s
+; RUN: opt < %s -passes='print<scalar-evolution>' 2>&1 | FileCheck %s
 
 ; CHECK: -->  (1 + (zext i4 {-8,+,-8}<%loop> to i32))<nuw><nsw>
 
diff --git a/llvm/test/Transforms/InstCombine/constant-fold-address-space-pointer.ll b/llvm/test/Transforms/InstCombine/constant-fold-address-space-pointer.ll
index 5a28354eb532d..04f15ece02ab5 100644
--- a/llvm/test/Transforms/InstCombine/constant-fold-address-space-pointer.ll
+++ b/llvm/test/Transforms/InstCombine/constant-fold-address-space-pointer.ll
@@ -225,7 +225,7 @@ define i32 @test_cast_gep_large_indices_as() {
 
 define i32 @test_constant_cast_gep_struct_indices_as() {
 ; CHECK-LABEL: @test_constant_cast_gep_struct_indices_as(
-; CHECK-NEXT:    [[Y:%.*]] = load i32, i32 addrspace(3)* getelementptr inbounds (%struct.foo, [[STRUCT_FOO:%.*]] addrspace(3)* @constant_fold_global_ptr, i16 0, i32 2, i16 2), align 8
+; CHECK-NEXT:    [[Y:%.*]] = load i32, i32 addrspace(3)* getelementptr inbounds (%struct.foo, [[STRUCT_FOO:%.*]] addrspace(3)* @constant_fold_global_ptr, i16 0, i32 2, i16 2), align 16
 ; CHECK-NEXT:    ret i32 [[Y]]
 ;
   %x = getelementptr %struct.foo, %struct.foo addrspace(3)* @constant_fold_global_ptr, i18 0, i32 2, i12 2
diff --git a/llvm/test/Transforms/InstCombine/constant-fold-gep.ll b/llvm/test/Transforms/InstCombine/constant-fold-gep.ll
index 277a900a48b7d..cd49353f707bf 100644
--- a/llvm/test/Transforms/InstCombine/constant-fold-gep.ll
+++ b/llvm/test/Transforms/InstCombine/constant-fold-gep.ll
@@ -15,20 +15,20 @@ define void @frob() {
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 1), align 4
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 2), align 8
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 1, i64 0), align 4
-; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 1, i64 1), align 4
+; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 1, i64 1), align 16
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 1, i64 2), align 4
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 0, i64 0), align 8
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 0, i64 1), align 4
-; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 0, i64 2), align 8
+; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 0, i64 2), align 16
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 1, i64 0), align 4
-; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 1, i64 1), align 4
+; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 1, i64 1), align 8
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 1, i32 1, i64 2), align 4
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 0, i64 0), align 16
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 0, i64 1), align 4
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 0, i64 2), align 8
-; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 1, i64 0), align 8
-; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 1, i64 1), align 8
-; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 1, i64 2), align 8
+; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 1, i64 0), align 4
+; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 1, i64 1), align 16
+; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 2, i32 1, i64 2), align 4
 ; CHECK-NEXT:    store i32 1, i32* getelementptr inbounds ([3 x %struct.X], [3 x %struct.X]* @Y, i64 1, i64 0, i32 0, i64 0), align 8
 ; CHECK-NEXT:    store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 2, i64 0, i32 0, i64 0), align 16
 ; CHECK-NEXT:    store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 1, i64 0, i32 0, i64 1), align 8
@@ -49,9 +49,9 @@ define void @frob() {
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 12), align 4
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 13), align 4
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 14), align 8
-  store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 15), align 8
+  store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 15), align 4
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 16), align 8
-  store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 17), align 8
+  store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 17), align 4
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 18), align 8
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 36), align 8
   store i32 1, i32* getelementptr ([3 x %struct.X], [3 x %struct.X]* @Y, i64 0, i64 0, i32 0, i64 19), align 8
@@ -98,3 +98,22 @@ entry:
 
   ret i16 %E
 }
+
+; Check that we improve the alignment information.
+; The base pointer is 16-byte aligned and we access the field at
+; an offset of 8-byte.
+; Every element in the @CallerInfos array is 16-byte aligned so
+; any access from the following gep is 8-byte aligned.
+%struct.CallerInfo = type { i8*, i32 }
+@CallerInfos = global [128 x %struct.CallerInfo] zeroinitializer, align 16
+
+define i32 @test_gep_in_struct(i64 %idx) {
+; CHECK-LABEL: @test_gep_in_struct(
+; CHECK-NEXT:    [[NS7:%.*]] = getelementptr inbounds [128 x %struct.CallerInfo], [128 x %struct.CallerInfo]* @CallerInfos, i64 0, i64 [[IDX:%.*]], i32 1
+; CHECK-NEXT:    [[RES:%.*]] = load i32, i32* [[NS7]], align 8
+; CHECK-NEXT:    ret i32 [[RES]]
+;
+  %NS7 = getelementptr inbounds [128 x %struct.CallerInfo], [128 x %struct.CallerInfo]* @CallerInfos, i64 0, i64 %idx, i32 1
+  %res = load i32, i32* %NS7, align 1
+  ret i32 %res
+}
diff --git a/llvm/test/Transforms/InstCombine/or.ll b/llvm/test/Transforms/InstCombine/or.ll
index 7e4115cc8934d..b5e3af2c76525 100644
--- a/llvm/test/Transforms/InstCombine/or.ll
+++ b/llvm/test/Transforms/InstCombine/or.ll
@@ -650,10 +650,12 @@ define i1 @test46(i8 signext %c)  {
 
 define <2 x i1> @test46_uniform(<2 x i8> %c)  {
 ; CHECK-LABEL: @test46_uniform(
-; CHECK-NEXT:    [[TMP1:%.*]] = and <2 x i8> [[C:%.*]], <i8 -33, i8 -33>
-; CHECK-NEXT:    [[TMP2:%.*]] = add <2 x i8> [[TMP1]], <i8 -65, i8 -65>
-; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult <2 x i8> [[TMP2]], <i8 26, i8 26>
-; CHECK-NEXT:    ret <2 x i1> [[TMP3]]
+; CHECK-NEXT:    [[C_OFF:%.*]] = add <2 x i8> [[C:%.*]], <i8 -97, i8 -97>
+; CHECK-NEXT:    [[CMP1:%.*]] = icmp ult <2 x i8> [[C_OFF]], <i8 26, i8 26>
+; CHECK-NEXT:    [[C_OFF17:%.*]] = add <2 x i8> [[C]], <i8 -65, i8 -65>
+; CHECK-NEXT:    [[CMP2:%.*]] = icmp ult <2 x i8> [[C_OFF17]], <i8 26, i8 26>
+; CHECK-NEXT:    [[OR:%.*]] = or <2 x i1> [[CMP1]], [[CMP2]]
+; CHECK-NEXT:    ret <2 x i1> [[OR]]
 ;
   %c.off = add <2 x i8> %c, <i8 -97, i8 -97>
   %cmp1 = icmp ult <2 x i8> %c.off, <i8 26, i8 26>
diff --git a/llvm/test/Transforms/InstSimplify/compare.ll b/llvm/test/Transforms/InstSimplify/compare.ll
index a029d0b3c60db..d0e5f3a73529a 100644
--- a/llvm/test/Transforms/InstSimplify/compare.ll
+++ b/llvm/test/Transforms/InstSimplify/compare.ll
@@ -1765,16 +1765,11 @@ define i1 @cmp_through_addrspacecast(i32 addrspace(1)* %p1) {
 
 ; Test simplifications for: icmp (X+Y), (X+Z) -> icmp Y,Z
 ; Test the overflow check when the RHS has NSW set and constant Z is greater
-; or equal than Y, then we know X+Y also can't overflow.
+; than Y, then we know X+Y also can't overflow.
 
 define i1 @icmp_nsw_1(i32 %V) {
 ; CHECK-LABEL: @icmp_nsw_1(
-; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], 5
-; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[V]], 6
-; CHECK-NEXT:    [[S1:%.*]] = sext i32 [[ADD5]] to i64
-; CHECK-NEXT:    [[S2:%.*]] = sext i32 [[ADD6]] to i64
-; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i64 [[S1]], [[S2]]
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %add5 = add i32 %V, 5
   %add6 = add nsw i32 %V, 6
@@ -1786,10 +1781,7 @@ define i1 @icmp_nsw_1(i32 %V) {
 
 define i1 @icmp_nsw_2(i32 %V) {
 ; CHECK-LABEL: @icmp_nsw_2(
-; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], 5
-; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[V]], 6
-; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[ADD5]], [[ADD6]]
-; CHECK-NEXT:    ret i1 [[CMP]]
+; CHECK-NEXT:    ret i1 true
 ;
   %add5 = add i32 %V, 5
   %add6 = add nsw i32 %V, 6
@@ -1797,6 +1789,158 @@ define i1 @icmp_nsw_2(i32 %V) {
   ret i1 %cmp
 }
 
+define i1 @icmp_nsw_commute(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_commute(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add i32 5, %V
+  %add6 = add nsw i32 %V, 6
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_commute2(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_commute2(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add i32 %V, 5
+  %add6 = add nsw i32 6, %V
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_commute3(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_commute3(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add i32 5, %V
+  %add6 = add nsw i32 6, %V
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_22(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_22(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add nsw i32 %V, 5
+  %add6 = add nsw i32 %V, 6
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_23(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_23(
+; CHECK-NEXT:    [[ADD5:%.*]] = add nsw i32 [[V:%.*]], 5
+; CHECK-NEXT:    [[ADD6:%.*]] = add i32 [[V]], 6
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[ADD5]], [[ADD6]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add5 = add nsw i32 %V, 5
+  %add6 = add i32 %V, 6
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_false(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_false(
+; CHECK-NEXT:    ret i1 false
+;
+  %add5 = add nsw i32 %V, 6
+  %add6 = add i32 %V, 5
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_false_2(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_false_2(
+; CHECK-NEXT:    ret i1 false
+;
+  %add5 = add nsw i32 %V, 6
+  %add6 = add nsw i32 %V, 5
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_false_3(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_false_3(
+; CHECK-NEXT:    [[ADD5:%.*]] = add nsw i32 [[V:%.*]], 5
+; CHECK-NEXT:    [[ADD6:%.*]] = add i32 [[V]], 5
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[ADD5]], [[ADD6]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add5 = add nsw i32 %V, 5
+  %add6 = add i32 %V, 5
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_false_4(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_false_4(
+; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], 6
+; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[V]], 5
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[ADD5]], [[ADD6]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add5 = add i32 %V, 6
+  %add6 = add nsw i32 %V, 5
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_false_5(i8 %V) {
+; CHECK-LABEL: @icmp_nsw_false_5(
+; CHECK-NEXT:    [[ADD:%.*]] = add i8 [[V:%.*]], 121
+; CHECK-NEXT:    [[ADDNSW:%.*]] = add nsw i8 [[V]], -104
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i8 [[ADD]], [[ADDNSW]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add = add i8 %V, 121
+  %addnsw = add nsw i8 %V, -104
+  %cmp = icmp slt i8 %add, %addnsw
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_i8(i8 %V) {
+; CHECK-LABEL: @icmp_nsw_i8(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add i8 %V, 5
+  %add6 = add nsw i8 %V, 6
+  %cmp = icmp slt i8 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_i16(i16 %V) {
+; CHECK-LABEL: @icmp_nsw_i16(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add i16 %V, 0
+  %add6 = add nsw i16 %V, 1
+  %cmp = icmp slt i16 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_i64(i64 %V) {
+; CHECK-LABEL: @icmp_nsw_i64(
+; CHECK-NEXT:    ret i1 true
+;
+  %add5 = add i64 %V, 5
+  %add6 = add nsw i64 %V, 6
+  %cmp = icmp slt i64 %add5, %add6
+  ret i1 %cmp
+}
+
+define <4 x i1> @icmp_nsw_vec(<4 x i32> %V) {
+; CHECK-LABEL: @icmp_nsw_vec(
+; CHECK-NEXT:    ret <4 x i1> <i1 true, i1 true, i1 true, i1 true>
+;
+  %add5 = add <4 x i32> %V, <i32 5, i32 5, i32 5, i32 5>
+  %add6 = add nsw <4 x i32> %V, <i32 6, i32 6, i32 6, i32 6>
+  %cmp = icmp slt <4 x i32> %add5, %add6
+  ret <4 x i1> %cmp
+}
+
 define i1 @icmp_nsw_3(i32 %V) {
 ; CHECK-LABEL: @icmp_nsw_3(
 ; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], 5
@@ -1878,14 +2022,49 @@ define i1 @icmp_nsw_9(i32 %V1, i32 %V2) {
 define i1 @icmp_nsw_10(i32 %V) {
 ; CHECK-LABEL: @icmp_nsw_10(
 ; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], 5
-; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[V]], 5
+; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[V]], 6
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp sgt i32 [[ADD6]], [[ADD5]]
 ; CHECK-NEXT:    ret i1 [[CMP]]
 ;
   %add5 = add i32 %V, 5
-  %add6 = add nsw i32 %V, 5
+  %add6 = add nsw i32 %V, 6
   %cmp = icmp sgt i32 %add6, %add5
   ret i1 %cmp
 }
 
+define i1 @icmp_nsw_11(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_11(
+; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], -125
+; CHECK-NEXT:    [[ADD6:%.*]] = add nsw i32 [[V]], -99
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[ADD5]], [[ADD6]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add5 = add i32 %V, -125
+  %add6 = add nsw i32 %V, -99
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_nonpos(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_nonpos(
+; CHECK-NEXT:    ret i1 false
+;
+  %add5 = add i32 %V, 0
+  %add6 = add nsw i32 %V, -1
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
+define i1 @icmp_nsw_nonpos2(i32 %V) {
+; CHECK-LABEL: @icmp_nsw_nonpos2(
+; CHECK-NEXT:    [[ADD5:%.*]] = add i32 [[V:%.*]], 1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[ADD5]], [[V]]
+; CHECK-NEXT:    ret i1 [[CMP]]
+;
+  %add5 = add i32 %V, 1
+  %add6 = add nsw i32 %V, 0
+  %cmp = icmp slt i32 %add5, %add6
+  ret i1 %cmp
+}
+
 attributes #0 = { null_pointer_is_valid }
diff --git a/llvm/test/Transforms/LoopDeletion/2017-07-11-incremental-dt.ll b/llvm/test/Transforms/LoopDeletion/2017-07-11-incremental-dt.ll
index ad4e434c11d9c..ad8b8b4c6c483 100644
--- a/llvm/test/Transforms/LoopDeletion/2017-07-11-incremental-dt.ll
+++ b/llvm/test/Transforms/LoopDeletion/2017-07-11-incremental-dt.ll
@@ -1,6 +1,7 @@
 ; RUN: opt < %s -loop-deletion -S
-; RUN: opt < %s -loop-deletion -analyze -domtree 2>&1 | FileCheck -check-prefix=DT %s
-; RUN: opt < %s -loop-deletion -analyze -verify-dom-info
+; RUN: opt < %s -loop-deletion -analyze -domtree 2>&1 -enable-new-pm=0 | FileCheck -check-prefix=DT %s
+; RUN: opt < %s -passes='loop(loop-deletion),print<domtree>' 2>&1 | FileCheck -check-prefix=DT %s
+; RUN: opt < %s -loop-deletion -verify-dom-info
 
 ; CHECK: for.body
 ; CHECK-NOT: for.cond1
diff --git a/llvm/test/Transforms/LoopDeletion/update-scev.ll b/llvm/test/Transforms/LoopDeletion/update-scev.ll
index 528a00a93f66b..44d23aa4060d1 100644
--- a/llvm/test/Transforms/LoopDeletion/update-scev.ll
+++ b/llvm/test/Transforms/LoopDeletion/update-scev.ll
@@ -1,4 +1,5 @@
-; RUN: opt -S -analyze -scalar-evolution -loop-deletion -scalar-evolution -verify-scev < %s | FileCheck %s --check-prefix=SCEV-EXPRS
+; RUN: opt -S -analyze -scalar-evolution -loop-deletion -scalar-evolution -verify-scev < %s -enable-new-pm=0 | FileCheck %s --check-prefix=SCEV-EXPRS
+; RUN: opt -S -passes='print<scalar-evolution>,loop-deletion,print<scalar-evolution>' -verify-scev < %s 2>&1 | FileCheck %s --check-prefix=SCEV-EXPRS
 ; RUN: opt -S -loop-deletion < %s | FileCheck %s --check-prefix=IR-AFTER-TRANSFORM
 ; RUN: opt -S -indvars -loop-deletion -indvars < %s | FileCheck %s --check-prefix=ORIGINAL-CRASH
 
diff --git a/llvm/test/Transforms/LoopDistribute/basic-with-memchecks.ll b/llvm/test/Transforms/LoopDistribute/basic-with-memchecks.ll
index 5f72b8d49b51b..4a48c16112103 100644
--- a/llvm/test/Transforms/LoopDistribute/basic-with-memchecks.ll
+++ b/llvm/test/Transforms/LoopDistribute/basic-with-memchecks.ll
@@ -6,7 +6,11 @@
 ; RUN:   FileCheck --check-prefix=VECTORIZE %s
 
 ; RUN: opt -basic-aa -loop-distribute -enable-loop-distribute -verify-loop-info -verify-dom-info \
-; RUN:   -loop-accesses -analyze < %s | FileCheck %s --check-prefix=ANALYSIS
+; RUN:   -loop-accesses -analyze < %s -enable-new-pm=0 | FileCheck %s --check-prefix=ANALYSIS
+
+; TODO: the following changes the order loop-access printing prints loops, remove legacy RUN and change after NPM switch
+; TODO: opt -aa-pipeline=basic-aa -passes='loop-distribute,print-access-info' -enable-loop-distribute \
+; TODO:   -verify-loop-info -verify-dom-info -disable-output < %s 2>&1 | FileCheck %s --check-prefix=ANALYSIS
 
 ; The memcheck version of basic.ll.  We should distribute and vectorize the
 ; second part of this loop with 5 memchecks (A+1 x {C, D, E} + C x {A, B})
diff --git a/llvm/test/Transforms/LoopDistribute/basic.ll b/llvm/test/Transforms/LoopDistribute/basic.ll
index d696c0cf3a45c..85795eed155d7 100644
--- a/llvm/test/Transforms/LoopDistribute/basic.ll
+++ b/llvm/test/Transforms/LoopDistribute/basic.ll
@@ -2,7 +2,11 @@
 ; RUN:   < %s | FileCheck %s
 
 ; RUN: opt -basic-aa -loop-distribute -enable-loop-distribute -verify-loop-info -verify-dom-info \
-; RUN:   -loop-accesses -analyze < %s | FileCheck %s --check-prefix=ANALYSIS
+; RUN:   -loop-accesses -analyze < %s -enable-new-pm=0 | FileCheck %s --check-prefix=ANALYSIS
+
+; TODO: the following changes the order loop-access printing prints loops, remove legacy RUN and change after NPM switch
+; TODO: opt -aa-pipeline=basic-aa -passes='loop-distribute,print-access-info' -enable-loop-distribute \
+; TODO:   -verify-loop-info -verify-dom-info -disable-output < %s 2>&1 | FileCheck %s --check-prefix=ANALYSIS
 
 ; RUN: opt -basic-aa -loop-distribute -enable-loop-distribute -loop-vectorize -force-vector-width=4 -S \
 ; RUN:   < %s | FileCheck %s --check-prefix=VECTORIZE
diff --git a/llvm/test/Transforms/LoopRotate/oz-disable.ll b/llvm/test/Transforms/LoopRotate/oz-disable.ll
index 7a6a9bf33a188..5ac7864b33aa3 100644
--- a/llvm/test/Transforms/LoopRotate/oz-disable.ll
+++ b/llvm/test/Transforms/LoopRotate/oz-disable.ll
@@ -1,6 +1,8 @@
 ; REQUIRES: asserts
 ; RUN: opt < %s -S -Os -debug -debug-only=loop-rotate 2>&1 | FileCheck %s -check-prefix=OS
 ; RUN: opt < %s -S -Oz -debug -debug-only=loop-rotate 2>&1 | FileCheck %s -check-prefix=OZ
+; RUN: opt < %s -S -passes='default<Os>' -debug -debug-only=loop-rotate 2>&1 | FileCheck %s -check-prefix=OS
+; RUN: opt < %s -S -passes='default<Oz>' -debug -debug-only=loop-rotate 2>&1 | FileCheck %s -check-prefix=OZ
 
 ; Loop should be rotated for -Os but not for -Oz.
 ; OS: rotating Loop at depth 1
diff --git a/llvm/test/Transforms/LoopSimplify/2010-07-15-IncorrectDomFrontierUpdate.ll b/llvm/test/Transforms/LoopSimplify/2010-07-15-IncorrectDomFrontierUpdate.ll
index f179da234cb14..4f215ab6d15a4 100644
--- a/llvm/test/Transforms/LoopSimplify/2010-07-15-IncorrectDomFrontierUpdate.ll
+++ b/llvm/test/Transforms/LoopSimplify/2010-07-15-IncorrectDomFrontierUpdate.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s -domfrontier -loop-simplify -domfrontier -verify-dom-info -analyze 
+; RUN: opt < %s -domfrontier -loop-simplify -domfrontier -verify-dom-info
 
 
 define void @a() nounwind {
diff --git a/llvm/test/Transforms/LoopSimplify/preserve-scev.ll b/llvm/test/Transforms/LoopSimplify/preserve-scev.ll
index 07486ef246861..275a54005a1f3 100644
--- a/llvm/test/Transforms/LoopSimplify/preserve-scev.ll
+++ b/llvm/test/Transforms/LoopSimplify/preserve-scev.ll
@@ -1,4 +1,5 @@
-; RUN: opt -S < %s -analyze -scalar-evolution -loop-simplify -scalar-evolution | FileCheck %s
+; RUN: opt -S < %s -analyze -scalar-evolution -loop-simplify -scalar-evolution -enable-new-pm=0 | FileCheck %s
+; RUN: opt -S < %s -passes='print<scalar-evolution>,loop-simplify,print<scalar-evolution>' 2>&1 | FileCheck %s
 
 ; Provide legal integer types.
 target datalayout = "n8:16:32:64"
diff --git a/llvm/test/Transforms/LoopStrengthReduce/pr3086.ll b/llvm/test/Transforms/LoopStrengthReduce/pr3086.ll
index 187c14f900a56..7e4e5b2807881 100644
--- a/llvm/test/Transforms/LoopStrengthReduce/pr3086.ll
+++ b/llvm/test/Transforms/LoopStrengthReduce/pr3086.ll
@@ -1,5 +1,5 @@
 ; RUN: opt < %s -loop-reduce
-; RUN: opt < %s -analyze -scalar-evolution
+; RUN: opt < %s -scalar-evolution
 ; PR 3086
 
 	%struct.Cls = type { i32, i8, [2 x %struct.Cls*], [2 x %struct.Lit*] }
diff --git a/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused.ll b/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused.ll
index 7300a5ff77030..edc3034c4917f 100644
--- a/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused.ll
+++ b/llvm/test/Transforms/LowerMatrixIntrinsics/multiply-fused.ll
@@ -1,4 +1,5 @@
 ; RUN: opt -lower-matrix-intrinsics -fuse-matrix-use-loops=false -fuse-matrix-tile-size=2 -matrix-allow-contract -force-fuse-matrix -instcombine -verify-dom-info %s -S | FileCheck %s
+; RUN: opt -passes=lower-matrix-intrinsics,instcombine -fuse-matrix-use-loops=false -fuse-matrix-tile-size=2 -matrix-allow-contract -force-fuse-matrix -verify-dom-info %s -S | FileCheck %s
 
 ; REQUIRES: aarch64-registered-target
 
diff --git a/llvm/test/Transforms/OpenMP/add_attributes.ll b/llvm/test/Transforms/OpenMP/add_attributes.ll
index cf1bd246d2b29..a448117b96a51 100644
--- a/llvm/test/Transforms/OpenMP/add_attributes.ll
+++ b/llvm/test/Transforms/OpenMP/add_attributes.ll
@@ -629,6 +629,10 @@ declare void @__kmpc_destroy_allocator(i32, i8*)
 
 declare void @__kmpc_push_target_tripcount(i64, i64)
 
+declare i32 @__kmpc_warp_active_thread_mask()
+
+declare void @__kmpc_syncwarp(i32)
+
 declare i32 @__tgt_target_mapper(i64, i8*, i32, i8**, i8**, i64*, i64*, i8**)
 
 declare i32 @__tgt_target_nowait_mapper(i64, i8*, i32, i8**, i8**, i64*, i64*, i8**)
@@ -1142,6 +1146,12 @@ declare void @__kmpc_proxy_task_completed_ooo(i8*)
 ; CHECK: ; Function Attrs: nounwind
 ; CHECK-NEXT: declare void @__kmpc_push_target_tripcount(i64, i64)
 
+; CHECK: ; Function Attrs: convergent nounwind
+; CHECK-NEXT: declare i32 @__kmpc_warp_active_thread_mask()
+
+; CHECK: ; Function Attrs: convergent nounwind
+; CHECK-NEXT: declare void @__kmpc_syncwarp(i32)
+
 ; CHECK: ; Function Attrs: nounwind
 ; CHECK-NEXT: declare i32 @__tgt_target_mapper(i64, i8*, i32, i8**, i8**, i64*, i64*, i8**)
 
@@ -1661,6 +1671,12 @@ declare void @__kmpc_proxy_task_completed_ooo(i8*)
 ; OPTIMISTIC: ; Function Attrs: inaccessiblememonly nofree nosync nounwind willreturn writeonly
 ; OPTIMISTIC-NEXT: declare void @__kmpc_push_target_tripcount(i64, i64)
 
+; OPTIMISTIC: ; Function Attrs: convergent nounwind
+; OPTIMISTIC-NEXT: declare i32 @__kmpc_warp_active_thread_mask()
+
+; OPTIMISTIC: ; Function Attrs: convergent nounwind
+; OPTIMISTIC-NEXT: declare void @__kmpc_syncwarp(i32)
+
 ; OPTIMISTIC: ; Function Attrs: nounwind
 ; OPTIMISTIC-NEXT: declare i32 @__tgt_target_mapper(i64, i8*, i32, i8**, i8**, i64*, i64*, i8**)
 
diff --git a/llvm/test/Transforms/OpenMP/add_attributes_amdgcn.ll b/llvm/test/Transforms/OpenMP/add_attributes_amdgcn.ll
new file mode 100644
index 0000000000000..acf547d4437f6
--- /dev/null
+++ b/llvm/test/Transforms/OpenMP/add_attributes_amdgcn.ll
@@ -0,0 +1,28 @@
+; RUN: opt < %s -S -openmpopt        | FileCheck %s
+; RUN: opt < %s -S -passes=openmpopt | FileCheck %s
+; RUN: opt < %s -S -openmpopt        -openmp-ir-builder-optimistic-attributes | FileCheck %s --check-prefix=OPTIMISTIC
+; RUN: opt < %s -S -passes=openmpopt -openmp-ir-builder-optimistic-attributes | FileCheck %s --check-prefix=OPTIMISTIC
+
+target triple = "amdgcn-amd-amdhsa"
+
+define void @call_all(i64 %arg) {
+  call void @__kmpc_syncwarp(i64 %arg)
+  call i64 @__kmpc_warp_active_thread_mask()
+  ret void
+}
+
+declare i64 @__kmpc_warp_active_thread_mask()
+
+declare void @__kmpc_syncwarp(i64)
+
+; CHECK: ; Function Attrs: convergent nounwind
+; CHECK-NEXT: declare i64 @__kmpc_warp_active_thread_mask()
+
+; CHECK: ; Function Attrs: convergent nounwind
+; CHECK-NEXT: declare void @__kmpc_syncwarp(i64)
+
+; OPTIMISTIC: ; Function Attrs: convergent nounwind
+; OPTIMISTIC-NEXT: declare i64 @__kmpc_warp_active_thread_mask()
+
+; OPTIMISTIC: ; Function Attrs: convergent nounwind
+; OPTIMISTIC-NEXT: declare void @__kmpc_syncwarp(i64)
diff --git a/llvm/test/Transforms/Util/PredicateInfo/condprop.ll b/llvm/test/Transforms/Util/PredicateInfo/condprop.ll
index 756457ab7fa9a..4f5f07878cd2a 100644
--- a/llvm/test/Transforms/Util/PredicateInfo/condprop.ll
+++ b/llvm/test/Transforms/Util/PredicateInfo/condprop.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -print-predicateinfo -analyze  < %s 2>&1 | FileCheck %s
+; RUN: opt -print-predicateinfo < %s 2>&1 | FileCheck %s
 
 @a = external global i32		; <i32*> [#uses=7]
 
diff --git a/llvm/test/Transforms/Util/PredicateInfo/edge.ll b/llvm/test/Transforms/Util/PredicateInfo/edge.ll
index 31c8faae496b0..dbd15dc70ba52 100644
--- a/llvm/test/Transforms/Util/PredicateInfo/edge.ll
+++ b/llvm/test/Transforms/Util/PredicateInfo/edge.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -print-predicateinfo -analyze  < %s 2>&1 | FileCheck %s
+; RUN: opt -print-predicateinfo < %s 2>&1 | FileCheck %s
 
 define i32 @f1(i32 %x) {
 ; CHECK-LABEL: @f1(
diff --git a/llvm/test/Transforms/Util/PredicateInfo/pr33456.ll b/llvm/test/Transforms/Util/PredicateInfo/pr33456.ll
index f1cc83a071b96..99e168acde465 100644
--- a/llvm/test/Transforms/Util/PredicateInfo/pr33456.ll
+++ b/llvm/test/Transforms/Util/PredicateInfo/pr33456.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -print-predicateinfo -analyze  < %s 2>&1 | FileCheck %s
+; RUN: opt -print-predicateinfo < %s 2>&1 | FileCheck %s
 ; Don't insert predicate info for conditions with a single target.
 @a = global i32 1, align 4
 @d = common global i32 0, align 4
diff --git a/llvm/test/Transforms/Util/PredicateInfo/pr33457.ll b/llvm/test/Transforms/Util/PredicateInfo/pr33457.ll
index b975ade9321db..4db0675bd80e1 100644
--- a/llvm/test/Transforms/Util/PredicateInfo/pr33457.ll
+++ b/llvm/test/Transforms/Util/PredicateInfo/pr33457.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt -print-predicateinfo -analyze  < %s 2>&1 | FileCheck %s
+; RUN: opt -print-predicateinfo < %s 2>&1 | FileCheck %s
 ; Don't insert predicate info for conditions with a single target.
 @a = global i32 6, align 4
 @c = global i32 -1, align 4
diff --git a/llvm/test/tools/llvm-mca/ARM/cortex-a57-memory-instructions.s b/llvm/test/tools/llvm-mca/ARM/cortex-a57-memory-instructions.s
new file mode 100644
index 0000000000000..d1e0880506d76
--- /dev/null
+++ b/llvm/test/tools/llvm-mca/ARM/cortex-a57-memory-instructions.s
@@ -0,0 +1,475 @@
+# NOTE: Assertions have been autogenerated by utils/update_mca_test_checks.py
+# RUN: llvm-mca -mtriple=armv8 -mcpu=cortex-a57 -instruction-tables < %s | FileCheck %s
+
+  .text
+  ldr	r5, [r7]
+  ldr	r6, [r3, #63]
+  ldr	r2, [r4, #4095]!
+  ldr	r1, [r2], #30
+  ldr	r3, [r1], #-30
+  ldr	r3, [r8, r1]
+  ldr	r2, [r5, -r3]
+  ldr	r1, [r5, r9]!
+  ldr	r6, [r7, -r8]!
+  ldr	r1, [r0, r2, lsr #3]!
+  ldr	r5, [r9], r2
+  ldr	r4, [r3], -r6
+  ldr	r3, [r8, -r2, lsl #15]
+  ldr	r1, [r5], r3, asr #15
+  ldrb	r3, [r8]
+  ldrb	r1, [sp, #63]
+  ldrb	r9, [r3, #4095]!
+  ldrb	r8, [r1], #22
+  ldrb	r2, [r7], #-19
+  ldrb	r9, [r8, r5]
+  ldrb	r1, [r5, -r1]
+  ldrb	r3, [r5, r2]!
+  ldrb	r6, [r9, -r3]!
+  ldrb	r2, [r1], r4
+  ldrb	r8, [r4], -r5
+  ldrb	r7, [r12, -r1, lsl #15]
+  ldrb	r5, [r2], r9, asr #15
+  ldrbt	r3, [r1], #4
+  ldrbt	r2, [r8], #-8
+  ldrbt	r8, [r7], r6
+  ldrbt	r1, [r2], -r6, lsl #12
+  ldrd	r0, r1, [r5]
+  ldrd	r8, r9, [r2, #15]
+  ldrd	r2, r3, [r9, #32]!
+  ldrd	r6, r7, [r1], #8
+  ldrd	r2, r3, [r8], #0
+  ldrd	r2, r3, [r8], #0
+  ldrd	r2, r3, [r8], #-0
+  ldrd	r4, r5, [r1, r3]
+  ldrd	r4, r5, [r7, r2]!
+  ldrd	r0, r1, [r8], r12
+  ldrd	r0, r1, [r8], -r12
+  ldrh	r3, [r4]
+  ldrh	r2, [r7, #4]
+  ldrh	r1, [r8, #64]!
+  ldrh	r12, [sp], #4
+  ldrh	r6, [r5, r4]
+  ldrh	r3, [r8, r11]!
+  ldrh	r1, [r2, -r1]!
+  ldrh	r9, [r7], r2
+  ldrh	r4, [r3], -r2
+  ldrht	r9, [r7], #128
+  ldrht	r4, [r3], #-75
+  ldrht	r9, [r7], r2
+  ldrht	r4, [r3], -r2
+  ldrsb	r3, [r4]
+  ldrsb	r2, [r7, #17]
+  ldrsb	r1, [r8, #255]!
+  ldrsb	r12, [sp], #9
+  ldrsb	r6, [r5, r4]
+  ldrsb	r3, [r8, r11]!
+  ldrsb	r1, [r2, -r1]!
+  ldrsb	r9, [r7], r2
+  ldrsb	r4, [r3], -r2
+  ldrsbt	r5, [r6], #1
+  ldrsbt	r3, [r8], #-12
+  ldrsbt	r8, [r9], r5
+  ldrsbt	r2, [r1], -r4
+  ldrsh	r5, [r9]
+  ldrsh	r4, [r5, #7]
+  ldrsh	r3, [r6, #55]!
+  ldrsh	r2, [r7], #-9
+  ldrsh	r3, [r1, r5]
+  ldrsh	r4, [r6, r1]!
+  ldrsh	r5, [r3, -r6]!
+  ldrsh	r6, [r9], r8
+  ldrsh	r7, [r8], -r3
+  ldrsht	r5, [r6], #1
+  ldrsht	r3, [r8], #-12
+  ldrsht	r8, [r9], r5
+  ldrsht	r2, [r1], -r4
+  ldm r2, {r1, r2, r4, r5, r6}
+  ldmia r2, {r1, r2, r4, r5, r6}
+  ldmia r2, {r1, r3, r4, r5, r6}
+  ldmib r2, {r1, r2}
+  ldmdb r2, {r1, r2}
+  ldmib r2, {r1, r3}
+  ldmib r2, {r1, r3, r5}
+  ldmib r2, {r1, r2, r5}
+  ldmdbeq r2, {r1, r2}
+  ldmibeq r2, {r1, r3}
+  ldmia r2, {r0, r1, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15}
+  ldmia r2, {r0, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, r13, r14, r15}
+  str	r8, [r12]
+  str	r7, [r1, #12]
+  str	r3, [r5, #40]!
+  str	r9, [sp], #4095
+  str	r1, [r7], #-128
+  str	r9, [r6, r3]
+  str	r8, [r0, -r2]
+  str	r7, [r1, r6]!
+  str	r7, [r1, r6, lsl #2]!
+  str	r6, [sp, -r1]!
+  str	r5, [r3], r9
+  str	r4, [r2], -r5
+  str	r3, [r4, -r2, lsl #2]
+  str	r2, [r7], r3, asr #24
+  strb	r9, [r2]
+  strb	r7, [r1, #3]
+  strb	r6, [r4, #405]!
+  strb	r5, [r7], #72
+  strb	r1, [sp], #-1
+  strb	r1, [r2, r9]
+  strb	r2, [r3, -r8]
+  strb	r3, [r4, r7]!
+  strb	r4, [r5, -r6]!
+  strb	r5, [r6], r5
+  strb	r6, [r2], -r4
+  strb	r7, [r12, -r3, lsl #5]
+  strb	sp, [r7], r2, asr #12
+  strbt	r6, [r2], #12
+  strbt	r5, [r6], #-13
+  strbt	r4, [r9], r5
+  strbt	r3, [r8], -r2, lsl #3
+  strd	r0, r1, [r4]
+  strd	r2, r3, [r6, #1]
+  strd	r2, r3, [r7, #22]!
+  strd	r4, r5, [r8], #7
+  strd	r4, r5, [sp], #0
+  strd	r6, r7, [lr], #0
+  strd	r6, r7, [r9], #-0
+  strd	r8, r9, [r4, r1]
+  strd	r6, r7, [r3, r9]!
+  strd	r6, r7, [r5], r8
+  strd	r4, r5, [r12], -r10
+  strh	r3, [r4]
+  strh	r2, [r7, #4]
+  strh	r1, [r8, #64]!
+  strh	r12, [sp], #4
+  strh	r6, [r5, r4]
+  strh	r3, [r8, r11]!
+  strh	r1, [r2, -r1]!
+  strh	r9, [r7], r2
+  strh	r4, [r3], -r2
+  strht	r2, [r5], #76
+  strht	r8, [r1], #-25
+  strht	r5, [r3], r4
+  strht	r6, [r8], -r0
+
+# CHECK:      Instruction Info:
+# CHECK-NEXT: [1]: #uOps
+# CHECK-NEXT: [2]: Latency
+# CHECK-NEXT: [3]: RThroughput
+# CHECK-NEXT: [4]: MayLoad
+# CHECK-NEXT: [5]: MayStore
+# CHECK-NEXT: [6]: HasSideEffects (U)
+
+# CHECK:      [1]    [2]    [3]    [4]    [5]    [6]    Instructions:
+# CHECK-NEXT:  1      4     1.00    *                   ldr	r5, [r7]
+# CHECK-NEXT:  1      4     1.00    *                   ldr	r6, [r3, #63]
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r2, [r4, #4095]!
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r1, [r2], #30
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r3, [r1], #-30
+# CHECK-NEXT:  1      4     1.00    *                   ldr	r3, [r8, r1]
+# CHECK-NEXT:  1      4     1.00    *                   ldr	r2, [r5, -r3]
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r1, [r5, r9]!
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r6, [r7, -r8]!
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r1, [r0, r2, lsr #3]!
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r5, [r9], r2
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r4, [r3], -r6
+# CHECK-NEXT:  1      4     1.00    *                   ldr	r3, [r8, -r2, lsl #15]
+# CHECK-NEXT:  2      4     1.00    *                   ldr	r1, [r5], r3, asr #15
+# CHECK-NEXT:  1      4     1.00    *                   ldrb	r3, [r8]
+# CHECK-NEXT:  1      4     1.00    *                   ldrb	r1, [sp, #63]
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r9, [r3, #4095]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r8, [r1], #22
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r2, [r7], #-19
+# CHECK-NEXT:  1      4     1.00    *                   ldrb	r9, [r8, r5]
+# CHECK-NEXT:  1      4     1.00    *                   ldrb	r1, [r5, -r1]
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r3, [r5, r2]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r6, [r9, -r3]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r2, [r1], r4
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r8, [r4], -r5
+# CHECK-NEXT:  1      4     1.00    *                   ldrb	r7, [r12, -r1, lsl #15]
+# CHECK-NEXT:  2      4     1.00    *                   ldrb	r5, [r2], r9, asr #15
+# CHECK-NEXT:  2      4     1.00    *                   ldrbt	r3, [r1], #4
+# CHECK-NEXT:  2      4     1.00    *                   ldrbt	r2, [r8], #-8
+# CHECK-NEXT:  2      4     1.00    *                   ldrbt	r8, [r7], r6
+# CHECK-NEXT:  2      4     1.00    *                   ldrbt	r1, [r2], -r6, lsl #12
+# CHECK-NEXT:  2      4     2.00    *                   ldrd	r0, r1, [r5]
+# CHECK-NEXT:  2      4     2.00    *                   ldrd	r8, r9, [r2, #15]
+# CHECK-NEXT:  4      5     2.00    *                   ldrd	r2, r3, [r9, #32]!
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r6, r7, [r1], #8
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r2, r3, [r8], #0
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r2, r3, [r8], #0
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r2, r3, [r8], #-0
+# CHECK-NEXT:  2      4     2.00    *                   ldrd	r4, r5, [r1, r3]
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r4, r5, [r7, r2]!
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r0, r1, [r8], r12
+# CHECK-NEXT:  4      4     2.00    *                   ldrd	r0, r1, [r8], -r12
+# CHECK-NEXT:  1      4     1.00    *                   ldrh	r3, [r4]
+# CHECK-NEXT:  1      4     1.00    *                   ldrh	r2, [r7, #4]
+# CHECK-NEXT:  1      4     1.00    *                   ldrh	r1, [r8, #64]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrh	r12, [sp], #4
+# CHECK-NEXT:  1      4     1.00    *                   ldrh	r6, [r5, r4]
+# CHECK-NEXT:  1      4     1.00    *                   ldrh	r3, [r8, r11]!
+# CHECK-NEXT:  1      4     1.00    *                   ldrh	r1, [r2, -r1]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrh	r9, [r7], r2
+# CHECK-NEXT:  2      4     1.00    *                   ldrh	r4, [r3], -r2
+# CHECK-NEXT:  2      4     1.00    *                   ldrht	r9, [r7], #128
+# CHECK-NEXT:  2      4     1.00    *                   ldrht	r4, [r3], #-75
+# CHECK-NEXT:  2      4     1.00    *                   ldrht	r9, [r7], r2
+# CHECK-NEXT:  2      4     1.00    *                   ldrht	r4, [r3], -r2
+# CHECK-NEXT:  1      4     1.00    *                   ldrsb	r3, [r4]
+# CHECK-NEXT:  1      4     1.00    *                   ldrsb	r2, [r7, #17]
+# CHECK-NEXT:  1      4     1.00    *                   ldrsb	r1, [r8, #255]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrsb	r12, [sp], #9
+# CHECK-NEXT:  1      4     1.00    *                   ldrsb	r6, [r5, r4]
+# CHECK-NEXT:  1      4     1.00    *                   ldrsb	r3, [r8, r11]!
+# CHECK-NEXT:  1      4     1.00    *                   ldrsb	r1, [r2, -r1]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrsb	r9, [r7], r2
+# CHECK-NEXT:  2      4     1.00    *                   ldrsb	r4, [r3], -r2
+# CHECK-NEXT:  2      4     1.00    *                   ldrsbt	r5, [r6], #1
+# CHECK-NEXT:  2      4     1.00    *                   ldrsbt	r3, [r8], #-12
+# CHECK-NEXT:  2      4     1.00    *                   ldrsbt	r8, [r9], r5
+# CHECK-NEXT:  2      4     1.00    *                   ldrsbt	r2, [r1], -r4
+# CHECK-NEXT:  1      4     1.00    *                   ldrsh	r5, [r9]
+# CHECK-NEXT:  1      4     1.00    *                   ldrsh	r4, [r5, #7]
+# CHECK-NEXT:  1      4     1.00    *                   ldrsh	r3, [r6, #55]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrsh	r2, [r7], #-9
+# CHECK-NEXT:  1      4     1.00    *                   ldrsh	r3, [r1, r5]
+# CHECK-NEXT:  1      4     1.00    *                   ldrsh	r4, [r6, r1]!
+# CHECK-NEXT:  1      4     1.00    *                   ldrsh	r5, [r3, -r6]!
+# CHECK-NEXT:  2      4     1.00    *                   ldrsh	r6, [r9], r8
+# CHECK-NEXT:  2      4     1.00    *                   ldrsh	r7, [r8], -r3
+# CHECK-NEXT:  2      4     1.00    *                   ldrsht	r5, [r6], #1
+# CHECK-NEXT:  2      4     1.00    *                   ldrsht	r3, [r8], #-12
+# CHECK-NEXT:  2      4     1.00    *                   ldrsht	r8, [r9], r5
+# CHECK-NEXT:  2      4     1.00    *                   ldrsht	r2, [r1], -r4
+# CHECK-NEXT:  16     10    16.00   *                   ldm	r2, {r1, r2, r4, r5, r6}
+# CHECK-NEXT:  16     10    16.00   *                   ldm	r2, {r1, r2, r4, r5, r6}
+# CHECK-NEXT:  16     10    16.00   *                   ldm	r2, {r1, r3, r4, r5, r6}
+# CHECK-NEXT:  16     10    16.00   *                   ldmib	r2, {r1, r2}
+# CHECK-NEXT:  16     10    16.00   *                   ldmdb	r2, {r1, r2}
+# CHECK-NEXT:  16     10    16.00   *                   ldmib	r2, {r1, r3}
+# CHECK-NEXT:  16     10    16.00   *                   ldmib	r2, {r1, r3, r5}
+# CHECK-NEXT:  16     10    16.00   *                   ldmib	r2, {r1, r2, r5}
+# CHECK-NEXT:  16     10    16.00   *                   ldmdbeq	r2, {r1, r2}
+# CHECK-NEXT:  16     10    16.00   *                   ldmibeq	r2, {r1, r3}
+# CHECK-NEXT:  16     10    16.00   *                   ldm	r2, {r0, r1, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, sp, lr, pc}
+# CHECK-NEXT:  16     10    16.00   *                   ldm	r2, {r0, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, sp, lr, pc}
+# CHECK-NEXT:  1      1     1.00           *            str	r8, [r12]
+# CHECK-NEXT:  1      1     1.00           *            str	r7, [r1, #12]
+# CHECK-NEXT:  2      1     1.00           *            str	r3, [r5, #40]!
+# CHECK-NEXT:  2      1     1.00           *            str	r9, [sp], #4095
+# CHECK-NEXT:  2      1     1.00           *            str	r1, [r7], #-128
+# CHECK-NEXT:  1      1     1.00           *            str	r9, [r6, r3]
+# CHECK-NEXT:  1      1     1.00           *            str	r8, [r0, -r2]
+# CHECK-NEXT:  2      1     1.00           *            str	r7, [r1, r6]!
+# CHECK-NEXT:  2      2     1.00           *            str	r7, [r1, r6, lsl #2]!
+# CHECK-NEXT:  2      1     1.00           *            str	r6, [sp, -r1]!
+# CHECK-NEXT:  2      2     1.00           *            str	r5, [r3], r9
+# CHECK-NEXT:  2      2     1.00           *            str	r4, [r2], -r5
+# CHECK-NEXT:  1      1     1.00           *            str	r3, [r4, -r2, lsl #2]
+# CHECK-NEXT:  2      2     1.00           *            str	r2, [r7], r3, asr #24
+# CHECK-NEXT:  1      1     1.00           *            strb	r9, [r2]
+# CHECK-NEXT:  1      1     1.00           *            strb	r7, [r1, #3]
+# CHECK-NEXT:  2      1     1.00           *            strb	r6, [r4, #405]!
+# CHECK-NEXT:  2      1     1.00           *            strb	r5, [r7], #72
+# CHECK-NEXT:  2      1     1.00           *            strb	r1, [sp], #-1
+# CHECK-NEXT:  1      1     1.00           *            strb	r1, [r2, r9]
+# CHECK-NEXT:  1      1     1.00           *            strb	r2, [r3, -r8]
+# CHECK-NEXT:  2      1     1.00           *            strb	r3, [r4, r7]!
+# CHECK-NEXT:  2      1     1.00           *            strb	r4, [r5, -r6]!
+# CHECK-NEXT:  2      2     1.00           *            strb	r5, [r6], r5
+# CHECK-NEXT:  2      2     1.00           *            strb	r6, [r2], -r4
+# CHECK-NEXT:  1      1     1.00           *            strb	r7, [r12, -r3, lsl #5]
+# CHECK-NEXT:  2      2     1.00           *            strb	sp, [r7], r2, asr #12
+# CHECK-NEXT:  2      1     1.00                  U     strbt	r6, [r2], #12
+# CHECK-NEXT:  2      1     1.00                  U     strbt	r5, [r6], #-13
+# CHECK-NEXT:  2      2     1.00                  U     strbt	r4, [r9], r5
+# CHECK-NEXT:  2      2     1.00                  U     strbt	r3, [r8], -r2, lsl #3
+# CHECK-NEXT:  1      1     1.00           *            strd	r0, r1, [r4]
+# CHECK-NEXT:  1      1     1.00           *            strd	r2, r3, [r6, #1]
+# CHECK-NEXT:  2      1     1.00           *            strd	r2, r3, [r7, #22]!
+# CHECK-NEXT:  2      1     1.00           *            strd	r4, r5, [r8], #7
+# CHECK-NEXT:  2      1     1.00           *            strd	r4, r5, [sp], #0
+# CHECK-NEXT:  2      1     1.00           *            strd	r6, r7, [lr], #0
+# CHECK-NEXT:  2      1     1.00           *            strd	r6, r7, [r9], #-0
+# CHECK-NEXT:  1      1     1.00           *            strd	r8, r9, [r4, r1]
+# CHECK-NEXT:  2      1     1.00           *            strd	r6, r7, [r3, r9]!
+# CHECK-NEXT:  2      1     1.00           *            strd	r6, r7, [r5], r8
+# CHECK-NEXT:  2      1     1.00           *            strd	r4, r5, [r12], -r10
+# CHECK-NEXT:  1      1     1.00           *            strh	r3, [r4]
+# CHECK-NEXT:  1      1     1.00           *            strh	r2, [r7, #4]
+# CHECK-NEXT:  2      1     1.00                  U     strh	r1, [r8, #64]!
+# CHECK-NEXT:  2      1     1.00           *            strh	r12, [sp], #4
+# CHECK-NEXT:  1      1     1.00           *            strh	r6, [r5, r4]
+# CHECK-NEXT:  2      1     1.00                  U     strh	r3, [r8, r11]!
+# CHECK-NEXT:  2      1     1.00                  U     strh	r1, [r2, -r1]!
+# CHECK-NEXT:  2      1     1.00           *            strh	r9, [r7], r2
+# CHECK-NEXT:  2      1     1.00           *            strh	r4, [r3], -r2
+# CHECK-NEXT:  2      1     1.00                  U     strht	r2, [r5], #76
+# CHECK-NEXT:  2      1     1.00                  U     strht	r8, [r1], #-25
+# CHECK-NEXT:  2      1     1.00                  U     strht	r5, [r3], r4
+# CHECK-NEXT:  2      1     1.00                  U     strht	r6, [r8], -r0
+
+# CHECK:      Resources:
+# CHECK-NEXT: [0]   - A57UnitB
+# CHECK-NEXT: [1.0] - A57UnitI
+# CHECK-NEXT: [1.1] - A57UnitI
+# CHECK-NEXT: [2]   - A57UnitL
+# CHECK-NEXT: [3]   - A57UnitM
+# CHECK-NEXT: [4]   - A57UnitS
+# CHECK-NEXT: [5]   - A57UnitW
+# CHECK-NEXT: [6]   - A57UnitX
+
+# CHECK:      Resource pressure per iteration:
+# CHECK-NEXT: [0]    [1.0]  [1.1]  [2]    [3]    [4]    [5]    [6]
+# CHECK-NEXT:  -     44.00  44.00  284.00 9.00   55.00   -      -
+
+# CHECK:      Resource pressure by instruction:
+# CHECK-NEXT: [0]    [1.0]  [1.1]  [2]    [3]    [4]    [5]    [6]    Instructions:
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldr	r5, [r7]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldr	r6, [r3, #63]
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r2, [r4, #4095]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r1, [r2], #30
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r3, [r1], #-30
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldr	r3, [r8, r1]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldr	r2, [r5, -r3]
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r1, [r5, r9]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r6, [r7, -r8]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r1, [r0, r2, lsr #3]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r5, [r9], r2
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r4, [r3], -r6
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldr	r3, [r8, -r2, lsl #15]
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldr	r1, [r5], r3, asr #15
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrb	r3, [r8]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrb	r1, [sp, #63]
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r9, [r3, #4095]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r8, [r1], #22
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r2, [r7], #-19
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrb	r9, [r8, r5]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrb	r1, [r5, -r1]
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r3, [r5, r2]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r6, [r9, -r3]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r2, [r1], r4
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r8, [r4], -r5
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrb	r7, [r12, -r1, lsl #15]
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrb	r5, [r2], r9, asr #15
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrbt	r3, [r1], #4
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrbt	r2, [r8], #-8
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrbt	r8, [r7], r6
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrbt	r1, [r2], -r6, lsl #12
+# CHECK-NEXT:  -      -      -     2.00    -      -      -      -     ldrd	r0, r1, [r5]
+# CHECK-NEXT:  -      -      -     2.00    -      -      -      -     ldrd	r8, r9, [r2, #15]
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r2, r3, [r9, #32]!
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r6, r7, [r1], #8
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r2, r3, [r8], #0
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r2, r3, [r8], #0
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r2, r3, [r8], #-0
+# CHECK-NEXT:  -      -      -     2.00    -      -      -      -     ldrd	r4, r5, [r1, r3]
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r4, r5, [r7, r2]!
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r0, r1, [r8], r12
+# CHECK-NEXT:  -     1.00   1.00   2.00    -      -      -      -     ldrd	r0, r1, [r8], -r12
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrh	r3, [r4]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrh	r2, [r7, #4]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrh	r1, [r8, #64]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrh	r12, [sp], #4
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrh	r6, [r5, r4]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrh	r3, [r8, r11]!
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrh	r1, [r2, -r1]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrh	r9, [r7], r2
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrh	r4, [r3], -r2
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrht	r9, [r7], #128
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrht	r4, [r3], #-75
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrht	r9, [r7], r2
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrht	r4, [r3], -r2
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsb	r3, [r4]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsb	r2, [r7, #17]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsb	r1, [r8, #255]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsb	r12, [sp], #9
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsb	r6, [r5, r4]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsb	r3, [r8, r11]!
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsb	r1, [r2, -r1]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsb	r9, [r7], r2
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsb	r4, [r3], -r2
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsbt	r5, [r6], #1
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsbt	r3, [r8], #-12
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsbt	r8, [r9], r5
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsbt	r2, [r1], -r4
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsh	r5, [r9]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsh	r4, [r5, #7]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsh	r3, [r6, #55]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsh	r2, [r7], #-9
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsh	r3, [r1, r5]
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsh	r4, [r6, r1]!
+# CHECK-NEXT:  -      -      -     1.00    -      -      -      -     ldrsh	r5, [r3, -r6]!
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsh	r6, [r9], r8
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsh	r7, [r8], -r3
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsht	r5, [r6], #1
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsht	r3, [r8], #-12
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsht	r8, [r9], r5
+# CHECK-NEXT:  -     0.50   0.50   1.00    -      -      -      -     ldrsht	r2, [r1], -r4
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldm	r2, {r1, r2, r4, r5, r6}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldm	r2, {r1, r2, r4, r5, r6}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldm	r2, {r1, r3, r4, r5, r6}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmib	r2, {r1, r2}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmdb	r2, {r1, r2}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmib	r2, {r1, r3}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmib	r2, {r1, r3, r5}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmib	r2, {r1, r2, r5}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmdbeq	r2, {r1, r2}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldmibeq	r2, {r1, r3}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldm	r2, {r0, r1, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, sp, lr, pc}
+# CHECK-NEXT:  -      -      -     16.00   -      -      -      -     ldm	r2, {r0, r2, r3, r4, r5, r6, r7, r8, r9, r10, r11, r12, sp, lr, pc}
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     str	r8, [r12]
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     str	r7, [r1, #12]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     str	r3, [r5, #40]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     str	r9, [sp], #4095
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     str	r1, [r7], #-128
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     str	r9, [r6, r3]
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     str	r8, [r0, -r2]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     str	r7, [r1, r6]!
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     str	r7, [r1, r6, lsl #2]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     str	r6, [sp, -r1]!
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     str	r5, [r3], r9
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     str	r4, [r2], -r5
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     str	r3, [r4, -r2, lsl #2]
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     str	r2, [r7], r3, asr #24
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strb	r9, [r2]
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strb	r7, [r1, #3]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strb	r6, [r4, #405]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strb	r5, [r7], #72
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strb	r1, [sp], #-1
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strb	r1, [r2, r9]
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strb	r2, [r3, -r8]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strb	r3, [r4, r7]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strb	r4, [r5, -r6]!
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     strb	r5, [r6], r5
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     strb	r6, [r2], -r4
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strb	r7, [r12, -r3, lsl #5]
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     strb	sp, [r7], r2, asr #12
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strbt	r6, [r2], #12
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strbt	r5, [r6], #-13
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     strbt	r4, [r9], r5
+# CHECK-NEXT:  -      -      -      -     1.00   1.00    -      -     strbt	r3, [r8], -r2, lsl #3
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strd	r0, r1, [r4]
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strd	r2, r3, [r6, #1]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r2, r3, [r7, #22]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r4, r5, [r8], #7
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r4, r5, [sp], #0
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r6, r7, [lr], #0
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r6, r7, [r9], #-0
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strd	r8, r9, [r4, r1]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r6, r7, [r3, r9]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r6, r7, [r5], r8
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strd	r4, r5, [r12], -r10
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strh	r3, [r4]
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strh	r2, [r7, #4]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strh	r1, [r8, #64]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strh	r12, [sp], #4
+# CHECK-NEXT:  -      -      -      -      -     1.00    -      -     strh	r6, [r5, r4]
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strh	r3, [r8, r11]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strh	r1, [r2, -r1]!
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strh	r9, [r7], r2
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strh	r4, [r3], -r2
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strht	r2, [r5], #76
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strht	r8, [r1], #-25
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strht	r5, [r3], r4
+# CHECK-NEXT:  -     0.50   0.50    -      -     1.00    -      -     strht	r6, [r8], -r0
diff --git a/llvm/test/tools/llvm-symbolizer/input-base.test b/llvm/test/tools/llvm-symbolizer/input-base.test
index f4b641ea626f2..66244a7203c08 100644
--- a/llvm/test/tools/llvm-symbolizer/input-base.test
+++ b/llvm/test/tools/llvm-symbolizer/input-base.test
@@ -1,34 +1,33 @@
-# RUN: llvm-mc %s -o %t -filetype=obj
-## llvm-symbolizer infers the number base from the form of the address.
-# RUN: llvm-symbolizer -e %t -a 0x1234 | FileCheck %s
-# RUN: llvm-symbolizer -e %t -a 0X1234 | FileCheck %s
-# RUN: llvm-symbolizer -e %t -a 4660 | FileCheck %s
-# RUN: llvm-symbolizer -e %t -a 011064 | FileCheck %s
-# RUN: llvm-symbolizer -e %t -a 0b1001000110100 | FileCheck %s
-# RUN: llvm-symbolizer -e %t -a 0B1001000110100 | FileCheck %s
-# RUN: llvm-symbolizer -e %t -a 0o11064 | FileCheck %s
+# llvm-symbolizer infers the number base from the form of the address.
+RUN: llvm-symbolizer -e /dev/null -a 0x1234 | FileCheck %s
+RUN: llvm-symbolizer -e /dev/null -a 0X1234 | FileCheck %s
+RUN: llvm-symbolizer -e /dev/null -a 4660 | FileCheck %s
+RUN: llvm-symbolizer -e /dev/null -a 011064 | FileCheck %s
+RUN: llvm-symbolizer -e /dev/null -a 0b1001000110100 | FileCheck %s
+RUN: llvm-symbolizer -e /dev/null -a 0B1001000110100 | FileCheck %s
+RUN: llvm-symbolizer -e /dev/null -a 0o11064 | FileCheck %s
 
-## llvm-symbolizer / StringRef::getAsInteger only accepts the 0o prefix in lowercase.
-# RUN: llvm-symbolizer -e %t -a 0O1234 | FileCheck %s --check-prefix=INVALID-NOT-OCTAL-UPPER
+# llvm-symbolizer / StringRef::getAsInteger only accepts the 0o prefix in lowercase.
+RUN: llvm-symbolizer -e /dev/null -a 0O1234 | FileCheck %s --check-prefix=INVALID-NOT-OCTAL-UPPER
 
-## llvm-addr2line always requires hexadecimal, but accepts an optional 0x prefix.
-# RUN: llvm-addr2line -e %t -a 0x1234 | FileCheck %s
-# RUN: llvm-addr2line -e %t -a 0X1234 | FileCheck %s
-# RUN: llvm-addr2line -e %t -a 1234 | FileCheck %s
-# RUN: llvm-addr2line -e %t -a 01234 | FileCheck %s
-# RUN: llvm-addr2line -e %t -a 0b1010 | FileCheck %s --check-prefix=HEXADECIMAL-NOT-BINARY
-# RUN: llvm-addr2line -e %t -a 0B1010 | FileCheck %s --check-prefix=HEXADECIMAL-NOT-BINARY
-# RUN: llvm-addr2line -e %t -a 0o1234 | FileCheck %s --check-prefix=INVALID-NOT-OCTAL-LOWER
-# RUN: llvm-addr2line -e %t -a 0O1234 | FileCheck %s --check-prefix=INVALID-NOT-OCTAL-UPPER
+# llvm-addr2line always requires hexadecimal, but accepts an optional 0x prefix.
+RUN: llvm-addr2line -e /dev/null -a 0x1234 | FileCheck %s
+RUN: llvm-addr2line -e /dev/null -a 0X1234 | FileCheck %s
+RUN: llvm-addr2line -e /dev/null -a 1234 | FileCheck %s
+RUN: llvm-addr2line -e /dev/null -a 01234 | FileCheck %s
+RUN: llvm-addr2line -e /dev/null -a 0b1010 | FileCheck %s --check-prefix=HEXADECIMAL-NOT-BINARY
+RUN: llvm-addr2line -e /dev/null -a 0B1010 | FileCheck %s --check-prefix=HEXADECIMAL-NOT-BINARY
+RUN: llvm-addr2line -e /dev/null -a 0o1234 | FileCheck %s --check-prefix=INVALID-NOT-OCTAL-LOWER
+RUN: llvm-addr2line -e /dev/null -a 0O1234 | FileCheck %s --check-prefix=INVALID-NOT-OCTAL-UPPER
 
-# CHECK: 0x1234
-# CHECK-NEXT: ??
+CHECK: 0x1234
+CHECK-NEXT: ??
 
-# HEXADECIMAL-NOT-BINARY: 0xb1010
-# HEXADECIMAL-NOT-BINARY: ??
+HEXADECIMAL-NOT-BINARY: 0xb1010
+HEXADECIMAL-NOT-BINARY: ??
 
-# INVALID-NOT-OCTAL-LOWER: 0o1234
-# INVALID-NOT-OCTAL-LOWER-NOT: ??
+INVALID-NOT-OCTAL-LOWER: 0o1234
+INVALID-NOT-OCTAL-LOWER-NOT: ??
 
-# INVALID-NOT-OCTAL-UPPER: 0O1234
-# INVALID-NOT-OCTAL-UPPER-NOT: ??
+INVALID-NOT-OCTAL-UPPER: 0O1234
+INVALID-NOT-OCTAL-UPPER-NOT: ??
diff --git a/llvm/test/tools/llvm-symbolizer/invalid-dwarf.s b/llvm/test/tools/llvm-symbolizer/invalid-dwarf.s
deleted file mode 100644
index 027ccef00faac..0000000000000
--- a/llvm/test/tools/llvm-symbolizer/invalid-dwarf.s
+++ /dev/null
@@ -1,205 +0,0 @@
-# REQUIRES: x86-registered-target
-# Source:
-# void f1() { }
-# void f2() { }
-#
-# Build as: clang -ffunction-sections -gdwarf-5 -c test.c
-# Hand modify the rnglist to include an invalid RLE encoding (42)
-#
-# RUN: llvm-mc -dwarf-version=5 %s -triple=x86_64-pc-linux -filetype=obj -o %t
-# RUN: not llvm-symbolizer -obj=%t 0x0 2>&1 | FileCheck %s
-#
-# CHECK: error: decoding address ranges: unknown rnglists encoding 0x2a at offset 0x10
-
-	.text
-	.file	"test.c"
-	.section	.text.f1,"ax",@progbits
-	.globl	f1                              # -- Begin function f1
-	.p2align	4, 0x90
-	.type	f1,@function
-f1:                                     # @f1
-.Lfunc_begin0:
-	.file	0 "/usr/local/google/home/blaikie/dev/scratch" "test.c" md5 0x39c2464ceaf7fd68a00d44ca40e99028
-	.loc	0 1 0                           # test.c:1:0
-	.cfi_startproc
-# %bb.0:                                # %entry
-	pushq	%rbp
-	.cfi_def_cfa_offset 16
-	.cfi_offset %rbp, -16
-	movq	%rsp, %rbp
-	.cfi_def_cfa_register %rbp
-.Ltmp0:
-	.loc	0 1 13 prologue_end             # test.c:1:13
-	popq	%rbp
-	.cfi_def_cfa %rsp, 8
-	retq
-.Ltmp1:
-.Lfunc_end0:
-	.size	f1, .Lfunc_end0-f1
-	.cfi_endproc
-                                        # -- End function
-	.section	.text.f2,"ax",@progbits
-	.globl	f2                              # -- Begin function f2
-	.p2align	4, 0x90
-	.type	f2,@function
-f2:                                     # @f2
-.Lfunc_begin1:
-	.loc	0 2 0                           # test.c:2:0
-	.cfi_startproc
-# %bb.0:                                # %entry
-	pushq	%rbp
-	.cfi_def_cfa_offset 16
-	.cfi_offset %rbp, -16
-	movq	%rsp, %rbp
-	.cfi_def_cfa_register %rbp
-.Ltmp2:
-	.loc	0 2 13 prologue_end             # test.c:2:13
-	popq	%rbp
-	.cfi_def_cfa %rsp, 8
-	retq
-.Ltmp3:
-.Lfunc_end1:
-	.size	f2, .Lfunc_end1-f2
-	.cfi_endproc
-                                        # -- End function
-	.section	.debug_abbrev,"",@progbits
-	.byte	1                               # Abbreviation Code
-	.byte	17                              # DW_TAG_compile_unit
-	.byte	1                               # DW_CHILDREN_yes
-	.byte	37                              # DW_AT_producer
-	.byte	37                              # DW_FORM_strx1
-	.byte	19                              # DW_AT_language
-	.byte	5                               # DW_FORM_data2
-	.byte	3                               # DW_AT_name
-	.byte	37                              # DW_FORM_strx1
-	.byte	114                             # DW_AT_str_offsets_base
-	.byte	23                              # DW_FORM_sec_offset
-	.byte	16                              # DW_AT_stmt_list
-	.byte	23                              # DW_FORM_sec_offset
-	.byte	27                              # DW_AT_comp_dir
-	.byte	37                              # DW_FORM_strx1
-	.byte	17                              # DW_AT_low_pc
-	.byte	1                               # DW_FORM_addr
-	.byte	85                              # DW_AT_ranges
-	.byte	35                              # DW_FORM_rnglistx
-	.byte	115                             # DW_AT_addr_base
-	.byte	23                              # DW_FORM_sec_offset
-	.byte	116                             # DW_AT_rnglists_base
-	.byte	23                              # DW_FORM_sec_offset
-	.byte	0                               # EOM(1)
-	.byte	0                               # EOM(2)
-	.byte	2                               # Abbreviation Code
-	.byte	46                              # DW_TAG_subprogram
-	.byte	0                               # DW_CHILDREN_no
-	.byte	17                              # DW_AT_low_pc
-	.byte	27                              # DW_FORM_addrx
-	.byte	18                              # DW_AT_high_pc
-	.byte	6                               # DW_FORM_data4
-	.byte	64                              # DW_AT_frame_base
-	.byte	24                              # DW_FORM_exprloc
-	.byte	3                               # DW_AT_name
-	.byte	37                              # DW_FORM_strx1
-	.byte	58                              # DW_AT_decl_file
-	.byte	11                              # DW_FORM_data1
-	.byte	59                              # DW_AT_decl_line
-	.byte	11                              # DW_FORM_data1
-	.byte	63                              # DW_AT_external
-	.byte	25                              # DW_FORM_flag_present
-	.byte	0                               # EOM(1)
-	.byte	0                               # EOM(2)
-	.byte	0                               # EOM(3)
-	.section	.debug_info,"",@progbits
-.Lcu_begin0:
-	.long	.Ldebug_info_end0-.Ldebug_info_start0 # Length of Unit
-.Ldebug_info_start0:
-	.short	5                               # DWARF version number
-	.byte	1                               # DWARF Unit Type
-	.byte	8                               # Address Size (in bytes)
-	.long	.debug_abbrev                   # Offset Into Abbrev. Section
-	.byte	1                               # Abbrev [1] 0xc:0x36 DW_TAG_compile_unit
-	.byte	0                               # DW_AT_producer
-	.short	12                              # DW_AT_language
-	.byte	1                               # DW_AT_name
-	.long	.Lstr_offsets_base0             # DW_AT_str_offsets_base
-	.long	.Lline_table_start0             # DW_AT_stmt_list
-	.byte	2                               # DW_AT_comp_dir
-	.quad	0                               # DW_AT_low_pc
-	.byte	0                               # DW_AT_ranges
-	.long	.Laddr_table_base0              # DW_AT_addr_base
-	.long	.Lrnglists_table_base0          # DW_AT_rnglists_base
-	.byte	2                               # Abbrev [2] 0x2b:0xb DW_TAG_subprogram
-	.byte	0                               # DW_AT_low_pc
-	.long	.Lfunc_end0-.Lfunc_begin0       # DW_AT_high_pc
-	.byte	1                               # DW_AT_frame_base
-	.byte	86
-	.byte	3                               # DW_AT_name
-	.byte	0                               # DW_AT_decl_file
-	.byte	1                               # DW_AT_decl_line
-                                        # DW_AT_external
-	.byte	2                               # Abbrev [2] 0x36:0xb DW_TAG_subprogram
-	.byte	1                               # DW_AT_low_pc
-	.long	.Lfunc_end1-.Lfunc_begin1       # DW_AT_high_pc
-	.byte	1                               # DW_AT_frame_base
-	.byte	86
-	.byte	4                               # DW_AT_name
-	.byte	0                               # DW_AT_decl_file
-	.byte	2                               # DW_AT_decl_line
-                                        # DW_AT_external
-	.byte	0                               # End Of Children Mark
-.Ldebug_info_end0:
-	.section	.debug_rnglists,"",@progbits
-	.long	.Ldebug_list_header_end0-.Ldebug_list_header_start0 # Length
-.Ldebug_list_header_start0:
-	.short	5                               # Version
-	.byte	8                               # Address size
-	.byte	0                               # Segment selector size
-	.long	1                               # Offset entry count
-.Lrnglists_table_base0:
-	.long	.Ldebug_ranges0-.Lrnglists_table_base0
-.Ldebug_ranges0:
-	.byte	42                               # DW_RLE_startx_length
-	.byte	0                               #   start index
-	.uleb128 .Lfunc_end0-.Lfunc_begin0      #   length
-	.byte	3                               # DW_RLE_startx_length
-	.byte	1                               #   start index
-	.uleb128 .Lfunc_end1-.Lfunc_begin1      #   length
-	.byte	0                               # DW_RLE_end_of_list
-.Ldebug_list_header_end0:
-	.section	.debug_str_offsets,"",@progbits
-	.long	24                              # Length of String Offsets Set
-	.short	5
-	.short	0
-.Lstr_offsets_base0:
-	.section	.debug_str,"MS",@progbits,1
-.Linfo_string0:
-	.asciz	"clang version 12.0.0 (git@github.com:llvm/llvm-project.git 9a33f027ac7d73e14ae287e78ab554142d1cbc8f)" # string offset=0
-.Linfo_string1:
-	.asciz	"test.c"                        # string offset=101
-.Linfo_string2:
-	.asciz	"/usr/local/google/home/blaikie/dev/scratch" # string offset=108
-.Linfo_string3:
-	.asciz	"f1"                            # string offset=151
-.Linfo_string4:
-	.asciz	"f2"                            # string offset=154
-	.section	.debug_str_offsets,"",@progbits
-	.long	.Linfo_string0
-	.long	.Linfo_string1
-	.long	.Linfo_string2
-	.long	.Linfo_string3
-	.long	.Linfo_string4
-	.section	.debug_addr,"",@progbits
-	.long	.Ldebug_addr_end0-.Ldebug_addr_start0 # Length of contribution
-.Ldebug_addr_start0:
-	.short	5                               # DWARF version number
-	.byte	8                               # Address size
-	.byte	0                               # Segment selector size
-.Laddr_table_base0:
-	.quad	.Lfunc_begin0
-	.quad	.Lfunc_begin1
-.Ldebug_addr_end0:
-	.ident	"clang version 12.0.0 (git@github.com:llvm/llvm-project.git 9a33f027ac7d73e14ae287e78ab554142d1cbc8f)"
-	.section	".note.GNU-stack","",@progbits
-	.addrsig
-	.section	.debug_line,"",@progbits
-.Lline_table_start0:
-
diff --git a/llvm/test/tools/llvm-symbolizer/pdb/missing_pdb.test b/llvm/test/tools/llvm-symbolizer/pdb/missing_pdb.test
index 7e7fa79c4b382..b83b46a752133 100644
--- a/llvm/test/tools/llvm-symbolizer/pdb/missing_pdb.test
+++ b/llvm/test/tools/llvm-symbolizer/pdb/missing_pdb.test
@@ -1,4 +1,4 @@
-RUN: not llvm-symbolizer 0x401000 0x401001 -obj="%p/Inputs/missing_pdb.exe" 2>%t.err \
+RUN: llvm-symbolizer 0x401000 0x401001 -obj="%p/Inputs/missing_pdb.exe" 2>%t.err \
 RUN:    | FileCheck %s
 RUN: FileCheck --check-prefix=ERROR %s < %t.err
 
diff --git a/llvm/test/tools/llvm-symbolizer/sym.test b/llvm/test/tools/llvm-symbolizer/sym.test
index 612485b8b1fe6..73097b2c4b9b6 100644
--- a/llvm/test/tools/llvm-symbolizer/sym.test
+++ b/llvm/test/tools/llvm-symbolizer/sym.test
@@ -31,8 +31,7 @@ RUN: llvm-symbolizer -i -print-address -p -obj=%p/Inputs/addr.exe < %p/Inputs/ad
 RUN: llvm-symbolizer --inlining=true --print-address -p --obj=%p/Inputs/addr.exe < %p/Inputs/addr.inp | FileCheck -check-prefix="PRETTY" %s
 
 RUN: echo "0x1" > %t.input
-## FIXME: Looks like this was meant to have an input file to test, but it isn't present - so this test probably isn't testing what it's intended to test
-RUN: not llvm-symbolizer -obj=%p/Inputs/zero < %t.input | FileCheck -check-prefix="ZERO" %s
+RUN: llvm-symbolizer -obj=%p/Inputs/zero < %t.input | FileCheck -check-prefix="ZERO" %s
 
 RUN: llvm-addr2line -obj=%p/Inputs/addr.exe < %p/Inputs/addr.inp | FileCheck -check-prefix=A2L %s
 RUN: llvm-addr2line -a -obj=%p/Inputs/addr.exe < %p/Inputs/addr.inp | FileCheck -check-prefixes=A2L,A2L_A %s
diff --git a/llvm/tools/llvm-symbolizer/llvm-symbolizer.cpp b/llvm/tools/llvm-symbolizer/llvm-symbolizer.cpp
index e4061408a821d..f57922b2c506e 100644
--- a/llvm/tools/llvm-symbolizer/llvm-symbolizer.cpp
+++ b/llvm/tools/llvm-symbolizer/llvm-symbolizer.cpp
@@ -73,15 +73,12 @@ static cl::list<std::string> ClInputAddresses(cl::Positional,
                                               cl::desc("<input addresses>..."),
                                               cl::ZeroOrMore);
 
-static bool HasError = false;
-
 template<typename T>
 static bool error(Expected<T> &ResOrErr) {
   if (ResOrErr)
     return false;
   logAllUnhandledErrors(ResOrErr.takeError(), errs(),
                         "LLVMSymbolizer: error reading file: ");
-  HasError = true;
   return true;
 }
 
@@ -293,10 +290,6 @@ int main(int argc, char **argv) {
       Args.hasFlag(OPT_untag_addresses, OPT_no_untag_addresses, !IsAddr2Line);
   Opts.UseNativePDBReader = Args.hasArg(OPT_use_native_pdb_reader);
   Opts.UseSymbolTable = true;
-  Opts.RecoverableErrorHandler = [&](Error E) {
-    HasError = true;
-    WithColor::defaultErrorHandler(std::move(E));
-  };
 
   for (const opt::Arg *A : Args.filtered(OPT_dsym_hint_EQ)) {
     StringRef Hint(A->getValue());
@@ -343,5 +336,5 @@ int main(int argc, char **argv) {
                      Symbolizer, Printer);
   }
 
-  return HasError;
+  return 0;
 }
diff --git a/llvm/tools/opt/opt.cpp b/llvm/tools/opt/opt.cpp
index 3be5c8160f8eb..7a225b388c91e 100644
--- a/llvm/tools/opt/opt.cpp
+++ b/llvm/tools/opt/opt.cpp
@@ -757,9 +757,13 @@ int main(int argc, char **argv) {
   // but `-enable-new-pm -codegenprepare` will still revert to legacy PM.
   if ((EnableNewPassManager && !CodegenPassSpecifiedInPassList()) ||
       PassPipeline.getNumOccurrences() > 0) {
+    if (AnalyzeOnly) {
+      errs() << "Cannot specify -analyze under new pass manager\n";
+      return 1;
+    }
     if (PassPipeline.getNumOccurrences() > 0 && PassList.size() > 0) {
       errs()
-          << "Cannot specify passes via both -foo-pass and --passes=foo-pass";
+          << "Cannot specify passes via both -foo-pass and --passes=foo-pass\n";
       return 1;
     }
     SmallVector<StringRef, 4> Passes;
diff --git a/llvm/unittests/ADT/ImmutableMapTest.cpp b/llvm/unittests/ADT/ImmutableMapTest.cpp
index fa61816d213cf..4541ea94742a9 100644
--- a/llvm/unittests/ADT/ImmutableMapTest.cpp
+++ b/llvm/unittests/ADT/ImmutableMapTest.cpp
@@ -46,4 +46,47 @@ TEST(ImmutableMapTest, MultiElemIntMapTest) {
   EXPECT_EQ(3U, S2.getHeight());
 }
 
+TEST(ImmutableMapTest, EmptyIntMapRefTest) {
+  using int_int_map = ImmutableMapRef<int, int>;
+  ImmutableMapRef<int, int>::FactoryTy *f =
+    new ImmutableMapRef<int, int>::FactoryTy();
+
+  EXPECT_TRUE(int_int_map::getEmptyMap(f) == int_int_map::getEmptyMap(f));
+  EXPECT_FALSE(int_int_map::getEmptyMap(f) != int_int_map::getEmptyMap(f));
+  EXPECT_TRUE(int_int_map::getEmptyMap(f).isEmpty());
+
+  int_int_map S = int_int_map::getEmptyMap(f);
+  EXPECT_EQ(0u, S.getHeight());
+  EXPECT_TRUE(S.begin() == S.end());
+  EXPECT_FALSE(S.begin() != S.end());
+}
+
+TEST(ImmutableMapTest, MultiElemIntMapRefTest) {
+  ImmutableMapRef<int, int>::FactoryTy *f =
+    new ImmutableMapRef<int, int>::FactoryTy();
+
+  ImmutableMapRef<int, int> S = ImmutableMapRef<int, int>::getEmptyMap(f);
+
+  ImmutableMapRef<int, int> S2 = S.add(3, 10).add(4, 11).add(5, 12);
+
+  EXPECT_TRUE(S.isEmpty());
+  EXPECT_FALSE(S2.isEmpty());
+
+  EXPECT_EQ(nullptr, S.lookup(3));
+  EXPECT_EQ(nullptr, S.lookup(9));
+
+  EXPECT_EQ(10, *S2.lookup(3));
+  EXPECT_EQ(11, *S2.lookup(4));
+  EXPECT_EQ(12, *S2.lookup(5));
+
+  EXPECT_EQ(5, S2.getMaxElement()->first);
+  EXPECT_EQ(3U, S2.getHeight());
+}
+
+  TEST(ImmutableMapTest, MapOfMapRefsTest) {
+  ImmutableMap<int, ImmutableMapRef<int, int>>::Factory f;
+
+  EXPECT_TRUE(f.getEmptyMap() == f.getEmptyMap());
+  }
+
 }
diff --git a/llvm/unittests/Analysis/ValueTrackingTest.cpp b/llvm/unittests/Analysis/ValueTrackingTest.cpp
index f2cafa6aa510b..5b5c3a36677ed 100644
--- a/llvm/unittests/Analysis/ValueTrackingTest.cpp
+++ b/llvm/unittests/Analysis/ValueTrackingTest.cpp
@@ -1208,6 +1208,66 @@ TEST_F(ComputeKnownBitsTest, ComputeKnownBitsAddWithRangeNoOverlap) {
   EXPECT_EQ(Known.getMaxValue(), 575);
 }
 
+TEST_F(ComputeKnownBitsTest, ComputeKnownBitsGEPWithRange) {
+  parseAssembly(
+      "define void @test(i64* %p) {\n"
+      "  %A = load i64, i64* %p, !range !{i64 64, i64 65536}\n"
+      "  %APtr = inttoptr i64 %A to float*"
+      "  %APtrPlus512 = getelementptr float, float* %APtr, i32 128\n"
+      "  %c = icmp ugt float* %APtrPlus512, inttoptr (i32 523 to float*)\n"
+      "  call void @llvm.assume(i1 %c)\n"
+      "  ret void\n"
+      "}\n"
+      "declare void @llvm.assume(i1)\n");
+  AssumptionCache AC(*F);
+  KnownBits Known = computeKnownBits(A, M->getDataLayout(), /* Depth */ 0, &AC,
+                                     F->front().getTerminator());
+  EXPECT_EQ(Known.Zero.getZExtValue(), ~(65536llu - 1));
+  EXPECT_EQ(Known.One.getZExtValue(), 0u);
+  Instruction &APtrPlus512 = findInstructionByName(F, "APtrPlus512");
+  Known = computeKnownBits(&APtrPlus512, M->getDataLayout(), /* Depth */ 0, &AC,
+                           F->front().getTerminator());
+  // We know of one less zero because 512 may have produced a 1 that
+  // got carried all the way to the first trailing zero.
+  EXPECT_EQ(Known.Zero.getZExtValue(), ~(65536llu - 1) << 1);
+  EXPECT_EQ(Known.One.getZExtValue(), 0u);
+  // The known range is not precise given computeKnownBits works
+  // with the masks of zeros and ones, not the ranges.
+  EXPECT_EQ(Known.getMinValue(), 0u);
+  EXPECT_EQ(Known.getMaxValue(), 131071);
+}
+
+// 4*128 + [32, 64) doesn't produce overlapping bits.
+// Make sure we get all the individual bits properly.
+// This test is useful to check that we account for the scaling factor
+// in the gep. Indeed, gep float, [32,64), 128 is not 128 + [32,64).
+TEST_F(ComputeKnownBitsTest, ComputeKnownBitsGEPWithRangeNoOverlap) {
+  parseAssembly(
+      "define void @test(i64* %p) {\n"
+      "  %A = load i64, i64* %p, !range !{i64 32, i64 64}\n"
+      "  %APtr = inttoptr i64 %A to float*"
+      "  %APtrPlus512 = getelementptr float, float* %APtr, i32 128\n"
+      "  %c = icmp ugt float* %APtrPlus512, inttoptr (i32 523 to float*)\n"
+      "  call void @llvm.assume(i1 %c)\n"
+      "  ret void\n"
+      "}\n"
+      "declare void @llvm.assume(i1)\n");
+  AssumptionCache AC(*F);
+  KnownBits Known = computeKnownBits(A, M->getDataLayout(), /* Depth */ 0, &AC,
+                                     F->front().getTerminator());
+  EXPECT_EQ(Known.Zero.getZExtValue(), ~(64llu - 1));
+  EXPECT_EQ(Known.One.getZExtValue(), 32u);
+  Instruction &APtrPlus512 = findInstructionByName(F, "APtrPlus512");
+  Known = computeKnownBits(&APtrPlus512, M->getDataLayout(), /* Depth */ 0, &AC,
+                           F->front().getTerminator());
+  EXPECT_EQ(Known.Zero.getZExtValue(), ~512llu & ~(64llu - 1));
+  EXPECT_EQ(Known.One.getZExtValue(), 512u | 32u);
+  // The known range is not precise given computeKnownBits works
+  // with the masks of zeros and ones, not the ranges.
+  EXPECT_EQ(Known.getMinValue(), 544);
+  EXPECT_EQ(Known.getMaxValue(), 575);
+}
+
 class IsBytewiseValueTest : public ValueTrackingTest,
                             public ::testing::WithParamInterface<
                                 std::pair<const char *, const char *>> {
diff --git a/llvm/unittests/CodeGen/ScalableVectorMVTsTest.cpp b/llvm/unittests/CodeGen/ScalableVectorMVTsTest.cpp
index 48b950fa74e9f..566b2e75c70c5 100644
--- a/llvm/unittests/CodeGen/ScalableVectorMVTsTest.cpp
+++ b/llvm/unittests/CodeGen/ScalableVectorMVTsTest.cpp
@@ -139,19 +139,23 @@ TEST(ScalableVectorMVTsTest, SizeQueries) {
   EXPECT_EQ(nxv4i32.getSizeInBits(), nxv2i64.getSizeInBits());
   EXPECT_EQ(nxv2f64.getSizeInBits(), nxv2i64.getSizeInBits());
   EXPECT_NE(nxv2i32.getSizeInBits(), nxv4i32.getSizeInBits());
-  EXPECT_LT(nxv2i32.getSizeInBits(), nxv2i64.getSizeInBits());
-  EXPECT_LE(nxv4i32.getSizeInBits(), nxv2i64.getSizeInBits());
-  EXPECT_GT(nxv4i32.getSizeInBits(), nxv2i32.getSizeInBits());
-  EXPECT_GE(nxv2i64.getSizeInBits(), nxv4i32.getSizeInBits());
+  EXPECT_LT(nxv2i32.getSizeInBits().getKnownMinSize(),
+            nxv2i64.getSizeInBits().getKnownMinSize());
+  EXPECT_LE(nxv4i32.getSizeInBits().getKnownMinSize(),
+            nxv2i64.getSizeInBits().getKnownMinSize());
+  EXPECT_GT(nxv4i32.getSizeInBits().getKnownMinSize(),
+            nxv2i32.getSizeInBits().getKnownMinSize());
+  EXPECT_GE(nxv2i64.getSizeInBits().getKnownMinSize(),
+            nxv4i32.getSizeInBits().getKnownMinSize());
 
   // Check equivalence and ordering on fixed types.
   EXPECT_EQ(v4i32.getSizeInBits(), v2i64.getSizeInBits());
   EXPECT_EQ(v2f64.getSizeInBits(), v2i64.getSizeInBits());
   EXPECT_NE(v2i32.getSizeInBits(), v4i32.getSizeInBits());
-  EXPECT_LT(v2i32.getSizeInBits(), v2i64.getSizeInBits());
-  EXPECT_LE(v4i32.getSizeInBits(), v2i64.getSizeInBits());
-  EXPECT_GT(v4i32.getSizeInBits(), v2i32.getSizeInBits());
-  EXPECT_GE(v2i64.getSizeInBits(), v4i32.getSizeInBits());
+  EXPECT_LT(v2i32.getFixedSizeInBits(), v2i64.getFixedSizeInBits());
+  EXPECT_LE(v4i32.getFixedSizeInBits(), v2i64.getFixedSizeInBits());
+  EXPECT_GT(v4i32.getFixedSizeInBits(), v2i32.getFixedSizeInBits());
+  EXPECT_GE(v2i64.getFixedSizeInBits(), v4i32.getFixedSizeInBits());
 
   // Check that scalable and non-scalable types with the same minimum size
   // are not considered equal.
@@ -159,7 +163,7 @@ TEST(ScalableVectorMVTsTest, SizeQueries) {
   ASSERT_FALSE(v2i64.getSizeInBits() == nxv2f64.getSizeInBits());
 
   // Check that we can obtain a known-exact size from a non-scalable type.
-  EXPECT_EQ(v4i32.getSizeInBits(), 128U);
+  EXPECT_EQ(v4i32.getFixedSizeInBits(), 128U);
   EXPECT_EQ(v2i64.getFixedSizeInBits(), 128U);
 
   // Check that we can query the known minimum size for both scalable and
diff --git a/llvm/unittests/IR/VectorTypesTest.cpp b/llvm/unittests/IR/VectorTypesTest.cpp
index 37137053bf157..c8a09808428fa 100644
--- a/llvm/unittests/IR/VectorTypesTest.cpp
+++ b/llvm/unittests/IR/VectorTypesTest.cpp
@@ -286,10 +286,10 @@ TEST(VectorTypesTest, FixedLenComparisons) {
   EXPECT_EQ(V2I32Len.getKnownMinSize(), 64U);
   EXPECT_FALSE(V2I32Len.isScalable());
 
-  EXPECT_LT(V2Int32Ty->getPrimitiveSizeInBits(),
-            V4Int32Ty->getPrimitiveSizeInBits());
-  EXPECT_GT(V2Int64Ty->getPrimitiveSizeInBits(),
-            V2Int32Ty->getPrimitiveSizeInBits());
+  EXPECT_LT(V2Int32Ty->getPrimitiveSizeInBits().getFixedSize(),
+            V4Int32Ty->getPrimitiveSizeInBits().getFixedSize());
+  EXPECT_GT(V2Int64Ty->getPrimitiveSizeInBits().getFixedSize(),
+            V2Int32Ty->getPrimitiveSizeInBits().getFixedSize());
   EXPECT_EQ(V4Int32Ty->getPrimitiveSizeInBits(),
             V2Int64Ty->getPrimitiveSizeInBits());
   EXPECT_NE(V2Int32Ty->getPrimitiveSizeInBits(),
@@ -332,14 +332,14 @@ TEST(VectorTypesTest, ScalableComparisons) {
   EXPECT_EQ(ScV2I32Len.getKnownMinSize(), 64U);
   EXPECT_TRUE(ScV2I32Len.isScalable());
 
-  EXPECT_LT(ScV2Int32Ty->getPrimitiveSizeInBits(),
-            ScV4Int32Ty->getPrimitiveSizeInBits());
-  EXPECT_GT(ScV2Int64Ty->getPrimitiveSizeInBits(),
-            ScV2Int32Ty->getPrimitiveSizeInBits());
-  EXPECT_EQ(ScV4Int32Ty->getPrimitiveSizeInBits(),
-            ScV2Int64Ty->getPrimitiveSizeInBits());
-  EXPECT_NE(ScV2Int32Ty->getPrimitiveSizeInBits(),
-            ScV2Int64Ty->getPrimitiveSizeInBits());
+  EXPECT_LT(ScV2Int32Ty->getPrimitiveSizeInBits().getKnownMinSize(),
+            ScV4Int32Ty->getPrimitiveSizeInBits().getKnownMinSize());
+  EXPECT_GT(ScV2Int64Ty->getPrimitiveSizeInBits().getKnownMinSize(),
+            ScV2Int32Ty->getPrimitiveSizeInBits().getKnownMinSize());
+  EXPECT_EQ(ScV4Int32Ty->getPrimitiveSizeInBits().getKnownMinSize(),
+            ScV2Int64Ty->getPrimitiveSizeInBits().getKnownMinSize());
+  EXPECT_NE(ScV2Int32Ty->getPrimitiveSizeInBits().getKnownMinSize(),
+            ScV2Int64Ty->getPrimitiveSizeInBits().getKnownMinSize());
 
   // Check the DataLayout interfaces.
   EXPECT_EQ(DL.getTypeSizeInBits(ScV2Int64Ty),
diff --git a/llvm/utils/gn/secondary/clang/lib/Headers/BUILD.gn b/llvm/utils/gn/secondary/clang/lib/Headers/BUILD.gn
index 5704470e30105..8509ccfafd178 100644
--- a/llvm/utils/gn/secondary/clang/lib/Headers/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/lib/Headers/BUILD.gn
@@ -195,6 +195,7 @@ copy("Headers") {
     "tgmath.h",
     "tmmintrin.h",
     "tsxldtrkintrin.h",
+    "uintrintrin.h",
     "unwind.h",
     "vadefs.h",
     "vaesintrin.h",
diff --git a/llvm/utils/gn/secondary/clang/tools/driver/BUILD.gn b/llvm/utils/gn/secondary/clang/tools/driver/BUILD.gn
index a91ee25224d85..24a95e1968bec 100644
--- a/llvm/utils/gn/secondary/clang/tools/driver/BUILD.gn
+++ b/llvm/utils/gn/secondary/clang/tools/driver/BUILD.gn
@@ -86,7 +86,7 @@ executable("clang") {
     # clang.  This is different from the CMake build, which requires devs to
     # explicitly build the "libcxx" target (which also needlessly compiles the
     # libcxx sources) to get a working compiler.
-    deps += [ "//libcxx/include" ]
+    deps += [ "//libcxx/include:copy_headers" ]
   }
   sources = [
     "cc1_main.cpp",
diff --git a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
index e30622f52195f..1a59c562c79ff 100644
--- a/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/include/BUILD.gn
@@ -1,4 +1,3 @@
-import("//clang/resource_dir.gni")
 import("//libcxx/config.gni")
 import("//llvm/utils/gn/build/write_cmake_config.gni")
 
@@ -7,249 +6,264 @@ declare_args() {
   libcxx_install_support_headers = true
 }
 
-libcxx_needs_site_config =
-    libcxx_abi_version != 1 || libcxx_abi_namespace != "" || libcxx_abi_unstable
+libcxx_generated_include_dir = "$root_build_dir/include/c++/v1"
 
-if (libcxx_needs_site_config) {
-  write_cmake_config("write_config") {
+# This is a bit weird. For now, we assume that __config_site is identical
+# in all toolchains, and only copy it (and all other libcxx headers)
+# to 'include' in the root build dir, so that it's the same for all toolchains.
+# Maybe we wnt to make this per-toolchain eventually (and then use root_out_dir
+# in libcxx_generated_include_dir) -- e.g. for cross-builds that for example
+# use for-linux-configured libc++ for the host build but for-windows-configured
+# libc++ for the target build.
+if (current_toolchain == default_toolchain) {
+  write_cmake_config("write_config_site") {
     input = "__config_site.in"
-    output = "$target_gen_dir/__config_site"
+    output = "$libcxx_generated_include_dir/__config_site"
 
-    values = []
+    values = [
+      "_LIBCPP_ABI_FORCE_ITANIUM=",
+      "_LIBCPP_ABI_FORCE_MICROSOFT=",
+      "_LIBCPP_HIDE_FROM_ABI_PER_TU_BY_DEFAULT=",
+      "_LIBCPP_HAS_NO_GLOBAL_FILESYSTEM_NAMESPACE=",
+      "_LIBCPP_HAS_NO_STDIN=",
+      "_LIBCPP_HAS_NO_STDOUT=",
+      "_LIBCPP_HAS_NO_THREADS=",
+      "_LIBCPP_HAS_NO_MONOTONIC_CLOCK=",
+      "_LIBCPP_HAS_NO_THREAD_UNSAFE_C_FUNCTIONS=",
+      "_LIBCPP_HAS_MUSL_LIBC=",
+      "_LIBCPP_HAS_THREAD_API_PTHREAD=",
+      "_LIBCPP_HAS_THREAD_API_EXTERNAL=",
+      "_LIBCPP_HAS_THREAD_API_WIN32=",
+      "_LIBCPP_HAS_THREAD_LIBRARY_EXTERNAL=",
+      "_LIBCPP_DISABLE_VISIBILITY_ANNOTATIONS=",
+      "_LIBCPP_NO_VCRUNTIME=",
+      "_LIBCPP_TYPEINFO_COMPARISON_IMPLEMENTATION=",
+      "_LIBCPP_HAS_PARALLEL_ALGORITHMS=",
+      "_LIBCPP_HAS_NO_RANDOM_DEVICE=",
+      "_LIBCPP_ABI_DEFINES=",
+    ]
     if (libcxx_abi_version != 1) {
       values += [ "_LIBCPP_ABI_VERSION=$libcxx_abi_version" ]
+    } else {
+      values += [ "_LIBCPP_ABI_VERSION=" ]
     }
-    if (libcxx_abi_namespace != "") {
-      values += [ "_LIBCPP_ABI_NAMESPACE=$libcxx_abi_namespace" ]
-    }
+    values += [ "_LIBCPP_ABI_NAMESPACE=$libcxx_abi_namespace" ]
     if (libcxx_abi_unstable) {
       values += [ "_LIBCPP_ABI_UNSTABLE=1" ]
+    } else {
+      values += [ "_LIBCPP_ABI_UNSTABLE=" ]
     }
   }
 
-  # Generate a custom __config header. The new header is created
-  # by prepending __config_site to the current __config header.
-  action("concat_config") {
-    script = "//libcxx/utils/cat_files.py"
-    inputs = [
-      "$target_gen_dir/__config_site",
+  copy("copy_headers") {
+    sources = [
+      "__bit_reference",
+      "__bsd_locale_defaults.h",
+      "__bsd_locale_fallbacks.h",
       "__config",
+      "__debug",
+      "__errc",
+      "__functional_03",
+      "__functional_base",
+      "__functional_base_03",
+      "__hash_table",
+      "__libcpp_version",
+      "__locale",
+      "__mutex_base",
+      "__node_handle",
+      "__nullptr",
+      "__split_buffer",
+      "__sso_allocator",
+      "__std_stream",
+      "__string",
+      "__threading_support",
+      "__tree",
+      "__tuple",
+      "__undef_macros",
+      "algorithm",
+      "any",
+      "array",
+      "atomic",
+      "barrier",
+      "bit",
+      "bitset",
+      "cassert",
+      "ccomplex",
+      "cctype",
+      "cerrno",
+      "cfenv",
+      "cfloat",
+      "charconv",
+      "chrono",
+      "cinttypes",
+      "ciso646",
+      "climits",
+      "clocale",
+      "cmath",
+      "codecvt",
+      "compare",
+      "complex",
+      "complex.h",
+      "condition_variable",
+      "csetjmp",
+      "csignal",
+      "cstdarg",
+      "cstdbool",
+      "cstddef",
+      "cstdint",
+      "cstdio",
+      "cstdlib",
+      "cstring",
+      "ctgmath",
+      "ctime",
+      "ctype.h",
+      "cwchar",
+      "cwctype",
+      "deque",
+      "errno.h",
+      "exception",
+      "experimental/__config",
+      "experimental/__memory",
+      "experimental/algorithm",
+      "experimental/coroutine",
+      "experimental/deque",
+      "experimental/filesystem",
+      "experimental/forward_list",
+      "experimental/functional",
+      "experimental/iterator",
+      "experimental/list",
+      "experimental/map",
+      "experimental/memory_resource",
+      "experimental/propagate_const",
+      "experimental/regex",
+      "experimental/set",
+      "experimental/simd",
+      "experimental/string",
+      "experimental/type_traits",
+      "experimental/unordered_map",
+      "experimental/unordered_set",
+      "experimental/utility",
+      "experimental/vector",
+      "ext/__hash",
+      "ext/hash_map",
+      "ext/hash_set",
+      "fenv.h",
+      "filesystem",
+      "float.h",
+      "forward_list",
+      "fstream",
+      "functional",
+      "future",
+      "initializer_list",
+      "inttypes.h",
+      "iomanip",
+      "ios",
+      "iosfwd",
+      "iostream",
+      "istream",
+      "iterator",
+      "latch",
+      "limits",
+      "limits.h",
+      "list",
+      "locale",
+      "locale.h",
+      "map",
+      "math.h",
+      "memory",
+      "module.modulemap",
+      "mutex",
+      "new",
+      "numeric",
+      "optional",
+      "ostream",
+      "queue",
+      "random",
+      "ratio",
+      "regex",
+      "scoped_allocator",
+      "semaphore",
+      "set",
+      "setjmp.h",
+      "shared_mutex",
+      "span",
+      "sstream",
+      "stack",
+      "stdbool.h",
+      "stddef.h",
+      "stdexcept",
+      "stdint.h",
+      "stdio.h",
+      "stdlib.h",
+      "streambuf",
+      "string",
+      "string.h",
+      "string_view",
+      "strstream",
+      "system_error",
+      "tgmath.h",
+      "thread",
+      "tuple",
+      "type_traits",
+      "typeindex",
+      "typeinfo",
+      "unordered_map",
+      "unordered_set",
+      "utility",
+      "valarray",
+      "variant",
+      "vector",
+      "version",
+      "wchar.h",
+      "wctype.h",
     ]
-    outputs = [ "$target_gen_dir/__config" ]
-    args = [
-      "$target_gen_dir/__config_site",
-      "__config",
-      "-o",
-      "$target_gen_dir/__config",
-    ]
-    deps = [ ":write_config" ]
+    deps = [ ":write_config_site" ]
+    if (target_os != "mac" && target_os != "win") {
+      # libcxx/cmake/Modules/HandleLibCXXABI.cmake sets
+      # LIBCXX_CXX_ABI_HEADER_TARGET if the libcxx abi library either of
+      # "libstdc++", "libsupc++", "libcxxabi", "libcxxrt", but not if it's "none",
+      # "default", or "vcruntime". So on Windows, these don't get copied due to
+      # LIBCXX_CXX_ABI_HEADER_TARGET not being set.
+      # On macOS, libcxx/CMakeLists.txt sets LIBCXX_CXX_ABI_SYSTEM to 1, which
+      # causes an empty header list to be passed to setup_abi_lib, so these
+      # don't get copied on macOS due to that.
+      deps += [ "//libcxxabi/include" ]
+    }
+    if (libcxx_install_support_headers) {
+      sources += [
+        "support/android/locale_bionic.h",
+        "support/fuchsia/xlocale.h",
+        "support/ibm/limits.h",
+        "support/ibm/locale_mgmt_aix.h",
+        "support/ibm/support.h",
+        "support/ibm/xlocale.h",
+        "support/musl/xlocale.h",
+        "support/newlib/xlocale.h",
+        "support/solaris/floatingpoint.h",
+        "support/solaris/wchar.h",
+        "support/solaris/xlocale.h",
+        "support/xlocale/__nop_locale_mgmt.h",
+        "support/xlocale/__posix_l_fallback.h",
+        "support/xlocale/__strtonum_fallback.h",
+      ]
+      if (target_os == "win") {
+        sources += [
+          "support/win32/limits_msvc_win32.h",
+          "support/win32/locale_win32.h",
+        ]
+      }
+    }
+    outputs = [ "$libcxx_generated_include_dir/{{source_target_relative}}" ]
   }
+}
 
-  copy("copy_config") {
-    sources = [ "$target_gen_dir/__config" ]
-    outputs = [ "$clang_resource_dir/include/c++/v1/{{source_file_part}}" ]
-    deps = [ ":concat_config" ]
-  }
+config("include_config") {
+  include_dirs = [ libcxx_generated_include_dir ]
 }
 
-copy("include") {
-  sources = [
-    "__bit_reference",
-    "__bsd_locale_defaults.h",
-    "__bsd_locale_fallbacks.h",
-    "__debug",
-    "__errc",
-    "__functional_03",
-    "__functional_base",
-    "__functional_base_03",
-    "__hash_table",
-    "__libcpp_version",
-    "__locale",
-    "__mutex_base",
-    "__node_handle",
-    "__nullptr",
-    "__split_buffer",
-    "__sso_allocator",
-    "__std_stream",
-    "__string",
-    "__threading_support",
-    "__tree",
-    "__tuple",
-    "__undef_macros",
-    "algorithm",
-    "any",
-    "array",
-    "atomic",
-    "bit",
-    "bitset",
-    "cassert",
-    "ccomplex",
-    "cctype",
-    "cerrno",
-    "cfenv",
-    "cfloat",
-    "charconv",
-    "chrono",
-    "cinttypes",
-    "ciso646",
-    "climits",
-    "clocale",
-    "cmath",
-    "codecvt",
-    "compare",
-    "complex",
-    "complex.h",
-    "condition_variable",
-    "csetjmp",
-    "csignal",
-    "cstdarg",
-    "cstdbool",
-    "cstddef",
-    "cstdint",
-    "cstdio",
-    "cstdlib",
-    "cstring",
-    "ctgmath",
-    "ctime",
-    "ctype.h",
-    "cwchar",
-    "cwctype",
-    "deque",
-    "errno.h",
-    "exception",
-    "experimental/__config",
-    "experimental/__memory",
-    "experimental/algorithm",
-    "experimental/coroutine",
-    "experimental/deque",
-    "experimental/filesystem",
-    "experimental/forward_list",
-    "experimental/functional",
-    "experimental/iterator",
-    "experimental/list",
-    "experimental/map",
-    "experimental/memory_resource",
-    "experimental/propagate_const",
-    "experimental/regex",
-    "experimental/set",
-    "experimental/simd",
-    "experimental/string",
-    "experimental/type_traits",
-    "experimental/unordered_map",
-    "experimental/unordered_set",
-    "experimental/utility",
-    "experimental/vector",
-    "ext/__hash",
-    "ext/hash_map",
-    "ext/hash_set",
-    "fenv.h",
-    "filesystem",
-    "float.h",
-    "forward_list",
-    "fstream",
-    "functional",
-    "future",
-    "initializer_list",
-    "inttypes.h",
-    "iomanip",
-    "ios",
-    "iosfwd",
-    "iostream",
-    "istream",
-    "iterator",
-    "limits",
-    "limits.h",
-    "list",
-    "locale",
-    "locale.h",
-    "map",
-    "math.h",
-    "memory",
-    "module.modulemap",
-    "mutex",
-    "new",
-    "numeric",
-    "optional",
-    "ostream",
-    "queue",
-    "random",
-    "ratio",
-    "regex",
-    "scoped_allocator",
-    "set",
-    "setjmp.h",
-    "shared_mutex",
-    "span",
-    "sstream",
-    "stack",
-    "stdbool.h",
-    "stddef.h",
-    "stdexcept",
-    "stdint.h",
-    "stdio.h",
-    "stdlib.h",
-    "streambuf",
-    "string",
-    "string.h",
-    "string_view",
-    "strstream",
-    "system_error",
-    "tgmath.h",
-    "thread",
-    "tuple",
-    "type_traits",
-    "typeindex",
-    "typeinfo",
-    "unordered_map",
-    "unordered_set",
-    "utility",
-    "valarray",
-    "variant",
-    "vector",
-    "version",
-    "wchar.h",
-    "wctype.h",
+group("include") {
+  deps = [
+    ":copy_headers($default_toolchain)",
   ]
-  deps = []
-  if (target_os != "mac" && target_os != "win") {
-    # libcxx/cmake/Modules/HandleLibCXXABI.cmake sets
-    # LIBCXX_CXX_ABI_HEADER_TARGET if the libcxx abi library either of
-    # "libstdc++", "libsupc++", "libcxxabi", "libcxxrt", but not if it's "none",
-    # "default", or "vcruntime". So on Windows, these don't get copied due to
-    # LIBCXX_CXX_ABI_HEADER_TARGET not being set.
-    # On macOS, libcxx/CMakeLists.txt sets LIBCXX_CXX_ABI_SYSTEM to 1, which
-    # causes an empty header list to be passed to setup_abi_lib, so these
-    # don't get copied on macOS due to that.
-    deps += [ "//libcxxabi/include" ]
-  }
-  if (!libcxx_needs_site_config) {
-    sources += [ "__config" ]
-  } else {
-    deps += [ ":copy_config" ]
-  }
-  if (libcxx_install_support_headers) {
-    sources += [
-      "support/android/locale_bionic.h",
-      "support/fuchsia/xlocale.h",
-      "support/ibm/limits.h",
-      "support/ibm/locale_mgmt_aix.h",
-      "support/ibm/support.h",
-      "support/ibm/xlocale.h",
-      "support/musl/xlocale.h",
-      "support/newlib/xlocale.h",
-      "support/solaris/floatingpoint.h",
-      "support/solaris/wchar.h",
-      "support/solaris/xlocale.h",
-      "support/xlocale/__nop_locale_mgmt.h",
-      "support/xlocale/__posix_l_fallback.h",
-      "support/xlocale/__strtonum_fallback.h",
-    ]
-    if (target_os == "win") {
-      sources += [
-        "support/win32/limits_msvc_win32.h",
-        "support/win32/locale_win32.h",
-      ]
-    }
-  }
-  outputs = [ "$root_build_dir/include/c++/v1/{{source_target_relative}}" ]
+  public_configs = [ ":include_config" ]
 }
diff --git a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
index 71668b65f72f5..d4b43d6808c75 100644
--- a/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxx/src/BUILD.gn
@@ -1,4 +1,5 @@
 import("//clang/runtimes.gni")
+import("//libcxx/config.gni")
 import("//llvm/utils/gn/build/symlink_or_copy.gni")
 
 declare_args() {
@@ -37,10 +38,7 @@ declare_args() {
 }
 
 config("cxx_config") {
-  include_dirs = [
-    "//libcxxabi/include",
-    "//libcxx/include",
-  ]
+  include_dirs = [ "//libcxxabi/include" ]
   cflags = [
     "-Wall",
     "-Wextra",
@@ -203,6 +201,7 @@ if (libcxx_enable_shared) {
     sources = cxx_sources
     deps = [
       "//compiler-rt/lib/builtins",
+      "//libcxx/include",
       "//libcxxabi/src:cxxabi_shared",
       "//libunwind/src:unwind_shared",
     ]
@@ -252,6 +251,7 @@ if (libcxx_enable_static) {
     }
     deps = [
       "//compiler-rt/lib/builtins",
+      "//libcxx/include",
       "//libcxxabi/src:cxxabi_static",
       "//libunwind/src:unwind_static",
     ]
@@ -268,6 +268,7 @@ if (libcxx_enable_experimental) {
     output_dir = runtimes_dir
     output_name = "c++experimental"
     sources = [ "experimental/memory_resource.cpp" ]
+    deps = [ "//libcxx/include" ]
     configs += [ ":cxx_config" ]
     configs -= [
       "//llvm/utils/gn/build:no_exceptions",
diff --git a/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn b/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn
index 3e6063a211753..7cccf866db2ba 100644
--- a/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn
+++ b/llvm/utils/gn/secondary/libcxxabi/src/BUILD.gn
@@ -59,10 +59,7 @@ if (target_os == "linux" || target_os == "fuchsia") {
 }
 
 config("cxxabi_config") {
-  include_dirs = [
-    "//libcxxabi/include",
-    "//libcxx/include",
-  ]
+  include_dirs = [ "//libcxxabi/include" ]
   cflags_cc = [ "-nostdinc++" ]
   defines = [ "_LIBCXXABI_BUILDING_LIBRARY" ]
   if (target_os == "win") {
@@ -86,6 +83,7 @@ if (libcxxabi_enable_shared) {
     public = cxxabi_headers
     deps = [
       "//compiler-rt/lib/builtins",
+      "//libcxx/include",
       "//libunwind/src:unwind_shared",
     ]
     configs += [ ":cxxabi_config" ]
@@ -116,6 +114,7 @@ if (libcxxabi_enable_static) {
     }
     deps = [
       "//compiler-rt/lib/builtins",
+      "//libcxx/include",
       "//libunwind/src:unwind_static",
     ]
     configs += [ ":cxxabi_config" ]
diff --git a/llvm/utils/lit/lit/LitTestCase.py b/llvm/utils/lit/lit/LitTestCase.py
index 951f7be958e2e..72d57689bf6a5 100644
--- a/llvm/utils/lit/lit/LitTestCase.py
+++ b/llvm/utils/lit/lit/LitTestCase.py
@@ -55,7 +55,7 @@ def load_test_suite(inputs):
         params={})
 
     # Perform test discovery.
-    tests = lit.discovery.find_tests_for_inputs(lit_config, inputs)
+    tests = lit.discovery.find_tests_for_inputs(lit_config, inputs, False)
     test_adaptors = [LitTestCase(t, lit_config) for t in tests]
 
     # Return a unittest test suite which just runs the tests in order.
diff --git a/llvm/utils/lit/lit/cl_arguments.py b/llvm/utils/lit/lit/cl_arguments.py
index 69166e00aba8c..591d4f9aaafb5 100644
--- a/llvm/utils/lit/lit/cl_arguments.py
+++ b/llvm/utils/lit/lit/cl_arguments.py
@@ -123,6 +123,12 @@ def parse_args():
     execution_group.add_argument("--allow-empty-runs",
             help="Do not fail the run if all tests are filtered out",
             action="store_true")
+    execution_group.add_argument("--no-indirectly-run-check",
+            dest="indirectlyRunCheck",
+            help="Do not error if a test would not be run if the user had "
+                 "specified the containing directory instead of naming the "
+                 "test directly.",
+            action="store_false")
 
     selection_group = parser.add_argument_group("Test Selection")
     selection_group.add_argument("--max-tests",
diff --git a/llvm/utils/lit/lit/discovery.py b/llvm/utils/lit/lit/discovery.py
index d8054543d018d..2f027a5b03fe4 100644
--- a/llvm/utils/lit/lit/discovery.py
+++ b/llvm/utils/lit/lit/discovery.py
@@ -125,7 +125,8 @@ def search(path_in_suite):
 
     return search(path_in_suite)
 
-def getTests(path, litConfig, testSuiteCache, localConfigCache):
+def getTests(path, litConfig, testSuiteCache,
+             localConfigCache, indirectlyRunCheck):
     # Find the test suite for this input and its relative path.
     ts,path_in_suite = getTestSuite(path, litConfig, testSuiteCache)
     if ts is None:
@@ -137,10 +138,10 @@ def getTests(path, litConfig, testSuiteCache, localConfigCache):
                                                         path_in_suite))
 
     return ts, getTestsInSuite(ts, path_in_suite, litConfig,
-                               testSuiteCache, localConfigCache)
+                               testSuiteCache, localConfigCache, indirectlyRunCheck)
 
 def getTestsInSuite(ts, path_in_suite, litConfig,
-                    testSuiteCache, localConfigCache):
+                    testSuiteCache, localConfigCache, indirectlyRunCheck):
     # Check that the source path exists (errors here are reported by the
     # caller).
     source_path = ts.getSourcePath(path_in_suite)
@@ -149,8 +150,30 @@ def getTestsInSuite(ts, path_in_suite, litConfig,
 
     # Check if the user named a test directly.
     if not os.path.isdir(source_path):
-        lc = getLocalConfig(ts, path_in_suite[:-1], litConfig, localConfigCache)
-        yield Test.Test(ts, path_in_suite, lc)
+        test_dir_in_suite = path_in_suite[:-1]
+        lc = getLocalConfig(ts, test_dir_in_suite, litConfig, localConfigCache)
+        test = Test.Test(ts, path_in_suite, lc)
+
+        # Issue a error if the specified test would not be run if
+        # the user had specified the containing directory instead of
+        # of naming the test directly. This helps to avoid writing
+        # tests which are not executed. The check adds some performance
+        # overhead which might be important if a large number of tests
+        # are being run directly.
+        # --no-indirectly-run-check: skips this check.
+        if indirectlyRunCheck and lc.test_format is not None:
+            found = False
+            for res in lc.test_format.getTestsInDirectory(ts, test_dir_in_suite,
+                                                          litConfig, lc):
+                if test.getFullName() == res.getFullName():
+                    found = True
+                    break
+            if not found:
+                litConfig.error(
+                    '%r would not be run indirectly: change name or LIT config'
+                    % test.getFullName())
+
+        yield test
         return
 
     # Otherwise we have a directory to search for tests, start by getting the
@@ -196,10 +219,11 @@ def getTestsInSuite(ts, path_in_suite, litConfig,
         # Otherwise, load from the nested test suite, if present.
         if sub_ts is not None:
             subiter = getTestsInSuite(sub_ts, subpath_in_suite, litConfig,
-                                      testSuiteCache, localConfigCache)
+                                      testSuiteCache, localConfigCache,
+                                      indirectlyRunCheck)
         else:
             subiter = getTestsInSuite(ts, subpath, litConfig, testSuiteCache,
-                                      localConfigCache)
+                                      localConfigCache, indirectlyRunCheck)
 
         N = 0
         for res in subiter:
@@ -208,7 +232,7 @@ def getTestsInSuite(ts, path_in_suite, litConfig,
         if sub_ts and not N:
             litConfig.warning('test suite %r contained no tests' % sub_ts.name)
 
-def find_tests_for_inputs(lit_config, inputs):
+def find_tests_for_inputs(lit_config, inputs, indirectlyRunCheck):
     """
     find_tests_for_inputs(lit_config, inputs) -> [Test]
 
@@ -237,8 +261,8 @@ def find_tests_for_inputs(lit_config, inputs):
     local_config_cache = {}
     for input in actual_inputs:
         prev = len(tests)
-        tests.extend(getTests(input, lit_config,
-                              test_suite_cache, local_config_cache)[1])
+        tests.extend(getTests(input, lit_config, test_suite_cache,
+                              local_config_cache, indirectlyRunCheck)[1])
         if prev == len(tests):
             lit_config.warning('input %r contained no tests' % input)
 
diff --git a/llvm/utils/lit/lit/main.py b/llvm/utils/lit/lit/main.py
index d94d7280809da..6c9885c4d4be7 100755
--- a/llvm/utils/lit/lit/main.py
+++ b/llvm/utils/lit/lit/main.py
@@ -39,7 +39,8 @@ def main(builtin_params={}):
         config_prefix=opts.configPrefix,
         echo_all_commands=opts.echoAllCommands)
 
-    discovered_tests = lit.discovery.find_tests_for_inputs(lit_config, opts.test_paths)
+    discovered_tests = lit.discovery.find_tests_for_inputs(lit_config, opts.test_paths,
+                                                           opts.indirectlyRunCheck)
     if not discovered_tests:
         sys.stderr.write('error: did not discover any tests for provided path(s)\n')
         sys.exit(2)
diff --git a/llvm/utils/lit/tests/Inputs/discovery/test.not-txt b/llvm/utils/lit/tests/Inputs/discovery/test.not-txt
new file mode 100644
index 0000000000000..b80b60b7a2794
--- /dev/null
+++ b/llvm/utils/lit/tests/Inputs/discovery/test.not-txt
@@ -0,0 +1 @@
+# RUN: true
diff --git a/llvm/utils/lit/tests/discovery.py b/llvm/utils/lit/tests/discovery.py
index 94b227f33fb3d..cb04edaab04c2 100644
--- a/llvm/utils/lit/tests/discovery.py
+++ b/llvm/utils/lit/tests/discovery.py
@@ -51,17 +51,17 @@
 # CHECK-CONFIG-MAP-ERR: resolved input '{{.*(/|\\\\)config-map-discovery(/|\\\\)main-config}}' to 'config-map'::()
 
 
-# Check discovery when exact test names are given.
+# Check discovery when tests are named directly.
 #
 # RUN: %{lit} \
 # RUN:     %{inputs}/discovery/subdir/test-three.py \
 # RUN:     %{inputs}/discovery/subsuite/test-one.txt \
 # RUN:   -j 1 --show-tests --show-suites -v > %t.out
-# RUN: FileCheck --check-prefix=CHECK-EXACT-TEST < %t.out %s
+# RUN: FileCheck --check-prefix=CHECK-DIRECT-TEST < %t.out %s
 #
-# CHECK-EXACT-TEST: -- Available Tests --
-# CHECK-EXACT-TEST: sub-suite :: test-one
-# CHECK-EXACT-TEST: top-level-suite :: subdir/test-three
+# CHECK-DIRECT-TEST: -- Available Tests --
+# CHECK-DIRECT-TEST: sub-suite :: test-one
+# CHECK-DIRECT-TEST: top-level-suite :: subdir/test-three
 
 # Check discovery when config files end in .py
 # RUN: %{lit} %{inputs}/py-config-discovery \
@@ -122,18 +122,31 @@
 # CHECK-ASEXEC-OUT: top-level-suite :: test-one
 # CHECK-ASEXEC-OUT: top-level-suite :: test-two
 
-# Check discovery when exact test names are given.
+# Check discovery when tests are named directly.
 #
 # FIXME: Note that using a path into a subsuite doesn't work correctly here.
 #
 # RUN: %{lit} \
 # RUN:     %{inputs}/exec-discovery/subdir/test-three.py \
 # RUN:   -j 1 --show-tests --show-suites -v > %t.out
-# RUN: FileCheck --check-prefix=CHECK-ASEXEC-EXACT-TEST < %t.out %s
+# RUN: FileCheck --check-prefix=CHECK-ASEXEC-DIRECT-TEST < %t.out %s
 #
-# CHECK-ASEXEC-EXACT-TEST: -- Available Tests --
-# CHECK-ASEXEC-EXACT-TEST: top-level-suite :: subdir/test-three
+# CHECK-ASEXEC-DIRECT-TEST: -- Available Tests --
+# CHECK-ASEXEC-DIRECT-TEST: top-level-suite :: subdir/test-three
 
+# Check an error is emitted when the directly named test would not be run
+# indirectly (e.g. when the directory containing the test is specified).
+#
+# RUN: not %{lit} \
+# RUN:     %{inputs}/discovery/test.not-txt -j 1 2>%t.err
+# RUN: FileCheck --check-prefix=CHECK-ERROR-INDIRECT-RUN-CHECK < %t.err %s
+#
+# CHECK-ERROR-INDIRECT-RUN-CHECK: error: 'top-level-suite :: test.not-txt' would not be run indirectly
+
+# Check that no error is emitted with --no-indirectly-run-check.
+#
+# RUN: %{lit} \
+# RUN:     %{inputs}/discovery/test.not-txt -j 1 --no-indirectly-run-check
 
 # Check that we don't recurse infinitely when loading an site specific test
 # suite located inside the test source root.
diff --git a/llvm/utils/lit/tests/unit/TestRunner.py b/llvm/utils/lit/tests/unit/TestRunner.py
index 33e4ebc716f40..b4eb1293f3f62 100644
--- a/llvm/utils/lit/tests/unit/TestRunner.py
+++ b/llvm/utils/lit/tests/unit/TestRunner.py
@@ -40,7 +40,7 @@ def load_keyword_parser_lit_tests():
         test_path = os.path.dirname(os.path.dirname(__file__))
         inputs = [os.path.join(test_path, 'Inputs/testrunner-custom-parsers/')]
         assert os.path.isdir(inputs[0])
-        tests = lit.discovery.find_tests_for_inputs(lit_config, inputs)
+        tests = lit.discovery.find_tests_for_inputs(lit_config, inputs, False)
         assert len(tests) == 1 and "there should only be one test"
         TestIntegratedTestKeywordParser.inputTestCase = tests[0]
 
diff --git a/mlir/docs/DeclarativeRewrites.md b/mlir/docs/DeclarativeRewrites.md
index 7e4f675166ab7..81a20f3c29f94 100644
--- a/mlir/docs/DeclarativeRewrites.md
+++ b/mlir/docs/DeclarativeRewrites.md
@@ -136,9 +136,11 @@ and `$attr` in result patterns and additional constraints.
 
 The pattern is position-based: the symbol names used for capturing here do not
 need to match with the op definition as shown in the above example. As another
-example, the pattern can be written as ` def : Pat<(AOp $a, F32Attr:$b), ...>;`
+example, the pattern can be written as `def : Pat<(AOp $a, F32Attr:$b), ...>;`
 and use `$a` and `$b` to refer to the captured input and attribute. But using
-the ODS name directly in the pattern is also allowed.
+the ODS name directly in the pattern is also allowed. Operands in the source
+pattern can have the same name. This bounds one operand to the name while
+verifying the rest are all equal.
 
 Also note that we only need to add `TypeConstraint` or `AttributeConstraint`
 when we need to further limit the match criteria. If all valid cases to the op
diff --git a/mlir/docs/OpDefinitions.md b/mlir/docs/OpDefinitions.md
index 96c257cef08cf..f44765204dc08 100644
--- a/mlir/docs/OpDefinitions.md
+++ b/mlir/docs/OpDefinitions.md
@@ -567,51 +567,104 @@ complete list.
 #### Custom builder methods
 
 However, if the above cases cannot satisfy all needs, you can define additional
-convenience build methods with `OpBuilder`.
+convenience build methods in the `builders` field as follows.
 
-`OpBuilder` is a class that takes the parameter list and the optional `build()`
-method body. They are separated because we need to generate op declaration and
-definition into separate files. The parameter list should not include `OpBuilder
-&builder, OperationState &state` as they will be inserted automatically and the
-placeholders `$_builder` and `$_state` used. For legacy/to be deprecated reason
-if the `OpBuilder` parameter starts with `OpBuilder` param, then the parameter
-is used. If the `body` is not provided, only the builder declaration will be
-generated; this provides a way to define complicated builders entirely in C++
-files.
+```tablegen
+def MyOp : Op<"my_op", []> {
+  let arguments = (ins F32Attr:$attr);
+
+  let builders = [
+    OpBuilderDAG<(ins "float":$val)>
+  ];
+}
+```
+
+The `builders` field is a list of custom builders that are added to the Op
+class. In this example, we provide a convenience builder that takes a floating
+point value instead of an attribute. The `ins` prefix is common to many function
+declarations in ODS, which use a TableGen [`dag`](#tablegen-syntax). What
+follows is a comma-separated list of types (quoted string) and names prefixed
+with the `$` sign. This will generate the declaration of a builder method that
+looks like:
+
+```c++
+class MyOp : /*...*/ {
+  /*...*/
+  static void build(::mlir::OpBuilder &builder, ::mlir::OperationState &state,
+                    float val);
+};
+```
 
-For example, for the following op:
+Note that the method has two additional leading arguments. These arguments are
+useful to construct the operation. In particular, the method must populate
+`state` with attributes, operands, regions and result types of the operation to
+be constructed. `builder` can be used to construct any IR objects that belong to
+the Op, such as types or nested operations. Since the type and name are
+generated as is in the C++ code, they should be valid C++ constructs for a type
+(in the namespace of the Op) and an identifier (e.g., `class` is not a valid
+identifier).
+
+Implementations of the builder can be provided directly in ODS, using TableGen
+code block as follows.
 
 ```tablegen
 def MyOp : Op<"my_op", []> {
   let arguments = (ins F32Attr:$attr);
 
-  let results = (outs);
+  let builders = [
+    OpBuilderDAG<(ins "float":$val), [{
+      $_state.addAttribute("attr", $_builder.getF32FloatAttr(val));
+    }]>
+  ];
 }
 ```
 
-If we want to define a builder with a default value for the only attribute, we
-can add into `MyOp`:
+The equivalents of `builder` and `state` arguments are available as `$_builder`
+and `$_state` special variables. The named arguments listed in the `ins` part
+are available directly, e.g. `val`. The body of the builder will be generated by
+substituting special variables and should otherwise be valid C++. While there is
+no limitation on the code size, we encourage one to define only short builders
+inline in ODS and put definitions of longer builders in C++ files.
+
+Finally, if some arguments need a default value, they can be defined using
+`CArg` to wrap the type and this value as follows.
 
 ```tablegen
-def MyOp : ... {
-  ...
+def MyOp : Op<"my_op", []> {
+  let arguments = (ins F32Attr:$attr);
 
   let builders = [
-    OpBuilder<"float val = 0.5f", [{
+    OpBuilderDAG<(ins CArg<"float", "0.5f">:$val), [{
       $_state.addAttribute("attr", $_builder.getF32FloatAttr(val));
     }]>
   ];
 }
 ```
 
-The generated builder will look like:
+The generated code will use default value in the declaration, but not in the
+definition, as required by C++.
 
 ```c++
-static void build(OpBuilder &builder, OperationState &state, float val = 0.5f) {
+/* Header file. */
+class MyOp : /*...*/ {
+  /*...*/
+  static void build(::mlir::OpBuilder &builder, ::mlir::OperationState &state,
+                    float val = 0.5f);
+};
+
+/* Source file. */
+MyOp::build(::mlir::OpBuilder &builder, ::mlir::OperationState &state,
+            float val) {
   state.addAttribute("attr", builder.getF32FloatAttr(val));
 }
 ```
 
+**Deprecated:** `OpBuilder` class allows one to specify the custom builder
+signature as a raw string, without separating parameters into different `dag`
+arguments. It also supports leading parameters of `OpBuilder &` and
+`OperationState &` types, which will be used instead of the autogenerated ones
+if present.
+
 ### Custom parser and printer methods
 
 Functions to parse and print the operation's custom assembly form.
diff --git a/mlir/docs/Traits.md b/mlir/docs/Traits.md
index 8b1cf0a03b99c..b8debb7d275fd 100644
--- a/mlir/docs/Traits.md
+++ b/mlir/docs/Traits.md
@@ -249,7 +249,7 @@ particular:
 -   Ops must be symbols, i.e. also have the `Symbol` trait;
 -   Ops have a single region with multiple blocks that corresponds to the body
     of the function;
--   the absence of a region corresponds to an external function;
+-   An op with a single empty region corresponds to an external function;
 -   arguments of the first block of the region are treated as function
     arguments;
 -   they can have argument and result attributes that are stored in dictionary
diff --git a/mlir/include/mlir-c/IR.h b/mlir/include/mlir-c/IR.h
index 8161234726477..a08fe77da37cd 100644
--- a/mlir/include/mlir-c/IR.h
+++ b/mlir/include/mlir-c/IR.h
@@ -50,6 +50,7 @@ extern "C" {
 DEFINE_C_API_STRUCT(MlirContext, void);
 DEFINE_C_API_STRUCT(MlirDialect, void);
 DEFINE_C_API_STRUCT(MlirOperation, void);
+DEFINE_C_API_STRUCT(MlirOpPrintingFlags, void);
 DEFINE_C_API_STRUCT(MlirBlock, void);
 DEFINE_C_API_STRUCT(MlirRegion, void);
 
@@ -228,6 +229,42 @@ void mlirOperationStateAddSuccessors(MlirOperationState *state, intptr_t n,
 void mlirOperationStateAddAttributes(MlirOperationState *state, intptr_t n,
                                      MlirNamedAttribute *attributes);
 
+/*============================================================================*/
+/* Op Printing flags API.                                                     */
+/* While many of these are simple settings that could be represented in a     */
+/* struct, they are wrapped in a heap allocated object and accessed via       */
+/* functions to maximize the possibility of compatibility over time.          */
+/*============================================================================*/
+
+/** Creates new printing flags with defaults, intended for customization.
+ * Must be freed with a call to mlirOpPrintingFlagsDestroy(). */
+MlirOpPrintingFlags mlirOpPrintingFlagsCreate();
+
+/** Destroys printing flags created with mlirOpPrintingFlagsCreate. */
+void mlirOpPrintingFlagsDestroy(MlirOpPrintingFlags flags);
+
+/** Enables the elision of large elements attributes by printing a lexically
+ * valid but otherwise meaningless form instead of the element data. The
+ * `largeElementLimit` is used to configure what is considered to be a "large"
+ * ElementsAttr by providing an upper limit to the number of elements. */
+void mlirOpPrintingFlagsElideLargeElementsAttrs(MlirOpPrintingFlags flags,
+                                                intptr_t largeElementLimit);
+
+/** Enable printing of debug information. If 'prettyForm' is set to true,
+ * debug information is printed in a more readable 'pretty' form. Note: The
+ * IR generated with 'prettyForm' is not parsable. */
+void mlirOpPrintingFlagsEnableDebugInfo(MlirOpPrintingFlags flags,
+                                        int prettyForm);
+
+/** Always print operations in the generic form. */
+void mlirOpPrintingFlagsPrintGenericOpForm(MlirOpPrintingFlags flags);
+
+/** Use local scope when printing the operation. This allows for using the
+ * printer in a more localized and thread-safe setting, but may not
+ * necessarily be identical to what the IR will look like when dumping
+ * the full module. */
+void mlirOpPrintingFlagsUseLocalScope(MlirOpPrintingFlags flags);
+
 /*============================================================================*/
 /* Operation API.                                                             */
 /*============================================================================*/
@@ -298,6 +335,11 @@ int mlirOperationRemoveAttributeByName(MlirOperation op, const char *name);
 void mlirOperationPrint(MlirOperation op, MlirStringCallback callback,
                         void *userData);
 
+/** Same as mlirOperationPrint but accepts flags controlling the printing
+ * behavior. */
+void mlirOperationPrintWithFlags(MlirOperation op, MlirOpPrintingFlags flags,
+                                 MlirStringCallback callback, void *userData);
+
 /** Prints an operation to stderr. */
 void mlirOperationDump(MlirOperation op);
 
@@ -432,6 +474,9 @@ intptr_t mlirOpResultGetResultNumber(MlirValue value);
 /** Returns the type of the value. */
 MlirType mlirValueGetType(MlirValue value);
 
+/** Prints the value to the standard error stream. */
+void mlirValueDump(MlirValue value);
+
 /** Prints a value by sending chunks of the string representation and
  * forwarding `userData to `callback`. Note that the callback may be called
  * several times with consecutive chunks of the string. */
diff --git a/mlir/include/mlir/CAPI/IR.h b/mlir/include/mlir/CAPI/IR.h
index dce293d05588d..b3e481dfb6655 100644
--- a/mlir/include/mlir/CAPI/IR.h
+++ b/mlir/include/mlir/CAPI/IR.h
@@ -24,6 +24,7 @@ DEFINE_C_API_PTR_METHODS(MlirContext, mlir::MLIRContext)
 DEFINE_C_API_PTR_METHODS(MlirDialect, mlir::Dialect)
 DEFINE_C_API_PTR_METHODS(MlirOperation, mlir::Operation)
 DEFINE_C_API_PTR_METHODS(MlirBlock, mlir::Block)
+DEFINE_C_API_PTR_METHODS(MlirOpPrintingFlags, mlir::OpPrintingFlags);
 DEFINE_C_API_PTR_METHODS(MlirRegion, mlir::Region)
 
 DEFINE_C_API_METHODS(MlirAttribute, mlir::Attribute)
diff --git a/mlir/include/mlir/Conversion/AsyncToLLVM/AsyncToLLVM.h b/mlir/include/mlir/Conversion/AsyncToLLVM/AsyncToLLVM.h
new file mode 100644
index 0000000000000..abb5d216028f4
--- /dev/null
+++ b/mlir/include/mlir/Conversion/AsyncToLLVM/AsyncToLLVM.h
@@ -0,0 +1,25 @@
+//===- AsyncToLLVM.h - Convert Async to LLVM dialect ------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_CONVERSION_ASYNCTOLLVM_ASYNCTOLLVM_H
+#define MLIR_CONVERSION_ASYNCTOLLVM_ASYNCTOLLVM_H
+
+#include <memory>
+
+namespace mlir {
+
+class ModuleOp;
+template <typename T>
+class OperationPass;
+
+/// Create a pass to convert Async operations to the LLVM dialect.
+std::unique_ptr<OperationPass<ModuleOp>> createConvertAsyncToLLVMPass();
+
+} // namespace mlir
+
+#endif // MLIR_CONVERSION_ASYNCTOLLVM_ASYNCTOLLVM_H
diff --git a/mlir/include/mlir/Conversion/Passes.h b/mlir/include/mlir/Conversion/Passes.h
index b4418bb2e0ac2..1039bd9fb150d 100644
--- a/mlir/include/mlir/Conversion/Passes.h
+++ b/mlir/include/mlir/Conversion/Passes.h
@@ -11,6 +11,7 @@
 
 #include "mlir/Conversion/AVX512ToLLVM/ConvertAVX512ToLLVM.h"
 #include "mlir/Conversion/AffineToStandard/AffineToStandard.h"
+#include "mlir/Conversion/AsyncToLLVM/AsyncToLLVM.h"
 #include "mlir/Conversion/GPUCommon/GPUCommonPass.h"
 #include "mlir/Conversion/GPUToNVVM/GPUToNVVMPass.h"
 #include "mlir/Conversion/GPUToROCDL/GPUToROCDLPass.h"
diff --git a/mlir/include/mlir/Conversion/Passes.td b/mlir/include/mlir/Conversion/Passes.td
index 36618384bb392..2458dca8b49ae 100644
--- a/mlir/include/mlir/Conversion/Passes.td
+++ b/mlir/include/mlir/Conversion/Passes.td
@@ -84,6 +84,21 @@ def ConvertAVX512ToLLVM : Pass<"convert-avx512-to-llvm", "ModuleOp"> {
   let dependentDialects = ["LLVM::LLVMDialect", "LLVM::LLVMAVX512Dialect"];
 }
 
+//===----------------------------------------------------------------------===//
+// AsyncToLLVM
+//===----------------------------------------------------------------------===//
+
+def ConvertAsyncToLLVM : Pass<"convert-async-to-llvm", "ModuleOp"> {
+  let summary = "Convert the operations from the async dialect into the LLVM "
+                "dialect";
+  let description = [{
+    Convert `async.execute` operations to LLVM coroutines and use async runtime
+    API to execute them.
+  }];
+  let constructor = "mlir::createConvertAsyncToLLVMPass()";
+  let dependentDialects = ["LLVM::LLVMDialect"];
+}
+
 //===----------------------------------------------------------------------===//
 // GPUCommon
 //===----------------------------------------------------------------------===//
@@ -91,6 +106,7 @@ def ConvertAVX512ToLLVM : Pass<"convert-avx512-to-llvm", "ModuleOp"> {
 def GpuToLLVMConversionPass : Pass<"gpu-to-llvm", "ModuleOp"> {
   let summary = "Convert GPU dialect to LLVM dialect with GPU runtime calls";
   let constructor = "mlir::createGpuToLLVMConversionPass()";
+  let dependentDialects = ["LLVM::LLVMDialect"];
   let options = [
     Option<"gpuBinaryAnnotation", "gpu-binary-annotation", "std::string",
            "", "Annotation attribute string for GPU binary">,
diff --git a/mlir/include/mlir/Dialect/Async/IR/AsyncOps.td b/mlir/include/mlir/Dialect/Async/IR/AsyncOps.td
index 56bb73f612724..0cc0c7fe25a8f 100644
--- a/mlir/include/mlir/Dialect/Async/IR/AsyncOps.td
+++ b/mlir/include/mlir/Dialect/Async/IR/AsyncOps.td
@@ -24,7 +24,9 @@ include "mlir/Interfaces/SideEffectInterfaces.td"
 class Async_Op<string mnemonic, list<OpTrait> traits = []> :
     Op<AsyncDialect, mnemonic, traits>;
 
-def Async_ExecuteOp : Async_Op<"execute", [AttrSizedOperandSegments]> {
+def Async_ExecuteOp :
+  Async_Op<"execute", [SingleBlockImplicitTerminator<"YieldOp">,
+                       AttrSizedOperandSegments]> {
   let summary = "Asynchronous execute operation";
   let description = [{
     The `body` region attached to the `async.execute` operation semantically
diff --git a/mlir/include/mlir/Dialect/GPU/GPUOps.td b/mlir/include/mlir/Dialect/GPU/GPUOps.td
index 1542241b5b021..180e9cf67d22b 100644
--- a/mlir/include/mlir/Dialect/GPU/GPUOps.td
+++ b/mlir/include/mlir/Dialect/GPU/GPUOps.td
@@ -18,10 +18,6 @@ include "mlir/Dialect/LLVMIR/LLVMOpBase.td"
 include "mlir/IR/SymbolInterfaces.td"
 include "mlir/Interfaces/SideEffectInterfaces.td"
 
-// Type constraint accepting standard integers, indices.
-def IntOrIndex : TypeConstraint<
-  Or<[AnySignlessInteger.predicate, Index.predicate]>, "integer or index">;
-
 //===----------------------------------------------------------------------===//
 // GPU Dialect operations.
 //===----------------------------------------------------------------------===//
@@ -296,9 +292,9 @@ def GPU_GPUFuncOp : GPU_Op<"func", [HasParent<"GPUModuleOp">,
 }
 
 def GPU_LaunchFuncOp : GPU_Op<"launch_func">,
-    Arguments<(ins IntOrIndex:$gridSizeX, IntOrIndex:$gridSizeY,
-               IntOrIndex:$gridSizeZ, IntOrIndex:$blockSizeX,
-               IntOrIndex:$blockSizeY, IntOrIndex:$blockSizeZ,
+    Arguments<(ins SymbolRefAttr:$kernel,
+               Index:$gridSizeX, Index:$gridSizeY, Index:$gridSizeZ,
+               Index:$blockSizeX, Index:$blockSizeY, Index:$blockSizeZ,
                Variadic<AnyType>:$operands)>,
     Results<(outs)> {
   let summary = "Launches a function as a GPU kernel";
@@ -312,8 +308,8 @@ def GPU_LaunchFuncOp : GPU_Op<"launch_func">,
     function is required to be a gpu.module. And finally, the module containing
     the kernel module (which thus cannot be the top-level module) is required
     to have the `gpu.container_module` attribute. The `gpu.launch_func`
-    operation has a symbol attribute named `kernel` to identify the fully specified
-    kernel function to launch (both the gpu.module and func).
+    operation has a symbol attribute named `kernel` to identify the fully 
+    specified kernel function to launch (both the gpu.module and func).
 
     The operation takes at least six operands, with the first three operands
     being grid sizes along x,y,z dimensions and the following three being block
@@ -321,8 +317,6 @@ def GPU_LaunchFuncOp : GPU_Op<"launch_func">,
     unused sizes must be explicitly set to `1`. The remaining operands are
     passed as arguments to the kernel function.
 
-    A custom syntax for this operation is currently not available.
-
     Example:
 
     ```mlir
@@ -357,13 +351,11 @@ def GPU_LaunchFuncOp : GPU_Op<"launch_func">,
         }
       }
 
-      "gpu.launch_func"(%cst, %cst, %cst,  // Grid sizes.
-                        %cst, %cst, %cst,  // Block sizes.
-                        %arg0, %arg1)      // Arguments passed to the kernel.
-            { kernel_module = @kernels,    // Module containing the kernel.
-              kernel = "kernel_1" }        // Kernel function.
-            : (index, index, index, index, index, index, f32, memref<?xf32, 1>)
-              -> ()
+      gpu.launch_func
+          @kernels::@kernel_1                          // Kernel function.
+          blocks in (%cst, %cst, %cst)                 // Grid size.
+          threads in (%cst, %cst, %cst)                // Block size.
+          args(%arg0 : f32, %arg1 : memref<?xf32, 1>)  // Kernel arguments.
     }
     ```
   }];
@@ -371,19 +363,12 @@ def GPU_LaunchFuncOp : GPU_Op<"launch_func">,
   let skipDefaultBuilders = 1;
 
   let builders = [
-    OpBuilder<"GPUFuncOp kernelFunc, "
-              "Value gridSizeX, Value gridSizeY, Value gridSizeZ, "
-              "Value blockSizeX, Value blockSizeY, Value blockSizeZ, "
-              "ValueRange kernelOperands">,
     OpBuilder<"GPUFuncOp kernelFunc, "
               "KernelDim3 gridSize, KernelDim3 blockSize, "
               "ValueRange kernelOperands">
   ];
 
   let extraClassDeclaration = [{
-    /// The kernel function specified by the operation's `kernel` attribute.
-    SymbolRefAttr kernel();
-
     /// The number of operands passed to the kernel function.
     unsigned getNumKernelOperands();
 
@@ -416,6 +401,13 @@ def GPU_LaunchFuncOp : GPU_Op<"launch_func">,
   }];
 
   let verifier = [{ return ::verify(*this); }];
+  let assemblyFormat = [{
+      $kernel
+      `blocks` `in` ` ` `(`$gridSizeX`,` $gridSizeY`,` $gridSizeZ`)`
+      `threads` `in` ` ` `(`$blockSizeX`,` $blockSizeY`,` $blockSizeZ`)`
+      custom<LaunchFuncOperands>($operands, type($operands))
+      attr-dict
+  }];
 }
 
 def GPU_LaunchOp : GPU_Op<"launch">,
diff --git a/mlir/include/mlir/Dialect/SCF/Passes.h b/mlir/include/mlir/Dialect/SCF/Passes.h
index 7edb2444e87c0..f3dda9bec335c 100644
--- a/mlir/include/mlir/Dialect/SCF/Passes.h
+++ b/mlir/include/mlir/Dialect/SCF/Passes.h
@@ -17,6 +17,9 @@
 
 namespace mlir {
 
+/// Creates a pass that bufferizes the SCF dialect.
+std::unique_ptr<Pass> createSCFBufferizePass();
+
 /// Creates a pass that specializes for loop for unrolling and
 /// vectorization.
 std::unique_ptr<Pass> createForLoopSpecializationPass();
diff --git a/mlir/include/mlir/Dialect/SCF/Passes.td b/mlir/include/mlir/Dialect/SCF/Passes.td
index 6f3cf0e126423..611869466214a 100644
--- a/mlir/include/mlir/Dialect/SCF/Passes.td
+++ b/mlir/include/mlir/Dialect/SCF/Passes.td
@@ -11,6 +11,11 @@
 
 include "mlir/Pass/PassBase.td"
 
+def SCFBufferize : FunctionPass<"scf-bufferize"> {
+  let summary = "Bufferize the scf dialect.";
+  let constructor = "mlir::createSCFBufferizePass()";
+}
+
 def SCFForLoopSpecialization
     : FunctionPass<"for-loop-specialization"> {
   let summary = "Specialize `for` loops for vectorization";
diff --git a/mlir/include/mlir/Dialect/SCF/Transforms.h b/mlir/include/mlir/Dialect/SCF/Transforms.h
index 222ad6bf5584b..3164d337b4775 100644
--- a/mlir/include/mlir/Dialect/SCF/Transforms.h
+++ b/mlir/include/mlir/Dialect/SCF/Transforms.h
@@ -17,7 +17,11 @@
 
 namespace mlir {
 
+class ConversionTarget;
+class MLIRContext;
+class OwningRewritePatternList;
 class Region;
+class TypeConverter;
 
 namespace scf {
 
@@ -42,6 +46,19 @@ void naivelyFuseParallelOps(Region &region);
 /// The old loop is replaced with the new one.
 void tileParallelLoop(ParallelOp op, llvm::ArrayRef<int64_t> tileSizes);
 
+/// Populates patterns for SCF structural type conversions and sets up the
+/// provided ConversionTarget with the appropriate legality configuration for
+/// the ops to get converted properly.
+///
+/// A "structural" type conversion is one where the underlying ops are
+/// completely agnostic to the actual types involved and simply need to update
+/// their types. An example of this is scf.if -- the scf.if op and the
+/// corresponding scf.yield ops need to update their types accordingly to the
+/// TypeConverter, but otherwise don't care what type conversions are happening.
+void populateSCFStructuralTypeConversionsAndLegality(
+    MLIRContext *context, TypeConverter &typeConverter,
+    OwningRewritePatternList &patterns, ConversionTarget &target);
+
 } // namespace scf
 } // namespace mlir
 
diff --git a/mlir/include/mlir/Dialect/Shape/IR/ShapeOps.td b/mlir/include/mlir/Dialect/Shape/IR/ShapeOps.td
index c6c52f2eb6ee0..6541cfadfc1b6 100644
--- a/mlir/include/mlir/Dialect/Shape/IR/ShapeOps.td
+++ b/mlir/include/mlir/Dialect/Shape/IR/ShapeOps.td
@@ -635,6 +635,7 @@ def Shape_AssumingOp : Shape_Op<"assuming",
 
   let printer = [{ return ::print(p, *this); }];
   let parser = [{ return ::parse$cppClass(parser, result); }];
+  let verifier = [{ return RegionBranchOpInterface::verifyTypes(*this); }];
 
   let extraClassDeclaration = [{
     // Inline the region into the region containing the AssumingOp and delete
diff --git a/mlir/include/mlir/Dialect/Shape/Transforms/Passes.h b/mlir/include/mlir/Dialect/Shape/Transforms/Passes.h
index f8976e9c75ebb..6df12998566ab 100644
--- a/mlir/include/mlir/Dialect/Shape/Transforms/Passes.h
+++ b/mlir/include/mlir/Dialect/Shape/Transforms/Passes.h
@@ -17,7 +17,8 @@
 #include "mlir/Pass/Pass.h"
 
 namespace mlir {
-class BufferizeTypeConverter;
+class ConversionTarget;
+class TypeConverter;
 } // namespace mlir
 
 namespace mlir {
@@ -40,9 +41,21 @@ void populateRemoveShapeConstraintsPatterns(OwningRewritePatternList &patterns,
                                             MLIRContext *ctx);
 std::unique_ptr<FunctionPass> createRemoveShapeConstraintsPass();
 
-void populateShapeTypeConversionPatterns(MLIRContext *ctx,
-                                         BufferizeTypeConverter &converter,
-                                         OwningRewritePatternList &patterns);
+/// Populates patterns for shape dialect structural type conversions and sets up
+/// the provided ConversionTarget with the appropriate legality configuration
+/// for the ops to get converted properly.
+///
+/// A "structural" type conversion is one where the underlying ops are
+/// completely agnostic to the actual types involved and simply need to update
+/// their types consistently. An example of this is shape.assuming -- the
+/// shape.assuming op and the corresponding shape.assuming_yield op need to have
+/// consistent types, but the exact types don't matter. So all that we need to
+/// do for a structural type conversion is to update both of their types
+/// consistently to the new types prescribed by the TypeConverter.
+void populateShapeStructuralTypeConversionsAndLegality(
+    MLIRContext *context, TypeConverter &typeConverter,
+    OwningRewritePatternList &patterns, ConversionTarget &target);
+
 // Bufferizes shape dialect ops.
 //
 // Note that most shape dialect ops must be converted to std before
diff --git a/mlir/include/mlir/Dialect/StandardOps/IR/Ops.td b/mlir/include/mlir/Dialect/StandardOps/IR/Ops.td
index 31fb780cf1450..e3e1930580b4e 100644
--- a/mlir/include/mlir/Dialect/StandardOps/IR/Ops.td
+++ b/mlir/include/mlir/Dialect/StandardOps/IR/Ops.td
@@ -2217,6 +2217,116 @@ def MemRefCastOp : CastOp<"memref_cast", [
   }];
 }
 
+
+//===----------------------------------------------------------------------===//
+// MemRefReinterpretCastOp
+//===----------------------------------------------------------------------===//
+
+def MemRefReinterpretCastOp:
+    BaseOpWithOffsetSizesAndStrides<"memref_reinterpret_cast", [
+      NoSideEffect, ViewLikeOpInterface
+    ]> {
+  let summary = "memref reinterpret cast operation";
+  let description = [{
+    Modify offset, sizes and strides of an unranked/ranked memref.
+
+    Example:
+    ```mlir
+    memref_reinterpret_cast %ranked to
+      offset: [0],
+      sizes: [%size0, 10],
+      strides: [1, %stride1]
+    : memref<?x?xf32> to memref<?x10xf32, offset: 0, strides: [1, ?]>
+
+    memref_reinterpret_cast %unranked to
+      offset: [%offset],
+      sizes: [%size0, %size1],
+      strides: [%stride0, %stride1]
+    : memref<*xf32> to memref<?x?xf32, offset: ?, strides: [?, ?]>
+    ```
+  }];
+
+  let arguments = (ins
+    Arg<AnyRankedOrUnrankedMemRef, "", []>:$source,
+    Variadic<Index>:$offsets,
+    Variadic<Index>:$sizes,
+    Variadic<Index>:$strides,
+    I64ArrayAttr:$static_offsets,
+    I64ArrayAttr:$static_sizes,
+    I64ArrayAttr:$static_strides
+  );
+  let results = (outs AnyMemRef:$result);
+  let extraClassDeclaration = extraBaseClassDeclaration # [{
+    // The result of the op is always a ranked memref.
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+    Value getViewSource() { return source(); }
+  }];
+}
+
+//===----------------------------------------------------------------------===//
+// MemRefReshapeOp
+//===----------------------------------------------------------------------===//
+
+def MemRefReshapeOp: Std_Op<"memref_reshape", [
+    ViewLikeOpInterface, NoSideEffect]>  {
+  let summary = "memref reshape operation";
+  let description = [{
+    The `memref_reshape` operation converts a memref from one type to an
+    equivalent type with a provided shape. The data is never copied or
+    modified.  The source and destination types are compatible if both have the
+    same element type, same number of elements, address space and identity
+    layout map. The following combinations are possible:
+
+    a. Source type is ranked or unranked. Shape argument has static size.
+    Result type is ranked.
+
+    ```mlir
+    // Reshape statically-shaped memref.
+    %dst = memref_reshape %src(%shape)
+             : (memref<4x1xf32>, memref<1xi32>) to memref<4xf32>
+    %dst0 = memref_reshape %src(%shape0)
+             : (memref<4x1xf32>, memref<2xi32>) to memref<2x2xf32>
+    // Flatten unranked memref.
+    %dst = memref_reshape %src(%shape)
+             : (memref<*xf32>, memref<1xi32>) to memref<?xf32>
+    ```
+
+    a. Source type is ranked or unranked. Shape argument has dynamic size.
+    Result type is unranked.
+
+    ```mlir
+    // Reshape dynamically-shaped 1D memref.
+    %dst = memref_reshape %src(%shape)
+             : (memref<?xf32>, memref<?xi32>) to memref<*xf32>
+    // Reshape unranked memref.
+    %dst = memref_reshape %src(%shape)
+             : (memref<*xf32>, memref<?xi32>) to memref<*xf32>
+    ```
+  }];
+
+  let arguments = (ins
+    AnyRankedOrUnrankedMemRef:$source,
+    MemRefRankOf<[AnySignlessInteger], [1]>:$shape
+  );
+  let results = (outs AnyRankedOrUnrankedMemRef:$result);
+
+  let builders = [OpBuilder<
+     "MemRefType resultType, Value operand, Value shape", [{
+       $_state.addOperands(operand);
+       $_state.addOperands(shape);
+       $_state.addTypes(resultType);
+     }]>];
+
+  let extraClassDeclaration = [{
+    MemRefType getType() { return getResult().getType().cast<MemRefType>(); }
+    Value getViewSource() { return source(); }
+  }];
+
+  let assemblyFormat = [{
+    $source `(` $shape `)` attr-dict `:` functional-type(operands, results)
+  }];
+}
+
 //===----------------------------------------------------------------------===//
 // MulFOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/include/mlir/ExecutionEngine/AsyncRuntime.h b/mlir/include/mlir/ExecutionEngine/AsyncRuntime.h
new file mode 100644
index 0000000000000..f6b739bad4616
--- /dev/null
+++ b/mlir/include/mlir/ExecutionEngine/AsyncRuntime.h
@@ -0,0 +1,73 @@
+//===- AsyncRuntime.h - Async runtime reference implementation ------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file declares basic Async runtime API for supporting Async dialect
+// to LLVM dialect lowering.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef MLIR_EXECUTIONENGINE_ASYNCRUNTIME_H_
+#define MLIR_EXECUTIONENGINE_ASYNCRUNTIME_H_
+
+#ifdef _WIN32
+#ifndef MLIR_ASYNCRUNTIME_EXPORT
+#ifdef mlir_c_runner_utils_EXPORTS
+/* We are building this library */
+#define MLIR_ASYNCRUNTIME_EXPORT __declspec(dllexport)
+#define MLIR_ASYNCRUNTIME_DEFINE_FUNCTIONS
+#else
+/* We are using this library */
+#define MLIR_ASYNCRUNTIME_EXPORT __declspec(dllimport)
+#endif // mlir_c_runner_utils_EXPORTS
+#endif // MLIR_ASYNCRUNTIME_EXPORT
+#else
+#define MLIR_ASYNCRUNTIME_EXPORT
+#define MLIR_ASYNCRUNTIME_DEFINE_FUNCTIONS
+#endif // _WIN32
+
+//===----------------------------------------------------------------------===//
+// Async runtime API.
+//===----------------------------------------------------------------------===//
+
+// Runtime implementation of `async.token` data type.
+typedef struct AsyncToken MLIR_AsyncToken;
+
+// Async runtime uses LLVM coroutines to represent asynchronous tasks. Task
+// function is a coroutine handle and a resume function that continue coroutine
+// execution from a suspension point.
+using CoroHandle = void *;           // coroutine handle
+using CoroResume = void (*)(void *); // coroutine resume function
+
+// Create a new `async.token` in not-ready state.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT AsyncToken *mlirAsyncRuntimeCreateToken();
+
+// Switches `async.token` to ready state and runs all awaiters.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeEmplaceToken(AsyncToken *);
+
+// Blocks the caller thread until the token becomes ready.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeAwaitToken(AsyncToken *);
+
+// Executes the task (coro handle + resume function) in one of the threads
+// managed by the runtime.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void mlirAsyncRuntimeExecute(CoroHandle,
+                                                                 CoroResume);
+
+// Executes the task (coro handle + resume function) in one of the threads
+// managed by the runtime after the token becomes ready.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeAwaitTokenAndExecute(AsyncToken *, CoroHandle, CoroResume);
+
+//===----------------------------------------------------------------------===//
+// Small async runtime support library for testing.
+//===----------------------------------------------------------------------===//
+
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void mlirAsyncRuntimePrintCurrentThreadId();
+
+#endif // MLIR_EXECUTIONENGINE_ASYNCRUNTIME_H_
diff --git a/mlir/include/mlir/IR/FunctionImplementation.h b/mlir/include/mlir/IR/FunctionImplementation.h
index 958cba51f6dcf..c19100c55219c 100644
--- a/mlir/include/mlir/IR/FunctionImplementation.h
+++ b/mlir/include/mlir/IR/FunctionImplementation.h
@@ -49,6 +49,16 @@ void addArgAndResultAttrs(Builder &builder, OperationState &result,
 using FuncTypeBuilder = function_ref<Type(
     Builder &, ArrayRef<Type>, ArrayRef<Type>, VariadicFlag, std::string &)>;
 
+/// Parses function arguments using `parser`. The `allowVariadic` argument
+/// indicates whether functions with variadic arguments are supported. The
+/// trailing arguments are populated by this function with names, types and
+/// attributes of the arguments.
+ParseResult parseFunctionArgumentList(
+    OpAsmParser &parser, bool allowAttributes, bool allowVariadic,
+    SmallVectorImpl<OpAsmParser::OperandType> &argNames,
+    SmallVectorImpl<Type> &argTypes, SmallVectorImpl<NamedAttrList> &argAttrs,
+    bool &isVariadic);
+
 /// Parses a function signature using `parser`. The `allowVariadic` argument
 /// indicates whether functions with variadic arguments are supported. The
 /// trailing arguments are populated by this function with names, types and
diff --git a/mlir/include/mlir/IR/FunctionSupport.h b/mlir/include/mlir/IR/FunctionSupport.h
index 3d467cd4f3642..7756761c2a523 100644
--- a/mlir/include/mlir/IR/FunctionSupport.h
+++ b/mlir/include/mlir/IR/FunctionSupport.h
@@ -79,7 +79,7 @@ namespace OpTrait {
 /// - Ops must be symbols, i.e. also have the `Symbol` trait;
 /// - Ops have a single region with multiple blocks that corresponds to the body
 ///   of the function;
-/// - the absence of a region corresponds to an external function;
+/// - An op with a single empty region corresponds to an external function;
 /// - leading arguments of the first block of the region are treated as function
 ///   arguments;
 /// - they can have argument attributes that are stored in a dictionary
diff --git a/mlir/include/mlir/IR/OpBase.td b/mlir/include/mlir/IR/OpBase.td
index e09c18c0e1a14..48412e9229a11 100644
--- a/mlir/include/mlir/IR/OpBase.td
+++ b/mlir/include/mlir/IR/OpBase.td
@@ -1804,6 +1804,13 @@ def NoRegionArguments : NativeOpTrait<"NoRegionArguments">;
 // Marker used to identify the argument list for an op or interface method.
 def ins;
 
+// This class represents a typed argument with optional default value for C
+// function signatures, e.g. builders or methods.
+class CArg<string ty, string value = ""> {
+  string type = ty;
+  string defaultValue = value;
+}
+
 // OpInterfaceTrait corresponds to a specific 'OpInterface' class defined in
 // C++. The purpose to wrap around C++ symbol string with this class is to make
 // interfaces specified for ops in TableGen less alien and more integrated.
@@ -1923,6 +1930,15 @@ def region;
 // Marker used to identify the successor list for an op.
 def successor;
 
+// Base class for custom builders. This is a transient class that will go away
+// when the transition to the DAG form of builder declaration is complete.
+// Should not be used directly.
+class OpBuilderBase<dag dp, code b> {
+  string params = ?;
+  dag dagParams = dp;
+  code body = b;
+}
+
 // Class for defining a custom builder.
 //
 // TableGen generates several generic builders for each op by default (see
@@ -1932,22 +1948,40 @@ def successor;
 // The signature of the builder is always
 //
 // ```c++
-// static void build(OpBuilder &builder, OperationState &state,
+// static void build(::mlir::OpBuilder &builder, ::mlir::OperationState &state,
 //                   <other-parameters>...) {
 //   <body>...
 // }
 // ```
 //
-// To define a custom builder, the parameter list (*excluding* the `Builder
-// *builder, OperationState &state` part) and body should be passed in
-// as separate template arguments to this class. This is because we generate
-// op declaration and definition into separate files. If an empty string is
-// passed in for `body`, then *only* the builder declaration will be
-// generated; this provides a way to define complicated builders entirely
-// in C++.
-class OpBuilder<string p, code b = ""> {
-  string params = p;
-  code body = b;
+// To define a custom builder, the parameter list (*excluding* the
+// `OpBuilder &builder, OperationState &state` part) and body should be passed
+// in as separate template arguments to this class. The parameter list is a
+// TableGen DAG with `ins` operation with named arguments, which has either:
+//   - string initializers ("Type":$name) to represent a typed parameter, or
+//   - CArg-typed initializers (CArg<"Type", "default">:$name) to represent a
+//     typed parameter that may have a default value.
+// The type string is used verbatim to produce code and, therefore, must be a
+// valid C++ type. It is used inside the C++ namespace of the parent Op's
+// dialect; explicit namespace qualification like `::mlir` may be necessary if
+// Ops are not placed inside the `mlir` namespace. The default value string is
+// used verbatim to produce code and must be a valid C++ initializer the given
+// type. For example, the following signature specification
+//
+// ```
+// OpBuilderDAG<(ins "int":$integerArg, CArg<"float", "3.0f">:$floatArg)>
+// ```
+//
+// has an integer parameter and a float parameter with a default value.
+//
+// If an empty string is passed in for `body`, then *only* the builder
+// declaration will be generated; this provides a way to define complicated
+// builders entirely in C++.
+class OpBuilderDAG<dag p, code b = ""> : OpBuilderBase<p, b>;
+
+// Deprecated version of OpBuilder that takes the builder signature as string.
+class OpBuilder<string p, code b = ""> : OpBuilderBase<(ins), b> {
+  let params = p;
 }
 
 // A base decorator class that may optionally be added to OpVariables.
@@ -2025,7 +2059,7 @@ class Op<Dialect dialect, string mnemonic, list<OpTrait> props = []> {
   //                   ValueRange operands,
   //                   ArrayRef<NamedAttribute> attributes);
   // ```
-  list<OpBuilder> builders = ?;
+  list<OpBuilderBase> builders = ?;
 
   // Avoid generating default build functions.  Custom builders must be
   // provided.
@@ -2258,8 +2292,11 @@ def addBenefit;
 //
 // In the source pattern, `argN` can be used to specify matchers (e.g., using
 // type/attribute type constraints, etc.) and bound to a name for later use.
-// We can also bound names to op instances to reference them later in
-// multi-entity constraints.
+// We can also bind names to op instances to reference them later in
+// multi-entity constraints. Operands in the source pattern can have
+// the same name. This bounds one operand to the name while verifying
+// the rest are all equal.
+//
 //
 // In the result pattern, `argN` can be used to refer to a previously bound
 // name, with potential transformations (e.g., using tAttr, etc.). `argN` can
@@ -2269,16 +2306,18 @@ def addBenefit;
 // For example,
 //
 // ```
-// def : Pattern<(OneResultOp1:$op1 $arg0, $arg1),
+// def : Pattern<(OneResultOp1:$op1 $arg0, $arg1, $arg0),
 //               [(OneResultOp2:$op2 $arg0, $arg1),
 //                (OneResultOp3 $op2 (OneResultOp4))],
 //               [(HasStaticShapePred $op1)]>;
 // ```
 //
-// `$argN` is bound to the `OneResultOp1`'s N-th argument and used later to
-// build `OneResultOp2`. `$op1` is bound to `OneResultOp1` and used to
-// check whether the result's shape is static. `$op2` is bound to
-// `OneResultOp2` and used to build `OneResultOp3`.
+// First `$arg0` and '$arg1' are bound to the `OneResultOp1`'s first
+// and second arguments and used later to build `OneResultOp2`. Second '$arg0'
+// is verified to be equal to the first '$arg0' operand.
+// `$op1` is bound to `OneResultOp1` and used to check whether the result's
+// shape is static. `$op2` is bound to `OneResultOp2` and used to
+// build `OneResultOp3`.
 //
 // ## Multi-result op
 //
diff --git a/mlir/include/mlir/IR/OperationSupport.h b/mlir/include/mlir/IR/OperationSupport.h
index e1e34f8da6c69..c745c1dedea3b 100644
--- a/mlir/include/mlir/IR/OperationSupport.h
+++ b/mlir/include/mlir/IR/OperationSupport.h
@@ -562,10 +562,10 @@ class OpPrintingFlags {
   OpPrintingFlags();
   OpPrintingFlags(llvm::NoneType) : OpPrintingFlags() {}
 
-  /// Enable the elision of large elements attributes, by printing a '...'
-  /// instead of the element data. Note: The IR generated with this option is
-  /// not parsable. `largeElementLimit` is used to configure what is considered
-  /// to be a "large" ElementsAttr by providing an upper limit to the number of
+  /// Enables the elision of large elements attributes by printing a lexically
+  /// valid but otherwise meaningless form instead of the element data. The
+  /// `largeElementLimit` is used to configure what is considered to be a
+  /// "large" ElementsAttr by providing an upper limit to the number of
   /// elements.
   OpPrintingFlags &elideLargeElementsAttrs(int64_t largeElementLimit = 16);
 
diff --git a/mlir/include/mlir/Transforms/Bufferize.h b/mlir/include/mlir/Transforms/Bufferize.h
index ddc00893cc472..5bee53ef01ce6 100644
--- a/mlir/include/mlir/Transforms/Bufferize.h
+++ b/mlir/include/mlir/Transforms/Bufferize.h
@@ -143,6 +143,15 @@ class BufferizeTypeConverter : public TypeConverter {
   SmallVector<DecomposeTypeConversionCallFn, 2> decomposeTypeConversions;
 };
 
+/// Marks ops used by bufferization for type conversion materializations as
+/// "legal" in the given ConversionTarget.
+///
+/// This function should be called by all bufferization passes using
+/// BufferizeTypeConverter so that materializations work proprely. One exception
+/// is bufferization passes doing "full" conversions, where it can be desirable
+/// for even the materializations to remain illegal so that they are eliminated.
+void populateBufferizeMaterializationLegality(ConversionTarget &target);
+
 /// Helper conversion pattern that encapsulates a BufferizeTypeConverter
 /// instance.
 template <typename SourceOp>
diff --git a/mlir/lib/Bindings/Python/IRModules.cpp b/mlir/lib/Bindings/Python/IRModules.cpp
index 2a768df0ffd9c..014b312971b7d 100644
--- a/mlir/lib/Bindings/Python/IRModules.cpp
+++ b/mlir/lib/Bindings/Python/IRModules.cpp
@@ -64,17 +64,46 @@ static const char kContextGetUnknownLocationDocstring[] =
 static const char kContextGetFileLocationDocstring[] =
     R"(Gets a Location representing a file, line and column)";
 
+static const char kOperationPrintDocstring[] =
+    R"(Prints the assembly form of the operation to a file like object.
+
+Args:
+  file: The file like object to write to. Defaults to sys.stdout.
+  binary: Whether to write bytes (True) or str (False). Defaults to False.
+  large_elements_limit: Whether to elide elements attributes above this
+    number of elements. Defaults to None (no limit).
+  enable_debug_info: Whether to print debug/location information. Defaults
+    to False.
+  pretty_debug_info: Whether to format debug information for easier reading
+    by a human (warning: the result is unparseable).
+  print_generic_op_form: Whether to print the generic assembly forms of all
+    ops. Defaults to False.
+  use_local_Scope: Whether to print in a way that is more optimized for
+    multi-threaded access but may not be consistent with how the overall
+    module prints.
+)";
+
+static const char kOperationGetAsmDocstring[] =
+    R"(Gets the assembly form of the operation with all options available.
+
+Args:
+  binary: Whether to return a bytes (True) or str (False) object. Defaults to
+    False.
+  ... others ...: See the print() method for common keyword arguments for
+    configuring the printout.
+Returns:
+  Either a bytes or str object, depending on the setting of the 'binary'
+  argument.
+)";
+
 static const char kOperationStrDunderDocstring[] =
-    R"(Prints the assembly form of the operation with default options.
+    R"(Gets the assembly form of the operation with default options.
 
 If more advanced control over the assembly formatting or I/O options is needed,
-use the dedicated print method, which supports keyword arguments to customize
-behavior.
+use the dedicated print or get_asm method, which supports keyword arguments to
+customize behavior.
 )";
 
-static const char kTypeStrDunderDocstring[] =
-    R"(Prints the assembly form of the type.)";
-
 static const char kDumpDocstring[] =
     R"(Dumps a debug representation of the object to stderr.)";
 
@@ -85,6 +114,14 @@ static const char kAppendBlockDocstring[] =
   The created block.
 )";
 
+static const char kValueDunderStrDocstring[] =
+    R"(Returns the string form of the value.
+
+If the value is a block argument, this is the assembly form of its type and the
+position in the argument list. If the value is an operation result, this is
+equivalent to printing the operation that produced it.
+)";
+
 //------------------------------------------------------------------------------
 // Conversion utilities.
 //------------------------------------------------------------------------------
@@ -113,6 +150,35 @@ struct PyPrintAccumulator {
   }
 };
 
+/// Accumulates int a python file-like object, either writing text (default)
+/// or binary.
+class PyFileAccumulator {
+public:
+  PyFileAccumulator(py::object fileObject, bool binary)
+      : pyWriteFunction(fileObject.attr("write")), binary(binary) {}
+
+  void *getUserData() { return this; }
+
+  MlirStringCallback getCallback() {
+    return [](const char *part, intptr_t size, void *userData) {
+      py::gil_scoped_acquire();
+      PyFileAccumulator *accum = static_cast<PyFileAccumulator *>(userData);
+      if (accum->binary) {
+        // Note: Still has to copy and not avoidable with this API.
+        py::bytes pyBytes(part, size);
+        accum->pyWriteFunction(pyBytes);
+      } else {
+        py::str pyStr(part, size); // Decodes as UTF-8 by default.
+        accum->pyWriteFunction(pyStr);
+      }
+    };
+  }
+
+private:
+  py::object pyWriteFunction;
+  bool binary;
+};
+
 /// Accumulates into a python string from a method that is expected to make
 /// one (no more, no less) call to the callback (asserts internally on
 /// violation).
@@ -707,6 +773,48 @@ void PyOperation::checkValid() {
   }
 }
 
+void PyOperation::print(py::object fileObject, bool binary,
+                        llvm::Optional<int64_t> largeElementsLimit,
+                        bool enableDebugInfo, bool prettyDebugInfo,
+                        bool printGenericOpForm, bool useLocalScope) {
+  checkValid();
+  if (fileObject.is_none())
+    fileObject = py::module::import("sys").attr("stdout");
+  MlirOpPrintingFlags flags = mlirOpPrintingFlagsCreate();
+  if (largeElementsLimit)
+    mlirOpPrintingFlagsElideLargeElementsAttrs(flags, *largeElementsLimit);
+  if (enableDebugInfo)
+    mlirOpPrintingFlagsEnableDebugInfo(flags, /*prettyForm=*/prettyDebugInfo);
+  if (printGenericOpForm)
+    mlirOpPrintingFlagsPrintGenericOpForm(flags);
+
+  PyFileAccumulator accum(fileObject, binary);
+  py::gil_scoped_release();
+  mlirOperationPrintWithFlags(get(), flags, accum.getCallback(),
+                              accum.getUserData());
+  mlirOpPrintingFlagsDestroy(flags);
+}
+
+py::object PyOperation::getAsm(bool binary,
+                               llvm::Optional<int64_t> largeElementsLimit,
+                               bool enableDebugInfo, bool prettyDebugInfo,
+                               bool printGenericOpForm, bool useLocalScope) {
+  py::object fileObject;
+  if (binary) {
+    fileObject = py::module::import("io").attr("BytesIO")();
+  } else {
+    fileObject = py::module::import("io").attr("StringIO")();
+  }
+  print(fileObject, /*binary=*/binary,
+        /*largeElementsLimit=*/largeElementsLimit,
+        /*enableDebugInfo=*/enableDebugInfo,
+        /*prettyDebugInfo=*/prettyDebugInfo,
+        /*printGenericOpForm=*/printGenericOpForm,
+        /*useLocalScope=*/useLocalScope);
+
+  return fileObject.attr("getvalue")();
+}
+
 //------------------------------------------------------------------------------
 // PyAttribute.
 //------------------------------------------------------------------------------
@@ -732,6 +840,169 @@ bool PyType::operator==(const PyType &other) {
   return mlirTypeEqual(type, other.type);
 }
 
+//------------------------------------------------------------------------------
+// PyValue and subclases.
+//------------------------------------------------------------------------------
+
+namespace {
+/// CRTP base class for Python MLIR values that subclass Value and should be
+/// castable from it. The value hierarchy is one level deep and is not supposed
+/// to accommodate other levels unless core MLIR changes.
+template <typename DerivedTy>
+class PyConcreteValue : public PyValue {
+public:
+  // Derived classes must define statics for:
+  //   IsAFunctionTy isaFunction
+  //   const char *pyClassName
+  // and redefine bindDerived.
+  using ClassTy = py::class_<DerivedTy, PyValue>;
+  using IsAFunctionTy = int (*)(MlirValue);
+
+  PyConcreteValue() = default;
+  PyConcreteValue(PyOperationRef operationRef, MlirValue value)
+      : PyValue(operationRef, value) {}
+  PyConcreteValue(PyValue &orig)
+      : PyConcreteValue(orig.getParentOperation(), castFrom(orig)) {}
+
+  /// Attempts to cast the original value to the derived type and throws on
+  /// type mismatches.
+  static MlirValue castFrom(PyValue &orig) {
+    if (!DerivedTy::isaFunction(orig.get())) {
+      auto origRepr = py::repr(py::cast(orig)).cast<std::string>();
+      throw SetPyError(PyExc_ValueError, llvm::Twine("Cannot cast value to ") +
+                                             DerivedTy::pyClassName +
+                                             " (from " + origRepr + ")");
+    }
+    return orig.get();
+  }
+
+  /// Binds the Python module objects to functions of this class.
+  static void bind(py::module &m) {
+    auto cls = ClassTy(m, DerivedTy::pyClassName);
+    cls.def(py::init<PyValue &>(), py::keep_alive<0, 1>());
+    DerivedTy::bindDerived(cls);
+  }
+
+  /// Implemented by derived classes to add methods to the Python subclass.
+  static void bindDerived(ClassTy &m) {}
+};
+
+/// Python wrapper for MlirBlockArgument.
+class PyBlockArgument : public PyConcreteValue<PyBlockArgument> {
+public:
+  static constexpr IsAFunctionTy isaFunction = mlirValueIsABlockArgument;
+  static constexpr const char *pyClassName = "BlockArgument";
+  using PyConcreteValue::PyConcreteValue;
+
+  static void bindDerived(ClassTy &c) {
+    c.def_property_readonly("owner", [](PyBlockArgument &self) {
+      return PyBlock(self.getParentOperation(),
+                     mlirBlockArgumentGetOwner(self.get()));
+    });
+    c.def_property_readonly("arg_number", [](PyBlockArgument &self) {
+      return mlirBlockArgumentGetArgNumber(self.get());
+    });
+    c.def("set_type", [](PyBlockArgument &self, PyType type) {
+      return mlirBlockArgumentSetType(self.get(), type);
+    });
+  }
+};
+
+/// Python wrapper for MlirOpResult.
+class PyOpResult : public PyConcreteValue<PyOpResult> {
+public:
+  static constexpr IsAFunctionTy isaFunction = mlirValueIsAOpResult;
+  static constexpr const char *pyClassName = "OpResult";
+  using PyConcreteValue::PyConcreteValue;
+
+  static void bindDerived(ClassTy &c) {
+    c.def_property_readonly("owner", [](PyOpResult &self) {
+      assert(
+          mlirOperationEqual(self.getParentOperation()->get(),
+                             mlirOpResultGetOwner(self.get())) &&
+          "expected the owner of the value in Python to match that in the IR");
+      return self.getParentOperation();
+    });
+    c.def_property_readonly("result_number", [](PyOpResult &self) {
+      return mlirOpResultGetResultNumber(self.get());
+    });
+  }
+};
+
+/// A list of block arguments. Internally, these are stored as consecutive
+/// elements, random access is cheap. The argument list is associated with the
+/// operation that contains the block (detached blocks are not allowed in
+/// Python bindings) and extends its lifetime.
+class PyBlockArgumentList {
+public:
+  PyBlockArgumentList(PyOperationRef operation, MlirBlock block)
+      : operation(std::move(operation)), block(block) {}
+
+  /// Returns the length of the block argument list.
+  intptr_t dunderLen() {
+    operation->checkValid();
+    return mlirBlockGetNumArguments(block);
+  }
+
+  /// Returns `index`-th element of the block argument list.
+  PyBlockArgument dunderGetItem(intptr_t index) {
+    if (index < 0 || index >= dunderLen()) {
+      throw SetPyError(PyExc_IndexError,
+                       "attempt to access out of bounds region");
+    }
+    PyValue value(operation, mlirBlockGetArgument(block, index));
+    return PyBlockArgument(value);
+  }
+
+  /// Defines a Python class in the bindings.
+  static void bind(py::module &m) {
+    py::class_<PyBlockArgumentList>(m, "BlockArgumentList")
+        .def("__len__", &PyBlockArgumentList::dunderLen)
+        .def("__getitem__", &PyBlockArgumentList::dunderGetItem);
+  }
+
+private:
+  PyOperationRef operation;
+  MlirBlock block;
+};
+
+/// A list of operation results. Internally, these are stored as consecutive
+/// elements, random access is cheap. The result list is associated with the
+/// operation whose results these are, and extends the lifetime of this
+/// operation.
+class PyOpResultList {
+public:
+  PyOpResultList(PyOperationRef operation) : operation(operation) {}
+
+  /// Returns the length of the result list.
+  intptr_t dunderLen() {
+    operation->checkValid();
+    return mlirOperationGetNumResults(operation->get());
+  }
+
+  /// Returns `index`-th element in the result list.
+  PyOpResult dunderGetItem(intptr_t index) {
+    if (index < 0 || index >= dunderLen()) {
+      throw SetPyError(PyExc_IndexError,
+                       "attempt to access out of bounds region");
+    }
+    PyValue value(operation, mlirOperationGetResult(operation->get(), index));
+    return PyOpResult(value);
+  }
+
+  /// Defines a Python class in the bindings.
+  static void bind(py::module &m) {
+    py::class_<PyOpResultList>(m, "OpResultList")
+        .def("__len__", &PyOpResultList::dunderLen)
+        .def("__getitem__", &PyOpResultList::dunderGetItem);
+  }
+
+private:
+  PyOperationRef operation;
+};
+
+} // end namespace
+
 //------------------------------------------------------------------------------
 // Standard attribute subclasses.
 //------------------------------------------------------------------------------
@@ -749,7 +1020,7 @@ class PyConcreteAttribute : public BaseTy {
   // Derived classes must define statics for:
   //   IsAFunctionTy isaFunction
   //   const char *pyClassName
-  using ClassTy = py::class_<DerivedTy, PyAttribute>;
+  using ClassTy = py::class_<DerivedTy, BaseTy>;
   using IsAFunctionTy = int (*)(MlirAttribute);
 
   PyConcreteAttribute() = default;
@@ -1793,18 +2064,39 @@ void mlir::python::populateIRSubmodule(py::module &m) {
       .def_property_readonly(
           "regions",
           [](PyOperation &self) { return PyRegionList(self.getRef()); })
+      .def_property_readonly(
+          "results",
+          [](PyOperation &self) { return PyOpResultList(self.getRef()); },
+          "Returns the list of Operation results.")
       .def("__iter__",
            [](PyOperation &self) { return PyRegionIterator(self.getRef()); })
       .def(
           "__str__",
           [](PyOperation &self) {
-            self.checkValid();
-            PyPrintAccumulator printAccum;
-            mlirOperationPrint(self.get(), printAccum.getCallback(),
-                               printAccum.getUserData());
-            return printAccum.join();
+            return self.getAsm(/*binary=*/false,
+                               /*largeElementsLimit=*/llvm::None,
+                               /*enableDebugInfo=*/false,
+                               /*prettyDebugInfo=*/false,
+                               /*printGenericOpForm=*/false,
+                               /*useLocalScope=*/false);
           },
-          kTypeStrDunderDocstring);
+          "Returns the assembly form of the operation.")
+      .def("print", &PyOperation::print,
+           // Careful: Lots of arguments must match up with print method.
+           py::arg("file") = py::none(), py::arg("binary") = false,
+           py::arg("large_elements_limit") = py::none(),
+           py::arg("enable_debug_info") = false,
+           py::arg("pretty_debug_info") = false,
+           py::arg("print_generic_op_form") = false,
+           py::arg("use_local_scope") = false, kOperationPrintDocstring)
+      .def("get_asm", &PyOperation::getAsm,
+           // Careful: Lots of arguments must match up with get_asm method.
+           py::arg("binary") = false,
+           py::arg("large_elements_limit") = py::none(),
+           py::arg("enable_debug_info") = false,
+           py::arg("pretty_debug_info") = false,
+           py::arg("print_generic_op_form") = false,
+           py::arg("use_local_scope") = false, kOperationGetAsmDocstring);
 
   // Mapping of PyRegion.
   py::class_<PyRegion>(m, "Region")
@@ -1833,6 +2125,12 @@ void mlir::python::populateIRSubmodule(py::module &m) {
 
   // Mapping of PyBlock.
   py::class_<PyBlock>(m, "Block")
+      .def_property_readonly(
+          "arguments",
+          [](PyBlock &self) {
+            return PyBlockArgumentList(self.getParentOperation(), self.get());
+          },
+          "Returns a list of block arguments.")
       .def_property_readonly(
           "operations",
           [](PyBlock &self) {
@@ -1867,9 +2165,9 @@ void mlir::python::populateIRSubmodule(py::module &m) {
                            printAccum.getUserData());
             return printAccum.join();
           },
-          kTypeStrDunderDocstring);
+          "Returns the assembly form of the block.");
 
-  // Mapping of Type.
+  // Mapping of PyAttribute.
   py::class_<PyAttribute>(m, "Attribute")
       .def_property_readonly(
           "context",
@@ -1906,7 +2204,7 @@ void mlir::python::populateIRSubmodule(py::module &m) {
                                printAccum.getUserData());
             return printAccum.join();
           },
-          kTypeStrDunderDocstring)
+          "Returns the assembly form of the Attribute.")
       .def("__repr__", [](PyAttribute &self) {
         // Generally, assembly formats are not printed for __repr__ because
         // this can cause exceptionally long debug output and exceptions.
@@ -1959,7 +2257,7 @@ void mlir::python::populateIRSubmodule(py::module &m) {
   PyStringAttribute::bind(m);
   PyDenseElementsAttribute::bind(m);
 
-  // Mapping of Type.
+  // Mapping of PyType.
   py::class_<PyType>(m, "Type")
       .def_property_readonly(
           "context", [](PyType &self) { return self.getContext().getObject(); },
@@ -1983,7 +2281,7 @@ void mlir::python::populateIRSubmodule(py::module &m) {
                           printAccum.getUserData());
             return printAccum.join();
           },
-          kTypeStrDunderDocstring)
+          "Returns the assembly form of the type.")
       .def("__repr__", [](PyType &self) {
         // Generally, assembly formats are not printed for __repr__ because
         // this can cause exceptionally long debug output and exceptions.
@@ -2015,11 +2313,40 @@ void mlir::python::populateIRSubmodule(py::module &m) {
   PyTupleType::bind(m);
   PyFunctionType::bind(m);
 
+  // Mapping of Value.
+  py::class_<PyValue>(m, "Value")
+      .def_property_readonly(
+          "context",
+          [](PyValue &self) { return self.getParentOperation()->getContext(); },
+          "Context in which the value lives.")
+      .def(
+          "dump", [](PyValue &self) { mlirValueDump(self.get()); },
+          kDumpDocstring)
+      .def(
+          "__str__",
+          [](PyValue &self) {
+            PyPrintAccumulator printAccum;
+            printAccum.parts.append("Value(");
+            mlirValuePrint(self.get(), printAccum.getCallback(),
+                           printAccum.getUserData());
+            printAccum.parts.append(")");
+            return printAccum.join();
+          },
+          kValueDunderStrDocstring)
+      .def_property_readonly("type", [](PyValue &self) {
+        return PyType(self.getParentOperation()->getContext(),
+                      mlirValueGetType(self.get()));
+      });
+  PyBlockArgument::bind(m);
+  PyOpResult::bind(m);
+
   // Container bindings.
+  PyBlockArgumentList::bind(m);
   PyBlockIterator::bind(m);
   PyBlockList::bind(m);
   PyOperationIterator::bind(m);
   PyOperationList::bind(m);
+  PyOpResultList::bind(m);
   PyRegionIterator::bind(m);
   PyRegionList::bind(m);
 }
diff --git a/mlir/lib/Bindings/Python/IRModules.h b/mlir/lib/Bindings/Python/IRModules.h
index c175018c8bb65..b438e8ac408df 100644
--- a/mlir/lib/Bindings/Python/IRModules.h
+++ b/mlir/lib/Bindings/Python/IRModules.h
@@ -23,6 +23,7 @@ class PyMlirContext;
 class PyModule;
 class PyOperation;
 class PyType;
+class PyValue;
 
 /// Template for a reference to a concrete type which captures a python
 /// reference to its underlying python object.
@@ -276,6 +277,15 @@ class PyOperation : public BaseContextObject {
   }
   void checkValid();
 
+  /// Implements the bound 'print' method and helps with others.
+  void print(pybind11::object fileObject, bool binary,
+             llvm::Optional<int64_t> largeElementsLimit, bool enableDebugInfo,
+             bool prettyDebugInfo, bool printGenericOpForm, bool useLocalScope);
+  pybind11::object getAsm(bool binary,
+                          llvm::Optional<int64_t> largeElementsLimit,
+                          bool enableDebugInfo, bool prettyDebugInfo,
+                          bool printGenericOpForm, bool useLocalScope);
+
 private:
   PyOperation(PyMlirContextRef contextRef, MlirOperation operation);
   static PyOperationRef createInstance(PyMlirContextRef contextRef,
@@ -381,6 +391,27 @@ class PyType : public BaseContextObject {
   MlirType type;
 };
 
+/// Wrapper around the generic MlirValue.
+/// Values are managed completely by the operation that resulted in their
+/// definition. For op result value, this is the operation that defines the
+/// value. For block argument values, this is the operation that contains the
+/// block to which the value is an argument (blocks cannot be detached in Python
+/// bindings so such operation always exists).
+class PyValue {
+public:
+  PyValue(PyOperationRef parentOperation, MlirValue value)
+      : parentOperation(parentOperation), value(value) {}
+
+  MlirValue get() { return value; }
+  PyOperationRef &getParentOperation() { return parentOperation; }
+
+  void checkValid() { return parentOperation->checkValid(); }
+
+private:
+  PyOperationRef parentOperation;
+  MlirValue value;
+};
+
 void populateIRSubmodule(pybind11::module &m);
 
 } // namespace python
diff --git a/mlir/lib/CAPI/IR/IR.cpp b/mlir/lib/CAPI/IR/IR.cpp
index 4bae43c424fdd..379770c8962f8 100644
--- a/mlir/lib/CAPI/IR/IR.cpp
+++ b/mlir/lib/CAPI/IR/IR.cpp
@@ -74,6 +74,36 @@ MlirStringRef mlirDialectGetNamespace(MlirDialect dialect) {
   return wrap(unwrap(dialect)->getNamespace());
 }
 
+/* ========================================================================== */
+/* Printing flags API.                                                        */
+/* ========================================================================== */
+
+MlirOpPrintingFlags mlirOpPrintingFlagsCreate() {
+  return wrap(new OpPrintingFlags());
+}
+
+void mlirOpPrintingFlagsDestroy(MlirOpPrintingFlags flags) {
+  delete unwrap(flags);
+}
+
+void mlirOpPrintingFlagsElideLargeElementsAttrs(MlirOpPrintingFlags flags,
+                                                intptr_t largeElementLimit) {
+  unwrap(flags)->elideLargeElementsAttrs(largeElementLimit);
+}
+
+void mlirOpPrintingFlagsEnableDebugInfo(MlirOpPrintingFlags flags,
+                                        int prettyForm) {
+  unwrap(flags)->enableDebugInfo(/*prettyForm=*/prettyForm);
+}
+
+void mlirOpPrintingFlagsPrintGenericOpForm(MlirOpPrintingFlags flags) {
+  unwrap(flags)->printGenericOpForm();
+}
+
+void mlirOpPrintingFlagsUseLocalScope(MlirOpPrintingFlags flags) {
+  unwrap(flags)->useLocalScope();
+}
+
 /* ========================================================================== */
 /* Location API.                                                              */
 /* ========================================================================== */
@@ -282,6 +312,13 @@ void mlirOperationPrint(MlirOperation op, MlirStringCallback callback,
   stream.flush();
 }
 
+void mlirOperationPrintWithFlags(MlirOperation op, MlirOpPrintingFlags flags,
+                                 MlirStringCallback callback, void *userData) {
+  detail::CallbackOstream stream(callback, userData);
+  unwrap(op)->print(stream, *unwrap(flags));
+  stream.flush();
+}
+
 void mlirOperationDump(MlirOperation op) { return unwrap(op)->dump(); }
 
 /* ========================================================================== */
@@ -454,6 +491,8 @@ MlirType mlirValueGetType(MlirValue value) {
   return wrap(unwrap(value).getType());
 }
 
+void mlirValueDump(MlirValue value) { unwrap(value).dump(); }
+
 void mlirValuePrint(MlirValue value, MlirStringCallback callback,
                     void *userData) {
   detail::CallbackOstream stream(callback, userData);
diff --git a/mlir/lib/Conversion/AsyncToLLVM/AsyncToLLVM.cpp b/mlir/lib/Conversion/AsyncToLLVM/AsyncToLLVM.cpp
new file mode 100644
index 0000000000000..13498e821132a
--- /dev/null
+++ b/mlir/lib/Conversion/AsyncToLLVM/AsyncToLLVM.cpp
@@ -0,0 +1,733 @@
+//===- AsyncToLLVM.cpp - Convert Async to LLVM dialect --------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Conversion/AsyncToLLVM/AsyncToLLVM.h"
+
+#include "../PassDetail.h"
+#include "mlir/Dialect/Async/IR/Async.h"
+#include "mlir/Dialect/LLVMIR/LLVMDialect.h"
+#include "mlir/Dialect/StandardOps/IR/Ops.h"
+#include "mlir/IR/BlockAndValueMapping.h"
+#include "mlir/IR/Builders.h"
+#include "mlir/IR/TypeUtilities.h"
+#include "mlir/Pass/Pass.h"
+#include "mlir/Transforms/DialectConversion.h"
+#include "mlir/Transforms/RegionUtils.h"
+#include "llvm/ADT/SetVector.h"
+#include "llvm/Support/FormatVariadic.h"
+
+#define DEBUG_TYPE "convert-async-to-llvm"
+
+using namespace mlir;
+using namespace mlir::async;
+
+// Prefix for functions outlined from `async.execute` op regions.
+static constexpr const char kAsyncFnPrefix[] = "async_execute_fn";
+
+//===----------------------------------------------------------------------===//
+// Async Runtime C API declaration.
+//===----------------------------------------------------------------------===//
+
+static constexpr const char *kCreateToken = "mlirAsyncRuntimeCreateToken";
+static constexpr const char *kEmplaceToken = "mlirAsyncRuntimeEmplaceToken";
+static constexpr const char *kAwaitToken = "mlirAsyncRuntimeAwaitToken";
+static constexpr const char *kExecute = "mlirAsyncRuntimeExecute";
+static constexpr const char *kAwaitAndExecute =
+    "mlirAsyncRuntimeAwaitTokenAndExecute";
+
+namespace {
+// Async Runtime API function types.
+struct AsyncAPI {
+  static FunctionType createTokenFunctionType(MLIRContext *ctx) {
+    return FunctionType::get({}, {TokenType::get(ctx)}, ctx);
+  }
+
+  static FunctionType emplaceTokenFunctionType(MLIRContext *ctx) {
+    return FunctionType::get({TokenType::get(ctx)}, {}, ctx);
+  }
+
+  static FunctionType awaitTokenFunctionType(MLIRContext *ctx) {
+    return FunctionType::get({TokenType::get(ctx)}, {}, ctx);
+  }
+
+  static FunctionType executeFunctionType(MLIRContext *ctx) {
+    auto hdl = LLVM::LLVMType::getInt8PtrTy(ctx);
+    auto resume = resumeFunctionType(ctx).getPointerTo();
+    return FunctionType::get({hdl, resume}, {}, ctx);
+  }
+
+  static FunctionType awaitAndExecuteFunctionType(MLIRContext *ctx) {
+    auto hdl = LLVM::LLVMType::getInt8PtrTy(ctx);
+    auto resume = resumeFunctionType(ctx).getPointerTo();
+    return FunctionType::get({TokenType::get(ctx), hdl, resume}, {}, ctx);
+  }
+
+  // Auxiliary coroutine resume intrinsic wrapper.
+  static LLVM::LLVMType resumeFunctionType(MLIRContext *ctx) {
+    auto voidTy = LLVM::LLVMType::getVoidTy(ctx);
+    auto i8Ptr = LLVM::LLVMType::getInt8PtrTy(ctx);
+    return LLVM::LLVMType::getFunctionTy(voidTy, {i8Ptr}, false);
+  }
+};
+} // namespace
+
+// Adds Async Runtime C API declarations to the module.
+static void addAsyncRuntimeApiDeclarations(ModuleOp module) {
+  auto builder = OpBuilder::atBlockTerminator(module.getBody());
+
+  MLIRContext *ctx = module.getContext();
+  Location loc = module.getLoc();
+
+  if (!module.lookupSymbol(kCreateToken))
+    builder.create<FuncOp>(loc, kCreateToken,
+                           AsyncAPI::createTokenFunctionType(ctx));
+
+  if (!module.lookupSymbol(kEmplaceToken))
+    builder.create<FuncOp>(loc, kEmplaceToken,
+                           AsyncAPI::emplaceTokenFunctionType(ctx));
+
+  if (!module.lookupSymbol(kAwaitToken))
+    builder.create<FuncOp>(loc, kAwaitToken,
+                           AsyncAPI::awaitTokenFunctionType(ctx));
+
+  if (!module.lookupSymbol(kExecute))
+    builder.create<FuncOp>(loc, kExecute, AsyncAPI::executeFunctionType(ctx));
+
+  if (!module.lookupSymbol(kAwaitAndExecute))
+    builder.create<FuncOp>(loc, kAwaitAndExecute,
+                           AsyncAPI::awaitAndExecuteFunctionType(ctx));
+}
+
+//===----------------------------------------------------------------------===//
+// LLVM coroutines intrinsics declarations.
+//===----------------------------------------------------------------------===//
+
+static constexpr const char *kCoroId = "llvm.coro.id";
+static constexpr const char *kCoroSizeI64 = "llvm.coro.size.i64";
+static constexpr const char *kCoroBegin = "llvm.coro.begin";
+static constexpr const char *kCoroSave = "llvm.coro.save";
+static constexpr const char *kCoroSuspend = "llvm.coro.suspend";
+static constexpr const char *kCoroEnd = "llvm.coro.end";
+static constexpr const char *kCoroFree = "llvm.coro.free";
+static constexpr const char *kCoroResume = "llvm.coro.resume";
+
+/// Adds coroutine intrinsics declarations to the module.
+static void addCoroutineIntrinsicsDeclarations(ModuleOp module) {
+  using namespace mlir::LLVM;
+
+  MLIRContext *ctx = module.getContext();
+  Location loc = module.getLoc();
+
+  OpBuilder builder(module.getBody()->getTerminator());
+
+  auto token = LLVMTokenType::get(ctx);
+  auto voidTy = LLVMType::getVoidTy(ctx);
+
+  auto i8 = LLVMType::getInt8Ty(ctx);
+  auto i1 = LLVMType::getInt1Ty(ctx);
+  auto i32 = LLVMType::getInt32Ty(ctx);
+  auto i64 = LLVMType::getInt64Ty(ctx);
+  auto i8Ptr = LLVMType::getInt8PtrTy(ctx);
+
+  if (!module.lookupSymbol(kCoroId))
+    builder.create<LLVMFuncOp>(
+        loc, kCoroId,
+        LLVMType::getFunctionTy(token, {i32, i8Ptr, i8Ptr, i8Ptr}, false));
+
+  if (!module.lookupSymbol(kCoroSizeI64))
+    builder.create<LLVMFuncOp>(loc, kCoroSizeI64,
+                               LLVMType::getFunctionTy(i64, false));
+
+  if (!module.lookupSymbol(kCoroBegin))
+    builder.create<LLVMFuncOp>(
+        loc, kCoroBegin, LLVMType::getFunctionTy(i8Ptr, {token, i8Ptr}, false));
+
+  if (!module.lookupSymbol(kCoroSave))
+    builder.create<LLVMFuncOp>(loc, kCoroSave,
+                               LLVMType::getFunctionTy(token, i8Ptr, false));
+
+  if (!module.lookupSymbol(kCoroSuspend))
+    builder.create<LLVMFuncOp>(loc, kCoroSuspend,
+                               LLVMType::getFunctionTy(i8, {token, i1}, false));
+
+  if (!module.lookupSymbol(kCoroEnd))
+    builder.create<LLVMFuncOp>(loc, kCoroEnd,
+                               LLVMType::getFunctionTy(i1, {i8Ptr, i1}, false));
+
+  if (!module.lookupSymbol(kCoroFree))
+    builder.create<LLVMFuncOp>(
+        loc, kCoroFree, LLVMType::getFunctionTy(i8Ptr, {token, i8Ptr}, false));
+
+  if (!module.lookupSymbol(kCoroResume))
+    builder.create<LLVMFuncOp>(loc, kCoroResume,
+                               LLVMType::getFunctionTy(voidTy, i8Ptr, false));
+}
+
+//===----------------------------------------------------------------------===//
+// Add malloc/free declarations to the module.
+//===----------------------------------------------------------------------===//
+
+static constexpr const char *kMalloc = "malloc";
+static constexpr const char *kFree = "free";
+
+/// Adds malloc/free declarations to the module.
+static void addCRuntimeDeclarations(ModuleOp module) {
+  using namespace mlir::LLVM;
+
+  MLIRContext *ctx = module.getContext();
+  Location loc = module.getLoc();
+
+  OpBuilder builder(module.getBody()->getTerminator());
+
+  auto voidTy = LLVMType::getVoidTy(ctx);
+  auto i64 = LLVMType::getInt64Ty(ctx);
+  auto i8Ptr = LLVMType::getInt8PtrTy(ctx);
+
+  if (!module.lookupSymbol(kMalloc))
+    builder.create<LLVM::LLVMFuncOp>(
+        loc, kMalloc, LLVMType::getFunctionTy(i8Ptr, {i64}, false));
+
+  if (!module.lookupSymbol(kFree))
+    builder.create<LLVM::LLVMFuncOp>(
+        loc, kFree, LLVMType::getFunctionTy(voidTy, i8Ptr, false));
+}
+
+//===----------------------------------------------------------------------===//
+// Coroutine resume function wrapper.
+//===----------------------------------------------------------------------===//
+
+static constexpr const char *kResume = "__resume";
+
+// A function that takes a coroutine handle and calls a `llvm.coro.resume`
+// intrinsics. We need this function to be able to pass it to the async
+// runtime execute API.
+static void addResumeFunction(ModuleOp module) {
+  MLIRContext *ctx = module.getContext();
+
+  OpBuilder moduleBuilder(module.getBody()->getTerminator());
+  Location loc = module.getLoc();
+
+  if (module.lookupSymbol(kResume))
+    return;
+
+  auto voidTy = LLVM::LLVMType::getVoidTy(ctx);
+  auto i8Ptr = LLVM::LLVMType::getInt8PtrTy(ctx);
+
+  auto resumeOp = moduleBuilder.create<LLVM::LLVMFuncOp>(
+      loc, kResume, LLVM::LLVMFunctionType::get(voidTy, {i8Ptr}));
+  SymbolTable::setSymbolVisibility(resumeOp, SymbolTable::Visibility::Private);
+
+  auto *block = resumeOp.addEntryBlock();
+  OpBuilder blockBuilder = OpBuilder::atBlockEnd(block);
+
+  blockBuilder.create<LLVM::CallOp>(loc, Type(),
+                                    blockBuilder.getSymbolRefAttr(kCoroResume),
+                                    resumeOp.getArgument(0));
+
+  blockBuilder.create<LLVM::ReturnOp>(loc, ValueRange());
+}
+
+//===----------------------------------------------------------------------===//
+// async.execute op outlining to the coroutine functions.
+//===----------------------------------------------------------------------===//
+
+// Function targeted for coroutine transformation has two additional blocks at
+// the end: coroutine cleanup and coroutine suspension.
+//
+// async.await op lowering additionaly creates a resume block for each
+// operation to enable non-blocking waiting via coroutine suspension.
+namespace {
+struct CoroMachinery {
+  Value asyncToken;
+  Value coroHandle;
+  Block *cleanup;
+  Block *suspend;
+};
+} // namespace
+
+// Builds an coroutine template compatible with LLVM coroutines lowering.
+//
+//  - `entry` block sets up the coroutine.
+//  - `cleanup` block cleans up the coroutine state.
+//  - `suspend block after the @llvm.coro.end() defines what value will be
+//    returned to the initial caller of a coroutine. Everything before the
+//    @llvm.coro.end() will be executed at every suspension point.
+//
+// Coroutine structure (only the important bits):
+//
+//   func @async_execute_fn(<function-arguments>) -> !async.token {
+//     ^entryBlock(<function-arguments>):
+//       %token = <async token> : !async.token // create async runtime token
+//       %hdl = llvm.call @llvm.coro.id(...)   // create a coroutine handle
+//       br ^cleanup
+//
+//     ^cleanup:
+//       llvm.call @llvm.coro.free(...)        // delete coroutine state
+//       br ^suspend
+//
+//     ^suspend:
+//       llvm.call @llvm.coro.end(...)         // marks the end of a coroutine
+//       return %token : !async.token
+//   }
+//
+// The actual code for the async.execute operation body region will be inserted
+// before the entry block terminator.
+//
+//
+static CoroMachinery setupCoroMachinery(FuncOp func) {
+  assert(func.getBody().empty() && "Function must have empty body");
+
+  MLIRContext *ctx = func.getContext();
+
+  auto token = LLVM::LLVMTokenType::get(ctx);
+  auto i1 = LLVM::LLVMType::getInt1Ty(ctx);
+  auto i32 = LLVM::LLVMType::getInt32Ty(ctx);
+  auto i64 = LLVM::LLVMType::getInt64Ty(ctx);
+  auto i8Ptr = LLVM::LLVMType::getInt8PtrTy(ctx);
+
+  Block *entryBlock = func.addEntryBlock();
+  Location loc = func.getBody().getLoc();
+
+  OpBuilder builder = OpBuilder::atBlockBegin(entryBlock);
+
+  // ------------------------------------------------------------------------ //
+  // Allocate async tokens/values that we will return from a ramp function.
+  // ------------------------------------------------------------------------ //
+  auto createToken =
+      builder.create<CallOp>(loc, kCreateToken, TokenType::get(ctx));
+
+  // ------------------------------------------------------------------------ //
+  // Initialize coroutine: allocate frame, get coroutine handle.
+  // ------------------------------------------------------------------------ //
+
+  // Constants for initializing coroutine frame.
+  auto constZero =
+      builder.create<LLVM::ConstantOp>(loc, i32, builder.getI32IntegerAttr(0));
+  auto constFalse =
+      builder.create<LLVM::ConstantOp>(loc, i1, builder.getBoolAttr(false));
+  auto nullPtr = builder.create<LLVM::NullOp>(loc, i8Ptr);
+
+  // Get coroutine id: @llvm.coro.id
+  auto coroId = builder.create<LLVM::CallOp>(
+      loc, token, builder.getSymbolRefAttr(kCoroId),
+      ValueRange({constZero, nullPtr, nullPtr, nullPtr}));
+
+  // Get coroutine frame size: @llvm.coro.size.i64
+  auto coroSize = builder.create<LLVM::CallOp>(
+      loc, i64, builder.getSymbolRefAttr(kCoroSizeI64), ValueRange());
+
+  // Allocate memory for coroutine frame.
+  auto coroAlloc = builder.create<LLVM::CallOp>(
+      loc, i8Ptr, builder.getSymbolRefAttr(kMalloc),
+      ValueRange(coroSize.getResult(0)));
+
+  // Begin a coroutine: @llvm.coro.begin
+  auto coroHdl = builder.create<LLVM::CallOp>(
+      loc, i8Ptr, builder.getSymbolRefAttr(kCoroBegin),
+      ValueRange({coroId.getResult(0), coroAlloc.getResult(0)}));
+
+  Block *cleanupBlock = func.addBlock();
+  Block *suspendBlock = func.addBlock();
+
+  // ------------------------------------------------------------------------ //
+  // Coroutine cleanup block: deallocate coroutine frame, free the memory.
+  // ------------------------------------------------------------------------ //
+  builder.setInsertionPointToStart(cleanupBlock);
+
+  // Get a pointer to the coroutine frame memory: @llvm.coro.free.
+  auto coroMem = builder.create<LLVM::CallOp>(
+      loc, i8Ptr, builder.getSymbolRefAttr(kCoroFree),
+      ValueRange({coroId.getResult(0), coroHdl.getResult(0)}));
+
+  // Free the memory.
+  builder.create<LLVM::CallOp>(loc, Type(), builder.getSymbolRefAttr(kFree),
+                               ValueRange(coroMem.getResult(0)));
+  // Branch into the suspend block.
+  builder.create<BranchOp>(loc, suspendBlock);
+
+  // ------------------------------------------------------------------------ //
+  // Coroutine suspend block: mark the end of a coroutine and return allocated
+  // async token.
+  // ------------------------------------------------------------------------ //
+  builder.setInsertionPointToStart(suspendBlock);
+
+  // Mark the end of a coroutine: @llvm.coro.end.
+  builder.create<LLVM::CallOp>(loc, i1, builder.getSymbolRefAttr(kCoroEnd),
+                               ValueRange({coroHdl.getResult(0), constFalse}));
+
+  // Return created `async.token` from the suspend block. This will be the
+  // return value of a coroutine ramp function.
+  builder.create<ReturnOp>(loc, createToken.getResult(0));
+
+  // Branch from the entry block to the cleanup block to create a valid CFG.
+  builder.setInsertionPointToEnd(entryBlock);
+
+  builder.create<BranchOp>(loc, cleanupBlock);
+
+  // `async.await` op lowering will create resume blocks for async
+  // continuations, and will conditionally branch to cleanup or suspend blocks.
+
+  return {createToken.getResult(0), coroHdl.getResult(0), cleanupBlock,
+          suspendBlock};
+}
+
+// Adds a suspension point before the `op`, and moves `op` and all operations
+// after it into the resume block. Returns a pointer to the resume block.
+//
+// `coroState` must be a value returned from the call to @llvm.coro.save(...)
+// intrinsic (saved coroutine state).
+//
+// Before:
+//
+//   ^bb0:
+//     "opBefore"(...)
+//     "op"(...)
+//   ^cleanup: ...
+//   ^suspend: ...
+//
+// After:
+//
+//   ^bb0:
+//     "opBefore"(...)
+//     %suspend = llmv.call @llvm.coro.suspend(...)
+//     switch %suspend [-1: ^suspend, 0: ^resume, 1: ^cleanup]
+//   ^resume:
+//     "op"(...)
+//   ^cleanup: ...
+//   ^suspend: ...
+//
+static Block *addSuspensionPoint(CoroMachinery coro, Value coroState,
+                                 Operation *op) {
+  MLIRContext *ctx = op->getContext();
+  auto i1 = LLVM::LLVMType::getInt1Ty(ctx);
+  auto i8 = LLVM::LLVMType::getInt8Ty(ctx);
+
+  Location loc = op->getLoc();
+  Block *splitBlock = op->getBlock();
+
+  // Split the block before `op`, newly added block is the resume block.
+  Block *resume = splitBlock->splitBlock(op);
+
+  // Add a coroutine suspension in place of original `op` in the split block.
+  OpBuilder builder = OpBuilder::atBlockEnd(splitBlock);
+
+  auto constFalse =
+      builder.create<LLVM::ConstantOp>(loc, i1, builder.getBoolAttr(false));
+
+  // Suspend a coroutine: @llvm.coro.suspend
+  auto coroSuspend = builder.create<LLVM::CallOp>(
+      loc, i8, builder.getSymbolRefAttr(kCoroSuspend),
+      ValueRange({coroState, constFalse}));
+
+  // After a suspension point decide if we should branch into resume, cleanup
+  // or suspend block of the coroutine (see @llvm.coro.suspend return code
+  // documentation).
+  auto constZero =
+      builder.create<LLVM::ConstantOp>(loc, i8, builder.getI8IntegerAttr(0));
+  auto constNegOne =
+      builder.create<LLVM::ConstantOp>(loc, i8, builder.getI8IntegerAttr(-1));
+
+  Block *resumeOrCleanup = builder.createBlock(resume);
+
+  // Suspend the coroutine ...?
+  builder.setInsertionPointToEnd(splitBlock);
+  auto isNegOne = builder.create<LLVM::ICmpOp>(
+      loc, LLVM::ICmpPredicate::eq, coroSuspend.getResult(0), constNegOne);
+  builder.create<LLVM::CondBrOp>(loc, isNegOne, /*trueDest=*/coro.suspend,
+                                 /*falseDest=*/resumeOrCleanup);
+
+  // ... or resume or cleanup the coroutine?
+  builder.setInsertionPointToStart(resumeOrCleanup);
+  auto isZero = builder.create<LLVM::ICmpOp>(
+      loc, LLVM::ICmpPredicate::eq, coroSuspend.getResult(0), constZero);
+  builder.create<LLVM::CondBrOp>(loc, isZero, /*trueDest=*/resume,
+                                 /*falseDest=*/coro.cleanup);
+
+  return resume;
+}
+
+// Outline the body region attached to the `async.execute` op into a standalone
+// function.
+static std::pair<FuncOp, CoroMachinery>
+outlineExecuteOp(SymbolTable &symbolTable, ExecuteOp execute) {
+  ModuleOp module = execute.getParentOfType<ModuleOp>();
+
+  MLIRContext *ctx = module.getContext();
+  Location loc = execute.getLoc();
+
+  OpBuilder moduleBuilder(module.getBody()->getTerminator());
+
+  // Get values captured by the async region
+  llvm::SetVector<mlir::Value> usedAbove;
+  getUsedValuesDefinedAbove(execute.body(), usedAbove);
+
+  // Collect types of the captured values.
+  auto usedAboveTypes =
+      llvm::map_range(usedAbove, [](Value value) { return value.getType(); });
+  SmallVector<Type, 4> inputTypes(usedAboveTypes.begin(), usedAboveTypes.end());
+  auto outputTypes = execute.getResultTypes();
+
+  auto funcType = moduleBuilder.getFunctionType(inputTypes, outputTypes);
+  auto funcAttrs = ArrayRef<NamedAttribute>();
+
+  // TODO: Derive outlined function name from the parent FuncOp (support
+  // multiple nested async.execute operations).
+  FuncOp func = FuncOp::create(loc, kAsyncFnPrefix, funcType, funcAttrs);
+  symbolTable.insert(func, moduleBuilder.getInsertionPoint());
+
+  SymbolTable::setSymbolVisibility(func, SymbolTable::Visibility::Private);
+
+  // Prepare a function for coroutine lowering by adding entry/cleanup/suspend
+  // blocks, adding llvm.coro instrinsics and setting up control flow.
+  CoroMachinery coro = setupCoroMachinery(func);
+
+  // Suspend async function at the end of an entry block, and resume it using
+  // Async execute API (execution will be resumed in a thread managed by the
+  // async runtime).
+  Block *entryBlock = &func.getBlocks().front();
+  OpBuilder builder = OpBuilder::atBlockTerminator(entryBlock);
+
+  // A pointer to coroutine resume intrinsic wrapper.
+  auto resumeFnTy = AsyncAPI::resumeFunctionType(ctx);
+  auto resumePtr = builder.create<LLVM::AddressOfOp>(
+      loc, resumeFnTy.getPointerTo(), kResume);
+
+  // Save the coroutine state: @llvm.coro.save
+  auto coroSave = builder.create<LLVM::CallOp>(
+      loc, LLVM::LLVMTokenType::get(ctx), builder.getSymbolRefAttr(kCoroSave),
+      ValueRange({coro.coroHandle}));
+
+  // Call async runtime API to execute a coroutine in the managed thread.
+  SmallVector<Value, 2> executeArgs = {coro.coroHandle, resumePtr.res()};
+  builder.create<CallOp>(loc, Type(), kExecute, executeArgs);
+
+  // Split the entry block before the terminator.
+  Block *resume = addSuspensionPoint(coro, coroSave.getResult(0),
+                                     entryBlock->getTerminator());
+
+  // Map from values defined above the execute op to the function arguments.
+  BlockAndValueMapping valueMapping;
+  valueMapping.map(usedAbove, func.getArguments());
+
+  // Clone all operations from the execute operation body into the outlined
+  // function body, and replace all `async.yield` operations with a call
+  // to async runtime to emplace the result token.
+  builder.setInsertionPointToStart(resume);
+  for (Operation &op : execute.body().getOps()) {
+    if (isa<async::YieldOp>(op)) {
+      builder.create<CallOp>(loc, kEmplaceToken, Type(), coro.asyncToken);
+      continue;
+    }
+    builder.clone(op, valueMapping);
+  }
+
+  // Replace the original `async.execute` with a call to outlined function.
+  OpBuilder callBuilder(execute);
+  SmallVector<Value, 4> usedAboveArgs(usedAbove.begin(), usedAbove.end());
+  auto callOutlinedFunc = callBuilder.create<CallOp>(
+      loc, func.getName(), execute.getResultTypes(), usedAboveArgs);
+  execute.replaceAllUsesWith(callOutlinedFunc.getResults());
+  execute.erase();
+
+  return {func, coro};
+}
+
+//===----------------------------------------------------------------------===//
+// Convert Async dialect types to LLVM types.
+//===----------------------------------------------------------------------===//
+
+namespace {
+class AsyncRuntimeTypeConverter : public TypeConverter {
+public:
+  AsyncRuntimeTypeConverter() { addConversion(convertType); }
+
+  static Type convertType(Type type) {
+    MLIRContext *ctx = type.getContext();
+    // Convert async tokens to opaque pointers.
+    if (type.isa<TokenType>())
+      return LLVM::LLVMType::getInt8PtrTy(ctx);
+    return type;
+  }
+};
+} // namespace
+
+//===----------------------------------------------------------------------===//
+// Convert types for all call operations to lowered async types.
+//===----------------------------------------------------------------------===//
+
+namespace {
+class CallOpOpConversion : public ConversionPattern {
+public:
+  explicit CallOpOpConversion(MLIRContext *ctx)
+      : ConversionPattern(CallOp::getOperationName(), 1, ctx) {}
+
+  LogicalResult
+  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    AsyncRuntimeTypeConverter converter;
+
+    SmallVector<Type, 5> resultTypes;
+    converter.convertTypes(op->getResultTypes(), resultTypes);
+
+    CallOp call = cast<CallOp>(op);
+    rewriter.replaceOpWithNewOp<CallOp>(op, resultTypes, call.callee(),
+                                        call.getOperands());
+
+    return success();
+  }
+};
+} // namespace
+
+//===----------------------------------------------------------------------===//
+// async.await op lowering to mlirAsyncRuntimeAwaitToken function call.
+//===----------------------------------------------------------------------===//
+
+namespace {
+class AwaitOpLowering : public ConversionPattern {
+public:
+  explicit AwaitOpLowering(
+      MLIRContext *ctx,
+      const llvm::DenseMap<FuncOp, CoroMachinery> &outlinedFunctions)
+      : ConversionPattern(AwaitOp::getOperationName(), 1, ctx),
+        outlinedFunctions(outlinedFunctions) {}
+
+  LogicalResult
+  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    // We can only await on the token operand. Async valus are not supported.
+    auto await = cast<AwaitOp>(op);
+    if (!await.operand().getType().isa<TokenType>())
+      return failure();
+
+    // Check if `async.await` is inside the outlined coroutine function.
+    auto func = await.getParentOfType<FuncOp>();
+    auto outlined = outlinedFunctions.find(func);
+    const bool isInCoroutine = outlined != outlinedFunctions.end();
+
+    Location loc = op->getLoc();
+
+    // Inside regular function we convert await operation to the blocking
+    // async API await function call.
+    if (!isInCoroutine)
+      rewriter.create<CallOp>(loc, Type(), kAwaitToken,
+                              ValueRange(op->getOperand(0)));
+
+    // Inside the coroutine we convert await operation into coroutine suspension
+    // point, and resume execution asynchronously.
+    if (isInCoroutine) {
+      const CoroMachinery &coro = outlined->getSecond();
+
+      OpBuilder builder(op);
+      MLIRContext *ctx = op->getContext();
+
+      // A pointer to coroutine resume intrinsic wrapper.
+      auto resumeFnTy = AsyncAPI::resumeFunctionType(ctx);
+      auto resumePtr = builder.create<LLVM::AddressOfOp>(
+          loc, resumeFnTy.getPointerTo(), kResume);
+
+      // Save the coroutine state: @llvm.coro.save
+      auto coroSave = builder.create<LLVM::CallOp>(
+          loc, LLVM::LLVMTokenType::get(ctx),
+          builder.getSymbolRefAttr(kCoroSave), ValueRange(coro.coroHandle));
+
+      // Call async runtime API to resume a coroutine in the managed thread when
+      // the async await argument becomes ready.
+      SmallVector<Value, 3> awaitAndExecuteArgs = {
+          await.getOperand(), coro.coroHandle, resumePtr.res()};
+      builder.create<CallOp>(loc, Type(), kAwaitAndExecute,
+                             awaitAndExecuteArgs);
+
+      // Split the entry block before the await operation.
+      addSuspensionPoint(coro, coroSave.getResult(0), op);
+    }
+
+    // Original operation was replaced by function call or suspension point.
+    rewriter.eraseOp(op);
+
+    return success();
+  }
+
+private:
+  const llvm::DenseMap<FuncOp, CoroMachinery> &outlinedFunctions;
+};
+} // namespace
+
+//===----------------------------------------------------------------------===//
+
+namespace {
+struct ConvertAsyncToLLVMPass
+    : public ConvertAsyncToLLVMBase<ConvertAsyncToLLVMPass> {
+  void runOnOperation() override;
+};
+
+void ConvertAsyncToLLVMPass::runOnOperation() {
+  ModuleOp module = getOperation();
+  SymbolTable symbolTable(module);
+
+  // Outline all `async.execute` body regions into async functions (coroutines).
+  llvm::DenseMap<FuncOp, CoroMachinery> outlinedFunctions;
+
+  WalkResult outlineResult = module.walk([&](ExecuteOp execute) {
+    // We currently do not support execute operations that take async
+    // token dependencies, async value arguments or produce async results.
+    if (!execute.dependencies().empty() || !execute.operands().empty() ||
+        !execute.results().empty()) {
+      execute.emitOpError(
+          "Can't outline async.execute op with async dependencies, arguments "
+          "or returned async results");
+      return WalkResult::interrupt();
+    }
+
+    outlinedFunctions.insert(outlineExecuteOp(symbolTable, execute));
+
+    return WalkResult::advance();
+  });
+
+  // Failed to outline all async execute operations.
+  if (outlineResult.wasInterrupted()) {
+    signalPassFailure();
+    return;
+  }
+
+  LLVM_DEBUG({
+    llvm::dbgs() << "Outlined " << outlinedFunctions.size()
+                 << " async functions\n";
+  });
+
+  // Add declarations for all functions required by the coroutines lowering.
+  addResumeFunction(module);
+  addAsyncRuntimeApiDeclarations(module);
+  addCoroutineIntrinsicsDeclarations(module);
+  addCRuntimeDeclarations(module);
+
+  MLIRContext *ctx = &getContext();
+
+  // Convert async dialect types and operations to LLVM dialect.
+  AsyncRuntimeTypeConverter converter;
+  OwningRewritePatternList patterns;
+
+  populateFuncOpTypeConversionPattern(patterns, ctx, converter);
+  patterns.insert<CallOpOpConversion>(ctx);
+  patterns.insert<AwaitOpLowering>(ctx, outlinedFunctions);
+
+  ConversionTarget target(*ctx);
+  target.addLegalDialect<LLVM::LLVMDialect>();
+  target.addIllegalDialect<AsyncDialect>();
+  target.addDynamicallyLegalOp<FuncOp>(
+      [&](FuncOp op) { return converter.isSignatureLegal(op.getType()); });
+  target.addDynamicallyLegalOp<CallOp>(
+      [&](CallOp op) { return converter.isLegal(op.getResultTypes()); });
+
+  if (failed(applyPartialConversion(module, target, patterns)))
+    signalPassFailure();
+}
+} // namespace
+
+std::unique_ptr<OperationPass<ModuleOp>> mlir::createConvertAsyncToLLVMPass() {
+  return std::make_unique<ConvertAsyncToLLVMPass>();
+}
diff --git a/mlir/lib/Conversion/AsyncToLLVM/CMakeLists.txt b/mlir/lib/Conversion/AsyncToLLVM/CMakeLists.txt
new file mode 100644
index 0000000000000..42e9021a6b8d4
--- /dev/null
+++ b/mlir/lib/Conversion/AsyncToLLVM/CMakeLists.txt
@@ -0,0 +1,16 @@
+add_mlir_conversion_library(MLIRAsyncToLLVM
+  AsyncToLLVM.cpp
+
+  ADDITIONAL_HEADER_DIRS
+  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Conversion/AsyncToLLVM
+
+  DEPENDS
+  MLIRConversionPassIncGen
+
+  LINK_COMPONENTS
+  Core
+
+  LINK_LIBS PUBLIC
+  MLIRLLVMIR
+  MLIRTransforms
+  )
diff --git a/mlir/lib/Conversion/CMakeLists.txt b/mlir/lib/Conversion/CMakeLists.txt
index dbb9ed699798d..73722bd964117 100644
--- a/mlir/lib/Conversion/CMakeLists.txt
+++ b/mlir/lib/Conversion/CMakeLists.txt
@@ -1,4 +1,5 @@
 add_subdirectory(AffineToStandard)
+add_subdirectory(AsyncToLLVM)
 add_subdirectory(AVX512ToLLVM)
 add_subdirectory(GPUCommon)
 add_subdirectory(GPUToNVVM)
diff --git a/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp b/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
index f7f5834e6351e..9d4c0c32dc82c 100644
--- a/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
+++ b/mlir/lib/Conversion/GPUCommon/ConvertLaunchFuncToRuntimeCalls.cpp
@@ -157,6 +157,34 @@ class ConvertHostRegisterOpToGpuRuntimeCallPattern
                   ConversionPatternRewriter &rewriter) const override;
 };
 
+/// A rewrite pattern to convert gpu.wait operations into a GPU runtime
+/// call. Currently it supports CUDA and ROCm (HIP).
+class ConvertWaitOpToGpuRuntimeCallPattern
+    : public ConvertOpToGpuRuntimeCallPattern<gpu::WaitOp> {
+public:
+  ConvertWaitOpToGpuRuntimeCallPattern(LLVMTypeConverter &typeConverter)
+      : ConvertOpToGpuRuntimeCallPattern<gpu::WaitOp>(typeConverter) {}
+
+private:
+  LogicalResult
+  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override;
+};
+
+/// A rewrite pattern to convert gpu.wait async operations into a GPU runtime
+/// call. Currently it supports CUDA and ROCm (HIP).
+class ConvertWaitAsyncOpToGpuRuntimeCallPattern
+    : public ConvertOpToGpuRuntimeCallPattern<gpu::WaitOp> {
+public:
+  ConvertWaitAsyncOpToGpuRuntimeCallPattern(LLVMTypeConverter &typeConverter)
+      : ConvertOpToGpuRuntimeCallPattern<gpu::WaitOp>(typeConverter) {}
+
+private:
+  LogicalResult
+  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override;
+};
+
 /// A rewrite patter to convert gpu.launch_func operations into a sequence of
 /// GPU runtime calls. Currently it supports CUDA and ROCm (HIP).
 ///
@@ -257,6 +285,69 @@ LogicalResult ConvertHostRegisterOpToGpuRuntimeCallPattern::matchAndRewrite(
   return success();
 }
 
+// Converts `gpu.wait` to runtime calls. The operands are all CUDA or ROCm
+// streams (i.e. void*). The converted op synchronizes the host with every
+// stream and then destroys it. That is, it assumes that the stream is not used
+// afterwards. In case this isn't correct, we will get a runtime error.
+// Eventually, we will have a pass that guarantees this property.
+LogicalResult ConvertWaitOpToGpuRuntimeCallPattern::matchAndRewrite(
+    Operation *op, ArrayRef<Value> operands,
+    ConversionPatternRewriter &rewriter) const {
+  if (cast<gpu::WaitOp>(op).asyncToken())
+    return failure(); // The gpu.wait is async.
+
+  Location loc = op->getLoc();
+
+  for (auto asyncDependency : operands)
+    streamSynchronizeCallBuilder.create(loc, rewriter, {asyncDependency});
+  for (auto asyncDependency : operands)
+    streamDestroyCallBuilder.create(loc, rewriter, {asyncDependency});
+
+  rewriter.eraseOp(op);
+  return success();
+}
+
+// Converts `gpu.wait async` to runtime calls. The result is a new stream that
+// is synchronized with all operands, which are CUDA or ROCm streams (i.e.
+// void*). We create and record an event after the definition of the stream
+// and make the new stream wait on that event before destroying it again. This
+// assumes that there is no other use between the definition and this op, and
+// the plan is to have a pass that guarantees this property.
+LogicalResult ConvertWaitAsyncOpToGpuRuntimeCallPattern::matchAndRewrite(
+    Operation *op, ArrayRef<Value> operands,
+    ConversionPatternRewriter &rewriter) const {
+  if (!cast<gpu::WaitOp>(op).asyncToken())
+    return failure(); // The gpu.wait is not async.
+
+  Location loc = op->getLoc();
+
+  auto insertionPoint = rewriter.saveInsertionPoint();
+  SmallVector<Value, 1> events;
+  for (auto pair : llvm::zip(op->getOperands(), operands)) {
+    auto token = std::get<0>(pair);
+    if (auto *defOp = token.getDefiningOp()) {
+      rewriter.setInsertionPointAfter(defOp);
+    } else {
+      // If we can't find the defining op, we record the event at block start,
+      // which is late and therefore misses parallelism, but still valid.
+      rewriter.setInsertionPointToStart(op->getBlock());
+    }
+    auto event = eventCreateCallBuilder.create(loc, rewriter, {}).getResult(0);
+    auto stream = std::get<1>(pair);
+    eventRecordCallBuilder.create(loc, rewriter, {event, stream});
+    events.push_back(event);
+  }
+  rewriter.restoreInsertionPoint(insertionPoint);
+  auto stream = streamCreateCallBuilder.create(loc, rewriter, {}).getResult(0);
+  for (auto event : events)
+    streamWaitEventCallBuilder.create(loc, rewriter, {stream, event});
+  for (auto event : events)
+    eventDestroyCallBuilder.create(loc, rewriter, {event});
+  rewriter.replaceOp(op, {stream});
+
+  return success();
+}
+
 // Creates a struct containing all kernel parameters on the stack and returns
 // an array of type-erased pointers to the fields of the struct. The array can
 // then be passed to the CUDA / ROCm (HIP) kernel launch calls.
@@ -411,7 +502,13 @@ mlir::createGpuToLLVMConversionPass(StringRef gpuBinaryAnnotation) {
 void mlir::populateGpuToLLVMConversionPatterns(
     LLVMTypeConverter &converter, OwningRewritePatternList &patterns,
     StringRef gpuBinaryAnnotation) {
-  patterns.insert<ConvertHostRegisterOpToGpuRuntimeCallPattern>(converter);
+  converter.addConversion(
+      [context = &converter.getContext()](gpu::AsyncTokenType type) -> Type {
+        return LLVM::LLVMType::getInt8PtrTy(context);
+      });
+  patterns.insert<ConvertHostRegisterOpToGpuRuntimeCallPattern,
+                  ConvertWaitOpToGpuRuntimeCallPattern,
+                  ConvertWaitAsyncOpToGpuRuntimeCallPattern>(converter);
   patterns.insert<ConvertLaunchFuncOpToGpuRuntimeCallPattern>(
       converter, gpuBinaryAnnotation);
   patterns.insert<EraseGpuModuleOpPattern>(&converter.getContext());
diff --git a/mlir/lib/Conversion/StandardToSPIRV/LegalizeStandardForSPIRV.cpp b/mlir/lib/Conversion/StandardToSPIRV/LegalizeStandardForSPIRV.cpp
index a2e608dcb7134..1cf3a326367c6 100644
--- a/mlir/lib/Conversion/StandardToSPIRV/LegalizeStandardForSPIRV.cpp
+++ b/mlir/lib/Conversion/StandardToSPIRV/LegalizeStandardForSPIRV.cpp
@@ -67,7 +67,8 @@ void LoadOpOfSubViewFolder<vector::TransferReadOp>::replaceOp(
     vector::TransferReadOp loadOp, SubViewOp subViewOp,
     ArrayRef<Value> sourceIndices, PatternRewriter &rewriter) const {
   rewriter.replaceOpWithNewOp<vector::TransferReadOp>(
-      loadOp, loadOp.getVectorType(), subViewOp.source(), sourceIndices);
+      loadOp, loadOp.getVectorType(), subViewOp.source(), sourceIndices,
+      loadOp.permutation_map(), loadOp.padding(), loadOp.maskedAttr());
 }
 
 template <>
@@ -84,7 +85,8 @@ void StoreOpOfSubViewFolder<vector::TransferWriteOp>::replaceOp(
     ArrayRef<Value> sourceIndices, PatternRewriter &rewriter) const {
   rewriter.replaceOpWithNewOp<vector::TransferWriteOp>(
       tranferWriteOp, tranferWriteOp.vector(), subViewOp.source(),
-      sourceIndices);
+      sourceIndices, tranferWriteOp.permutation_map(),
+      tranferWriteOp.maskedAttr());
 }
 } // namespace
 
diff --git a/mlir/lib/Dialect/GPU/IR/GPUDialect.cpp b/mlir/lib/Dialect/GPU/IR/GPUDialect.cpp
index 79fe969dbe175..7abefd7a54996 100644
--- a/mlir/lib/Dialect/GPU/IR/GPUDialect.cpp
+++ b/mlir/lib/Dialect/GPU/IR/GPUDialect.cpp
@@ -428,12 +428,11 @@ static ParseResult parseLaunchOp(OpAsmParser &parser, OperationState &result) {
 //===----------------------------------------------------------------------===//
 
 void LaunchFuncOp::build(OpBuilder &builder, OperationState &result,
-                         GPUFuncOp kernelFunc, Value gridSizeX, Value gridSizeY,
-                         Value gridSizeZ, Value blockSizeX, Value blockSizeY,
-                         Value blockSizeZ, ValueRange kernelOperands) {
+                         GPUFuncOp kernelFunc, KernelDim3 gridSize,
+                         KernelDim3 blockSize, ValueRange kernelOperands) {
   // Add grid and block sizes as op operands, followed by the data operands.
-  result.addOperands(
-      {gridSizeX, gridSizeY, gridSizeZ, blockSizeX, blockSizeY, blockSizeZ});
+  result.addOperands({gridSize.x, gridSize.y, gridSize.z, blockSize.x,
+                      blockSize.y, blockSize.z});
   result.addOperands(kernelOperands);
   auto kernelModule = kernelFunc.getParentOfType<GPUModuleOp>();
   auto kernelSymbol = builder.getSymbolRefAttr(
@@ -441,17 +440,6 @@ void LaunchFuncOp::build(OpBuilder &builder, OperationState &result,
   result.addAttribute(getKernelAttrName(), kernelSymbol);
 }
 
-void LaunchFuncOp::build(OpBuilder &builder, OperationState &result,
-                         GPUFuncOp kernelFunc, KernelDim3 gridSize,
-                         KernelDim3 blockSize, ValueRange kernelOperands) {
-  build(builder, result, kernelFunc, gridSize.x, gridSize.y, gridSize.z,
-        blockSize.x, blockSize.y, blockSize.z, kernelOperands);
-}
-
-SymbolRefAttr LaunchFuncOp::kernel() {
-  return getAttrOfType<SymbolRefAttr>(getKernelAttrName());
-}
-
 unsigned LaunchFuncOp::getNumKernelOperands() {
   return getNumOperands() - kNumConfigOperands;
 }
@@ -492,6 +480,33 @@ static LogicalResult verify(LaunchFuncOp op) {
   return success();
 }
 
+static ParseResult
+parseLaunchFuncOperands(OpAsmParser &parser,
+                        SmallVectorImpl<OpAsmParser::OperandType> &argNames,
+                        SmallVectorImpl<Type> &argTypes) {
+  if (parser.parseOptionalKeyword("args"))
+    return success();
+  SmallVector<NamedAttrList, 4> argAttrs;
+  bool isVariadic = false;
+  return impl::parseFunctionArgumentList(parser, /*allowAttributes=*/false,
+                                         /*allowVariadic=*/false, argNames,
+                                         argTypes, argAttrs, isVariadic);
+}
+
+static void printLaunchFuncOperands(OpAsmPrinter &printer,
+                                    OperandRange operands, TypeRange types) {
+  if (operands.empty())
+    return;
+  printer << "args(";
+  llvm::interleaveComma(llvm::zip(operands, types), printer,
+                        [&](const auto &pair) {
+                          printer.printOperand(std::get<0>(pair));
+                          printer << " : ";
+                          printer.printType(std::get<1>(pair));
+                        });
+  printer << ")";
+}
+
 //===----------------------------------------------------------------------===//
 // GPUFuncOp
 //===----------------------------------------------------------------------===//
diff --git a/mlir/lib/Dialect/SCF/Transforms/Bufferize.cpp b/mlir/lib/Dialect/SCF/Transforms/Bufferize.cpp
new file mode 100644
index 0000000000000..23cf72f6ed2a0
--- /dev/null
+++ b/mlir/lib/Dialect/SCF/Transforms/Bufferize.cpp
@@ -0,0 +1,41 @@
+//===- Bufferize.cpp - scf bufferize pass ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/Transforms/Bufferize.h"
+#include "PassDetail.h"
+#include "mlir/Dialect/SCF/Passes.h"
+#include "mlir/Dialect/SCF/SCF.h"
+#include "mlir/Dialect/SCF/Transforms.h"
+#include "mlir/Dialect/StandardOps/IR/Ops.h"
+#include "mlir/Transforms/DialectConversion.h"
+
+using namespace mlir;
+using namespace mlir::scf;
+
+namespace {
+struct SCFBufferizePass : public SCFBufferizeBase<SCFBufferizePass> {
+  void runOnFunction() override {
+    auto func = getOperation();
+    auto *context = &getContext();
+
+    BufferizeTypeConverter typeConverter;
+    OwningRewritePatternList patterns;
+    ConversionTarget target(*context);
+
+    populateBufferizeMaterializationLegality(target);
+    populateSCFStructuralTypeConversionsAndLegality(context, typeConverter,
+                                                    patterns, target);
+    if (failed(applyPartialConversion(func, target, patterns)))
+      return signalPassFailure();
+  };
+};
+} // end anonymous namespace
+
+std::unique_ptr<Pass> mlir::createSCFBufferizePass() {
+  return std::make_unique<SCFBufferizePass>();
+}
diff --git a/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt b/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt
index b3b20027896e1..bd36cf7586507 100644
--- a/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/SCF/Transforms/CMakeLists.txt
@@ -1,7 +1,9 @@
 add_mlir_dialect_library(MLIRSCFTransforms
+  Bufferize.cpp
   LoopSpecialization.cpp
   ParallelLoopFusion.cpp
   ParallelLoopTiling.cpp
+  StructuralTypeConversions.cpp
   Utils.cpp
 
   ADDITIONAL_HEADER_DIRS
@@ -17,5 +19,6 @@ add_mlir_dialect_library(MLIRSCFTransforms
   MLIRSCF
   MLIRStandard
   MLIRSupport
+  MLIRTransforms
   MLIRTransformUtils
 )
diff --git a/mlir/lib/Dialect/SCF/Transforms/StructuralTypeConversions.cpp b/mlir/lib/Dialect/SCF/Transforms/StructuralTypeConversions.cpp
new file mode 100644
index 0000000000000..30a2272f39a24
--- /dev/null
+++ b/mlir/lib/Dialect/SCF/Transforms/StructuralTypeConversions.cpp
@@ -0,0 +1,117 @@
+//===- StructuralTypeConversions.cpp - scf structural type conversions ----===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "PassDetail.h"
+#include "mlir/Dialect/SCF/Passes.h"
+#include "mlir/Dialect/SCF/SCF.h"
+#include "mlir/Dialect/SCF/Transforms.h"
+#include "mlir/Dialect/StandardOps/IR/Ops.h"
+#include "mlir/Transforms/DialectConversion.h"
+
+using namespace mlir;
+using namespace mlir::scf;
+
+namespace {
+class ConvertForOpTypes : public OpConversionPattern<ForOp> {
+public:
+  using OpConversionPattern::OpConversionPattern;
+  LogicalResult
+  matchAndRewrite(ForOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    SmallVector<Type, 6> newResultTypes;
+    for (auto type : op.getResultTypes()) {
+      Type newType = typeConverter->convertType(type);
+      if (!newType)
+        return rewriter.notifyMatchFailure(op, "not a 1:1 type conversion");
+      newResultTypes.push_back(newType);
+    }
+
+    // Clone and replace.
+    ForOp newOp = cast<ForOp>(rewriter.clone(*op.getOperation()));
+    newOp.getOperation()->setOperands(operands);
+    for (auto t : llvm::zip(newOp.getResults(), newResultTypes))
+      std::get<0>(t).setType(std::get<1>(t));
+    auto bodyArgs = newOp.getBody()->getArguments();
+    for (auto t : llvm::zip(llvm::drop_begin(bodyArgs, 1), newResultTypes))
+      std::get<0>(t).setType(std::get<1>(t));
+    rewriter.replaceOp(op, newOp.getResults());
+
+    return success();
+  }
+};
+} // namespace
+
+namespace {
+class ConvertIfOpTypes : public OpConversionPattern<IfOp> {
+public:
+  using OpConversionPattern::OpConversionPattern;
+  LogicalResult
+  matchAndRewrite(IfOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    // TODO: Generalize this to any type conversion, not just 1:1.
+    //
+    // We need to implement something more sophisticated here that tracks which
+    // types convert to which other types and does the appropriate
+    // materialization logic.
+    // For example, it's possible that one result type converts to 0 types and
+    // another to 2 types, so newResultTypes would at least be the right size to
+    // not crash in the llvm::zip call below, but then we would set the the
+    // wrong type on the SSA values! These edge cases are also why we cannot
+    // safely use the TypeConverter::convertTypes helper here.
+    SmallVector<Type, 6> newResultTypes;
+    for (auto type : op.getResultTypes()) {
+      Type newType = typeConverter->convertType(type);
+      if (!newType)
+        return rewriter.notifyMatchFailure(op, "not a 1:1 type conversion");
+      newResultTypes.push_back(newType);
+    }
+
+    // TODO: Write this with updateRootInPlace once the conversion infra
+    // supports source materializations on ops updated in place.
+    IfOp newOp = cast<IfOp>(rewriter.clone(*op.getOperation()));
+    newOp.getOperation()->setOperands(operands);
+    for (auto t : llvm::zip(newOp.getResults(), newResultTypes))
+      std::get<0>(t).setType(std::get<1>(t));
+    rewriter.replaceOp(op, newOp.getResults());
+    return success();
+  }
+};
+} // namespace
+
+namespace {
+// When the result types of a ForOp/IfOp get changed, the operand types of the
+// corresponding yield op need to be changed. In order to trigger the
+// appropriate type conversions / materializations, we need a dummy pattern.
+class ConvertYieldOpTypes : public OpConversionPattern<scf::YieldOp> {
+public:
+  using OpConversionPattern::OpConversionPattern;
+  LogicalResult
+  matchAndRewrite(scf::YieldOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override {
+    rewriter.replaceOpWithNewOp<scf::YieldOp>(op, operands);
+    return success();
+  }
+};
+} // namespace
+
+void mlir::scf::populateSCFStructuralTypeConversionsAndLegality(
+    MLIRContext *context, TypeConverter &typeConverter,
+    OwningRewritePatternList &patterns, ConversionTarget &target) {
+  patterns.insert<ConvertForOpTypes, ConvertIfOpTypes, ConvertYieldOpTypes>(
+      typeConverter, context);
+  target.addDynamicallyLegalOp<ForOp, IfOp>([&](Operation *op) {
+    return typeConverter.isLegal(op->getResultTypes());
+  });
+  target.addDynamicallyLegalOp<scf::YieldOp>([&](scf::YieldOp op) {
+    // We only have conversions for a subset of ops that use scf.yield
+    // terminators.
+    if (!isa<ForOp, IfOp>(op.getParentOp()))
+      return true;
+    return typeConverter.isLegal(op.getOperandTypes());
+  });
+}
diff --git a/mlir/lib/Dialect/Shape/IR/ShapeCanonicalization.td b/mlir/lib/Dialect/Shape/IR/ShapeCanonicalization.td
index b03e512999073..c57ad8c8d17c8 100644
--- a/mlir/lib/Dialect/Shape/IR/ShapeCanonicalization.td
+++ b/mlir/lib/Dialect/Shape/IR/ShapeCanonicalization.td
@@ -1,7 +1,5 @@
 include "mlir/Dialect/Shape/IR/ShapeOps.td"
 
-def EqualBinaryOperands : Constraint<CPred<"$0 == $1">>;
-
 def AllInputShapesEq : Constraint<CPred< [{
   llvm::all_of($0, [&](mlir::Value val) {
     return $0[0] == val;
@@ -10,9 +8,8 @@ def AllInputShapesEq : Constraint<CPred< [{
 
 // Canonicalization patterns.
 
-def CstrBroadcastableEqOps : Pat<(Shape_CstrBroadcastableOp:$op $lhs, $rhs),
-  (Shape_ConstWitnessOp ConstBoolAttrTrue),
-  [(EqualBinaryOperands $lhs, $rhs)]>;
+def CstrBroadcastableEqOps : Pat<(Shape_CstrBroadcastableOp:$op $x, $x),
+  (Shape_ConstWitnessOp ConstBoolAttrTrue)>;
 
 def CstrEqEqOps : Pat<(Shape_CstrEqOp:$op $shapes),
   (Shape_ConstWitnessOp ConstBoolAttrTrue),
diff --git a/mlir/lib/Dialect/Shape/Transforms/Bufferize.cpp b/mlir/lib/Dialect/Shape/Transforms/Bufferize.cpp
index bdebfa9a32d7b..20cd960e040fe 100644
--- a/mlir/lib/Dialect/Shape/Transforms/Bufferize.cpp
+++ b/mlir/lib/Dialect/Shape/Transforms/Bufferize.cpp
@@ -8,82 +8,30 @@
 
 #include "mlir/Transforms/Bufferize.h"
 #include "PassDetail.h"
-#include "mlir/Dialect/Shape/IR/Shape.h"
 #include "mlir/Dialect/Shape/Transforms/Passes.h"
-#include "mlir/Dialect/StandardOps/IR/Ops.h"
-#include "mlir/IR/Operation.h"
-#include "mlir/IR/StandardTypes.h"
 #include "mlir/Pass/Pass.h"
 
 using namespace mlir;
-using namespace mlir::shape;
 
 namespace {
-// Propagate tensor to memref conversions through shape.assuming ops.
-class TypeConversionAssumingOpConverter
-    : public BufferizeOpConversionPattern<shape::AssumingOp> {
-public:
-  using BufferizeOpConversionPattern<
-      shape::AssumingOp>::BufferizeOpConversionPattern;
-
-  LogicalResult
-  matchAndRewrite(shape::AssumingOp assumingOp, ArrayRef<Value> operands,
-                  ConversionPatternRewriter &rewriter) const final {
-    SmallVector<Type, 2> newResultTypes;
-    newResultTypes.reserve(assumingOp.getNumResults());
-    for (auto result : assumingOp.getResults()) {
-      auto originalType = result.getType();
-      Type convertedType = converter.convertType(originalType);
-      newResultTypes.push_back(convertedType);
-    }
-
-    auto newAssumingOp = rewriter.create<shape::AssumingOp>(
-        assumingOp.getLoc(), newResultTypes, assumingOp.witness());
-
-    rewriter.replaceOp(assumingOp, newAssumingOp.getResults());
-    rewriter.inlineRegionBefore(assumingOp.doRegion(), newAssumingOp.doRegion(),
-                                newAssumingOp.doRegion().end());
-
-    return success();
-  }
-};
-
 struct ShapeBufferizePass : public ShapeBufferizeBase<ShapeBufferizePass> {
   void runOnFunction() override {
     MLIRContext &ctx = getContext();
 
     OwningRewritePatternList patterns;
-    BufferizeTypeConverter converter;
-    populateShapeTypeConversionPatterns(&ctx, converter, patterns);
-
+    BufferizeTypeConverter typeConverter;
     ConversionTarget target(getContext());
-    auto isMemRefType = [](Type type) { return type.isa<BaseMemRefType>(); };
 
-    target.addDynamicallyLegalOp<AssumingOp>([&](shape::AssumingOp op) {
-      return std::all_of(op.result_type_begin(), op.result_type_end(),
-                         isMemRefType);
-    });
+    populateBufferizeMaterializationLegality(target);
+    populateShapeStructuralTypeConversionsAndLegality(&ctx, typeConverter,
+                                                      patterns, target);
 
-    if (failed(mlir::applyPartialConversion(getFunction(), target, patterns)))
+    if (failed(applyPartialConversion(getFunction(), target, patterns)))
       signalPassFailure();
   }
 };
-
 } // namespace
 
-/// Populates `patterns` with the conversion patterns of tensor->memref.
-//
-// TODO: Change this to work generally with any type conversions.
-void mlir::populateShapeTypeConversionPatterns(
-    MLIRContext *context, BufferizeTypeConverter &converter,
-    OwningRewritePatternList &patterns) {
-  patterns.insert<TypeConversionAssumingOpConverter>(context, converter);
-}
-
-//===----------------------------------------------------------------------===//
-// ShapeBufferizePass construction
-//===----------------------------------------------------------------------===//
-
 std::unique_ptr<FunctionPass> mlir::createShapeBufferizePass() {
   return std::make_unique<ShapeBufferizePass>();
 }
diff --git a/mlir/lib/Dialect/Shape/Transforms/CMakeLists.txt b/mlir/lib/Dialect/Shape/Transforms/CMakeLists.txt
index ce413f57d989e..123a3664df894 100644
--- a/mlir/lib/Dialect/Shape/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/Shape/Transforms/CMakeLists.txt
@@ -2,6 +2,7 @@ add_mlir_dialect_library(MLIRShapeOpsTransforms
   Bufferize.cpp
   RemoveShapeConstraints.cpp
   ShapeToShapeLowering.cpp
+  StructuralTypeConversions.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/ShapeOps/Transforms
diff --git a/mlir/lib/Dialect/Shape/Transforms/StructuralTypeConversions.cpp b/mlir/lib/Dialect/Shape/Transforms/StructuralTypeConversions.cpp
new file mode 100644
index 0000000000000..61e862836a733
--- /dev/null
+++ b/mlir/lib/Dialect/Shape/Transforms/StructuralTypeConversions.cpp
@@ -0,0 +1,71 @@
+//===- StructuralTypeConversions.cpp - Shape structural type conversions --===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "PassDetail.h"
+#include "mlir/Dialect/Shape/IR/Shape.h"
+#include "mlir/Dialect/Shape/Transforms/Passes.h"
+#include "mlir/Transforms/DialectConversion.h"
+
+using namespace mlir;
+using namespace mlir::shape;
+
+namespace {
+class ConvertAssumingOpTypes : public OpConversionPattern<AssumingOp> {
+public:
+  using OpConversionPattern::OpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(AssumingOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const final {
+    SmallVector<Type, 2> newResultTypes;
+    newResultTypes.reserve(op.getNumResults());
+    for (auto result : op.getResults()) {
+      auto originalType = result.getType();
+      Type convertedType = getTypeConverter()->convertType(originalType);
+      newResultTypes.push_back(convertedType);
+    }
+
+    auto newAssumingOp =
+        rewriter.create<AssumingOp>(op.getLoc(), newResultTypes, op.witness());
+
+    rewriter.replaceOp(op, newAssumingOp.getResults());
+    rewriter.inlineRegionBefore(op.doRegion(), newAssumingOp.doRegion(),
+                                newAssumingOp.doRegion().end());
+
+    return success();
+  }
+};
+} // namespace
+
+namespace {
+class ConvertAssumingYieldOpTypes
+    : public OpConversionPattern<AssumingYieldOp> {
+public:
+  using OpConversionPattern::OpConversionPattern;
+
+  LogicalResult
+  matchAndRewrite(AssumingYieldOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const final {
+    rewriter.replaceOpWithNewOp<AssumingYieldOp>(op, operands);
+    return success();
+  }
+};
+} // namespace
+
+void mlir::populateShapeStructuralTypeConversionsAndLegality(
+    MLIRContext *context, TypeConverter &typeConverter,
+    OwningRewritePatternList &patterns, ConversionTarget &target) {
+  patterns.insert<ConvertAssumingOpTypes, ConvertAssumingYieldOpTypes>(
+      typeConverter, context);
+  target.addDynamicallyLegalOp<AssumingOp>([&](AssumingOp op) {
+    return typeConverter.isLegal(op.getResultTypes());
+  });
+  target.addDynamicallyLegalOp<AssumingYieldOp>([&](AssumingYieldOp op) {
+    return typeConverter.isLegal(op.getOperandTypes());
+  });
+}
diff --git a/mlir/lib/Dialect/StandardOps/IR/Ops.cpp b/mlir/lib/Dialect/StandardOps/IR/Ops.cpp
index 9fe94fe753279..4638ae4e92680 100644
--- a/mlir/lib/Dialect/StandardOps/IR/Ops.cpp
+++ b/mlir/lib/Dialect/StandardOps/IR/Ops.cpp
@@ -261,6 +261,126 @@ OpFoldResult AddIOp::fold(ArrayRef<Attribute> operands) {
                                         [](APInt a, APInt b) { return a + b; });
 }
 
+//===----------------------------------------------------------------------===//
+// BaseOpWithOffsetSizesAndStridesOp
+//===----------------------------------------------------------------------===//
+
+/// Print a list with either (1) the static integer value in `arrayAttr` if
+/// `isDynamic` evaluates to false or (2) the next value otherwise.
+/// This allows idiomatic printing of mixed value and integer attributes in a
+/// list. E.g. `[%arg0, 7, 42, %arg42]`.
+static void
+printListOfOperandsOrIntegers(OpAsmPrinter &p, ValueRange values,
+                              ArrayAttr arrayAttr,
+                              llvm::function_ref<bool(int64_t)> isDynamic) {
+  p << '[';
+  unsigned idx = 0;
+  llvm::interleaveComma(arrayAttr, p, [&](Attribute a) {
+    int64_t val = a.cast<IntegerAttr>().getInt();
+    if (isDynamic(val))
+      p << values[idx++];
+    else
+      p << val;
+  });
+  p << ']';
+}
+
+/// Parse a mixed list with either (1) static integer values or (2) SSA values.
+/// Fill `result` with the integer ArrayAttr named `attrName` where `dynVal`
+/// encode the position of SSA values. Add the parsed SSA values to `ssa`
+/// in-order.
+//
+/// E.g. after parsing "[%arg0, 7, 42, %arg42]":
+///   1. `result` is filled with the i64 ArrayAttr "[`dynVal`, 7, 42, `dynVal`]"
+///   2. `ssa` is filled with "[%arg0, %arg1]".
+static ParseResult
+parseListOfOperandsOrIntegers(OpAsmParser &parser, OperationState &result,
+                              StringRef attrName, int64_t dynVal,
+                              SmallVectorImpl<OpAsmParser::OperandType> &ssa) {
+  if (failed(parser.parseLSquare()))
+    return failure();
+  // 0-D.
+  if (succeeded(parser.parseOptionalRSquare())) {
+    result.addAttribute(attrName, parser.getBuilder().getArrayAttr({}));
+    return success();
+  }
+
+  SmallVector<int64_t, 4> attrVals;
+  while (true) {
+    OpAsmParser::OperandType operand;
+    auto res = parser.parseOptionalOperand(operand);
+    if (res.hasValue() && succeeded(res.getValue())) {
+      ssa.push_back(operand);
+      attrVals.push_back(dynVal);
+    } else {
+      IntegerAttr attr;
+      if (failed(parser.parseAttribute<IntegerAttr>(attr)))
+        return parser.emitError(parser.getNameLoc())
+               << "expected SSA value or integer";
+      attrVals.push_back(attr.getInt());
+    }
+
+    if (succeeded(parser.parseOptionalComma()))
+      continue;
+    if (failed(parser.parseRSquare()))
+      return failure();
+    break;
+  }
+
+  auto arrayAttr = parser.getBuilder().getI64ArrayAttr(attrVals);
+  result.addAttribute(attrName, arrayAttr);
+  return success();
+}
+
+/// Verify that a particular offset/size/stride static attribute is well-formed.
+template <typename OpType>
+static LogicalResult verifyOpWithOffsetSizesAndStridesPart(
+    OpType op, StringRef name, unsigned expectedNumElements, StringRef attrName,
+    ArrayAttr attr, llvm::function_ref<bool(int64_t)> isDynamic,
+    ValueRange values) {
+  /// Check static and dynamic offsets/sizes/strides breakdown.
+  if (attr.size() != expectedNumElements)
+    return op.emitError("expected ")
+           << expectedNumElements << " " << name << " values";
+  unsigned expectedNumDynamicEntries =
+      llvm::count_if(attr.getValue(), [&](Attribute attr) {
+        return isDynamic(attr.cast<IntegerAttr>().getInt());
+      });
+  if (values.size() != expectedNumDynamicEntries)
+    return op.emitError("expected ")
+           << expectedNumDynamicEntries << " dynamic " << name << " values";
+  return success();
+}
+
+/// Extract int64_t values from the assumed ArrayAttr of IntegerAttr.
+static SmallVector<int64_t, 4> extractFromI64ArrayAttr(Attribute attr) {
+  return llvm::to_vector<4>(
+      llvm::map_range(attr.cast<ArrayAttr>(), [](Attribute a) -> int64_t {
+        return a.cast<IntegerAttr>().getInt();
+      }));
+}
+
+/// Verify static attributes offsets/sizes/strides.
+template <typename OpType>
+static LogicalResult verifyOpWithOffsetSizesAndStrides(OpType op) {
+  unsigned srcRank = op.getSourceRank();
+  if (failed(verifyOpWithOffsetSizesAndStridesPart(
+          op, "offset", srcRank, op.getStaticOffsetsAttrName(),
+          op.static_offsets(), ShapedType::isDynamicStrideOrOffset,
+          op.offsets())))
+    return failure();
+  if (failed(verifyOpWithOffsetSizesAndStridesPart(
+          op, "size", srcRank, op.getStaticSizesAttrName(), op.static_sizes(),
+          ShapedType::isDynamic, op.sizes())))
+    return failure();
+  if (failed(verifyOpWithOffsetSizesAndStridesPart(
+          op, "stride", srcRank, op.getStaticStridesAttrName(),
+          op.static_strides(), ShapedType::isDynamicStrideOrOffset,
+          op.strides())))
+    return failure();
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // AllocOp / AllocaOp
 //===----------------------------------------------------------------------===//
@@ -2145,6 +2265,204 @@ OpFoldResult MemRefCastOp::fold(ArrayRef<Attribute> operands) {
   return impl::foldCastOp(*this);
 }
 
+//===----------------------------------------------------------------------===//
+// MemRefReinterpretCastOp
+//===----------------------------------------------------------------------===//
+
+/// Print of the form:
+/// ```
+///   `name` ssa-name to
+///       offset: `[` offset `]`
+///       sizes: `[` size-list `]`
+///       strides:`[` stride-list `]`
+///   `:` any-memref-type to strided-memref-type
+/// ```
+static void print(OpAsmPrinter &p, MemRefReinterpretCastOp op) {
+  int stdDotLen = StandardOpsDialect::getDialectNamespace().size() + 1;
+  p << op.getOperationName().drop_front(stdDotLen) << " " << op.source()
+    << " to offset: ";
+  printListOfOperandsOrIntegers(p, op.offsets(), op.static_offsets(),
+                                ShapedType::isDynamicStrideOrOffset);
+  p << ", sizes: ";
+  printListOfOperandsOrIntegers(p, op.sizes(), op.static_sizes(),
+                                ShapedType::isDynamic);
+  p << ", strides: ";
+  printListOfOperandsOrIntegers(p, op.strides(), op.static_strides(),
+                                ShapedType::isDynamicStrideOrOffset);
+  p.printOptionalAttrDict(
+      op.getAttrs(),
+      /*elidedAttrs=*/{MemRefReinterpretCastOp::getOperandSegmentSizeAttr(),
+                       MemRefReinterpretCastOp::getStaticOffsetsAttrName(),
+                       MemRefReinterpretCastOp::getStaticSizesAttrName(),
+                       MemRefReinterpretCastOp::getStaticStridesAttrName()});
+  p << ": " << op.source().getType() << " to " << op.getType();
+}
+
+/// Parse of the form:
+/// ```
+///   `name` ssa-name to
+///       offset: `[` offset `]`
+///       sizes: `[` size-list `]`
+///       strides:`[` stride-list `]`
+///   `:` any-memref-type to strided-memref-type
+/// ```
+static ParseResult parseMemRefReinterpretCastOp(OpAsmParser &parser,
+                                                OperationState &result) {
+  // Parse `operand` and `offset`.
+  OpAsmParser::OperandType operand;
+  if (parser.parseOperand(operand))
+    return failure();
+
+  // Parse offset.
+  SmallVector<OpAsmParser::OperandType, 1> offset;
+  if (parser.parseKeyword("to") || parser.parseKeyword("offset") ||
+      parser.parseColon() ||
+      parseListOfOperandsOrIntegers(
+          parser, result, MemRefReinterpretCastOp::getStaticOffsetsAttrName(),
+          ShapedType::kDynamicStrideOrOffset, offset) ||
+      parser.parseComma())
+    return failure();
+
+  // Parse `sizes`.
+  SmallVector<OpAsmParser::OperandType, 4> sizes;
+  if (parser.parseKeyword("sizes") || parser.parseColon() ||
+      parseListOfOperandsOrIntegers(
+          parser, result, MemRefReinterpretCastOp::getStaticSizesAttrName(),
+          ShapedType::kDynamicSize, sizes) ||
+      parser.parseComma())
+    return failure();
+
+  // Parse `strides`.
+  SmallVector<OpAsmParser::OperandType, 4> strides;
+  if (parser.parseKeyword("strides") || parser.parseColon() ||
+      parseListOfOperandsOrIntegers(
+          parser, result, MemRefReinterpretCastOp::getStaticStridesAttrName(),
+          ShapedType::kDynamicStrideOrOffset, strides))
+    return failure();
+
+  // Handle segment sizes.
+  auto b = parser.getBuilder();
+  SmallVector<int, 4> segmentSizes = {1, static_cast<int>(offset.size()),
+                                      static_cast<int>(sizes.size()),
+                                      static_cast<int>(strides.size())};
+  result.addAttribute(MemRefReinterpretCastOp::getOperandSegmentSizeAttr(),
+
+                      b.getI32VectorAttr(segmentSizes));
+
+  // Parse types and resolve.
+  Type indexType = b.getIndexType();
+  Type operandType, resultType;
+  return failure(
+      (parser.parseOptionalAttrDict(result.attributes) ||
+       parser.parseColonType(operandType) || parser.parseKeyword("to") ||
+       parser.parseType(resultType) ||
+       parser.resolveOperand(operand, operandType, result.operands) ||
+       parser.resolveOperands(offset, indexType, result.operands) ||
+       parser.resolveOperands(sizes, indexType, result.operands) ||
+       parser.resolveOperands(strides, indexType, result.operands) ||
+       parser.addTypeToList(resultType, result.types)));
+}
+
+static LogicalResult verify(MemRefReinterpretCastOp op) {
+  // The source and result memrefs should be in the same memory space.
+  auto srcType = op.source().getType().cast<BaseMemRefType>();
+  auto resultType = op.getType().cast<MemRefType>();
+  if (srcType.getMemorySpace() != resultType.getMemorySpace())
+    return op.emitError("different memory spaces specified for source type ")
+           << srcType << " and result memref type " << resultType;
+  if (srcType.getElementType() != resultType.getElementType())
+    return op.emitError("different element types specified for source type ")
+           << srcType << " and result memref type " << resultType;
+
+  // Verify that dynamic and static offset/sizes/strides arguments/attributes
+  // are consistent.
+  if (failed(verifyOpWithOffsetSizesAndStridesPart(
+          op, "offset", 1, op.getStaticOffsetsAttrName(), op.static_offsets(),
+          ShapedType::isDynamicStrideOrOffset, op.offsets())))
+    return failure();
+  unsigned resultRank = op.getResultRank();
+  if (failed(verifyOpWithOffsetSizesAndStridesPart(
+          op, "size", resultRank, op.getStaticSizesAttrName(),
+          op.static_sizes(), ShapedType::isDynamic, op.sizes())))
+    return failure();
+  if (failed(verifyOpWithOffsetSizesAndStridesPart(
+          op, "stride", resultRank, op.getStaticStridesAttrName(),
+          op.static_strides(), ShapedType::isDynamicStrideOrOffset,
+          op.strides())))
+    return failure();
+
+  // Extract source offset and strides.
+  int64_t resultOffset;
+  SmallVector<int64_t, 4> resultStrides;
+  if (failed(getStridesAndOffset(resultType, resultStrides, resultOffset)))
+    return failure();
+
+  // Match offset in result memref type and in static_offsets attribute.
+  int64_t expectedOffset = extractFromI64ArrayAttr(op.static_offsets()).front();
+  if (resultOffset != expectedOffset)
+    return op.emitError("expected result type with offset = ")
+           << resultOffset << " instead of " << expectedOffset;
+
+  // Match sizes in result memref type and in static_sizes attribute.
+  for (auto &en :
+       llvm::enumerate(llvm::zip(resultType.getShape(),
+                                 extractFromI64ArrayAttr(op.static_sizes())))) {
+    int64_t resultSize = std::get<0>(en.value());
+    int64_t expectedSize = std::get<1>(en.value());
+    if (resultSize != expectedSize)
+      return op.emitError("expected result type with size = ")
+             << expectedSize << " instead of " << resultSize
+             << " in dim = " << en.index();
+  }
+
+  // Match strides in result memref type and in static_strides attribute.
+  for (auto &en : llvm::enumerate(llvm::zip(
+           resultStrides, extractFromI64ArrayAttr(op.static_strides())))) {
+    int64_t resultStride = std::get<0>(en.value());
+    int64_t expectedStride = std::get<1>(en.value());
+    if (resultStride != expectedStride)
+      return op.emitError("expected result type with stride = ")
+             << expectedStride << " instead of " << resultStride
+             << " in dim = " << en.index();
+  }
+  return success();
+}
+
+//===----------------------------------------------------------------------===//
+// MemRefReshapeOp
+//===----------------------------------------------------------------------===//
+
+static LogicalResult verify(MemRefReshapeOp op) {
+  Type operandType = op.source().getType();
+  Type resultType = op.result().getType();
+
+  Type operandElementType = operandType.cast<ShapedType>().getElementType();
+  Type resultElementType = resultType.cast<ShapedType>().getElementType();
+  if (operandElementType != resultElementType)
+    return op.emitOpError("element types of source and destination memref "
+                          "types should be the same");
+
+  if (auto operandMemRefType = operandType.dyn_cast<MemRefType>())
+    if (!operandMemRefType.getAffineMaps().empty())
+      return op.emitOpError(
+          "source memref type should have identity affine map");
+
+  int64_t shapeSize = op.shape().getType().cast<MemRefType>().getDimSize(0);
+  auto resultMemRefType = resultType.dyn_cast<MemRefType>();
+  if (resultMemRefType) {
+    if (!resultMemRefType.getAffineMaps().empty())
+      return op.emitOpError(
+          "result memref type should have identity affine map");
+    if (shapeSize == ShapedType::kDynamicSize)
+      return op.emitOpError("cannot use shape operand with dynamic length to "
+                            "reshape to statically-ranked memref type");
+    if (shapeSize != resultMemRefType.getRank())
+      return op.emitOpError(
+          "length of shape operand differs from the result's memref rank");
+  }
+  return success();
+}
+
 //===----------------------------------------------------------------------===//
 // MulFOp
 //===----------------------------------------------------------------------===//
@@ -2542,75 +2860,6 @@ bool UIToFPOp::areCastCompatible(Type a, Type b) {
 // SubViewOp
 //===----------------------------------------------------------------------===//
 
-/// Print a list with either (1) the static integer value in `arrayAttr` if
-/// `isDynamic` evaluates to false or (2) the next value otherwise.
-/// This allows idiomatic printing of mixed value and integer attributes in a
-/// list. E.g. `[%arg0, 7, 42, %arg42]`.
-static void printSubViewListOfOperandsOrIntegers(
-    OpAsmPrinter &p, ValueRange values, ArrayAttr arrayAttr,
-    llvm::function_ref<bool(int64_t)> isDynamic) {
-  p << "[";
-  unsigned idx = 0;
-  llvm::interleaveComma(arrayAttr, p, [&](Attribute a) {
-    int64_t val = a.cast<IntegerAttr>().getInt();
-    if (isDynamic(val))
-      p << values[idx++];
-    else
-      p << val;
-  });
-  p << "] ";
-}
-
-/// Parse a mixed list with either (1) static integer values or (2) SSA values.
-/// Fill `result` with the integer ArrayAttr named `attrName` where `dynVal`
-/// encode the position of SSA values. Add the parsed SSA values to `ssa`
-/// in-order.
-//
-/// E.g. after parsing "[%arg0, 7, 42, %arg42]":
-///   1. `result` is filled with the i64 ArrayAttr "[`dynVal`, 7, 42, `dynVal`]"
-///   2. `ssa` is filled with "[%arg0, %arg1]".
-static ParseResult
-parseListOfOperandsOrIntegers(OpAsmParser &parser, OperationState &result,
-                              StringRef attrName, int64_t dynVal,
-                              SmallVectorImpl<OpAsmParser::OperandType> &ssa) {
-  if (failed(parser.parseLSquare()))
-    return failure();
-  // 0-D.
-  if (succeeded(parser.parseOptionalRSquare())) {
-    result.addAttribute(attrName, parser.getBuilder().getArrayAttr({}));
-    return success();
-  }
-
-  SmallVector<int64_t, 4> attrVals;
-  while (true) {
-    OpAsmParser::OperandType operand;
-    auto res = parser.parseOptionalOperand(operand);
-    if (res.hasValue() && succeeded(res.getValue())) {
-      ssa.push_back(operand);
-      attrVals.push_back(dynVal);
-    } else {
-      Attribute attr;
-      NamedAttrList placeholder;
-      if (failed(parser.parseAttribute(attr, "_", placeholder)) ||
-          !attr.isa<IntegerAttr>())
-        return parser.emitError(parser.getNameLoc())
-               << "expected SSA value or integer";
-      attrVals.push_back(attr.cast<IntegerAttr>().getInt());
-    }
-
-    if (succeeded(parser.parseOptionalComma()))
-      continue;
-    if (failed(parser.parseRSquare()))
-      return failure();
-    else
-      break;
-  }
-
-  auto arrayAttr = parser.getBuilder().getI64ArrayAttr(attrVals);
-  result.addAttribute(attrName, arrayAttr);
-  return success();
-}
-
 namespace {
 /// Helpers to write more idiomatic operations.
 namespace saturated_arith {
@@ -2698,12 +2947,15 @@ static void printOpWithOffsetsSizesAndStrides(
   p << op.getOperation()->getName().getStringRef().drop_front(stdDotLen) << ' ';
   p << op.source();
   printExtraOperands(p, op);
-  printSubViewListOfOperandsOrIntegers(p, op.offsets(), op.static_offsets(),
-                                       ShapedType::isDynamicStrideOrOffset);
-  printSubViewListOfOperandsOrIntegers(p, op.sizes(), op.static_sizes(),
-                                       ShapedType::isDynamic);
-  printSubViewListOfOperandsOrIntegers(p, op.strides(), op.static_strides(),
-                                       ShapedType::isDynamicStrideOrOffset);
+  printListOfOperandsOrIntegers(p, op.offsets(), op.static_offsets(),
+                                ShapedType::isDynamicStrideOrOffset);
+  p << ' ';
+  printListOfOperandsOrIntegers(p, op.sizes(), op.static_sizes(),
+                                ShapedType::isDynamic);
+  p << ' ';
+  printListOfOperandsOrIntegers(p, op.strides(), op.static_strides(),
+                                ShapedType::isDynamicStrideOrOffset);
+  p << ' ';
   p.printOptionalAttrDict(op.getAttrs(),
                           /*elidedAttrs=*/{OpType::getSpecialAttrNames()});
   p << " : " << op.getSourceType() << " " << resultTypeKeyword << " "
@@ -2843,33 +3095,6 @@ void mlir::SubViewOp::build(OpBuilder &b, OperationState &result,
 /// For ViewLikeOpInterface.
 Value SubViewOp::getViewSource() { return source(); }
 
-/// Verify that a particular offset/size/stride static attribute is well-formed.
-template <typename OpType>
-static LogicalResult verifyOpWithOffsetSizesAndStridesPart(
-    OpType op, StringRef name, StringRef attrName, ArrayAttr attr,
-    llvm::function_ref<bool(int64_t)> isDynamic, ValueRange values) {
-  /// Check static and dynamic offsets/sizes/strides breakdown.
-  if (attr.size() != op.getSourceRank())
-    return op.emitError("expected ")
-           << op.getSourceRank() << " " << name << " values";
-  unsigned expectedNumDynamicEntries =
-      llvm::count_if(attr.getValue(), [&](Attribute attr) {
-        return isDynamic(attr.cast<IntegerAttr>().getInt());
-      });
-  if (values.size() != expectedNumDynamicEntries)
-    return op.emitError("expected ")
-           << expectedNumDynamicEntries << " dynamic " << name << " values";
-  return success();
-}
-
-/// Helper function extracts int64_t from the assumedArrayAttr of IntegerAttr.
-static SmallVector<int64_t, 4> extractFromI64ArrayAttr(Attribute attr) {
-  return llvm::to_vector<4>(
-      llvm::map_range(attr.cast<ArrayAttr>(), [](Attribute a) -> int64_t {
-        return a.cast<IntegerAttr>().getInt();
-      }));
-}
-
 llvm::Optional<SmallVector<bool, 4>>
 mlir::computeRankReductionMask(ArrayRef<int64_t> originalShape,
                                ArrayRef<int64_t> reducedShape) {
@@ -3006,24 +3231,6 @@ static LogicalResult produceSubViewErrorMsg(SubViewVerificationResult result,
   llvm_unreachable("unexpected subview verification result");
 }
 
-template <typename OpType>
-static LogicalResult verifyOpWithOffsetSizesAndStrides(OpType op) {
-  // Verify static attributes offsets/sizes/strides.
-  if (failed(verifyOpWithOffsetSizesAndStridesPart(
-          op, "offset", op.getStaticOffsetsAttrName(), op.static_offsets(),
-          ShapedType::isDynamicStrideOrOffset, op.offsets())))
-    return failure();
-
-  if (failed(verifyOpWithOffsetSizesAndStridesPart(
-          op, "size", op.getStaticSizesAttrName(), op.static_sizes(),
-          ShapedType::isDynamic, op.sizes())))
-    return failure();
-  if (failed(verifyOpWithOffsetSizesAndStridesPart(
-          op, "stride", op.getStaticStridesAttrName(), op.static_strides(),
-          ShapedType::isDynamicStrideOrOffset, op.strides())))
-    return failure();
-  return success();
-}
 
 /// Verifier for SubViewOp.
 static LogicalResult verify(SubViewOp op) {
diff --git a/mlir/lib/ExecutionEngine/AsyncRuntime.cpp b/mlir/lib/ExecutionEngine/AsyncRuntime.cpp
new file mode 100644
index 0000000000000..b450a8334ae49
--- /dev/null
+++ b/mlir/lib/ExecutionEngine/AsyncRuntime.cpp
@@ -0,0 +1,91 @@
+//===- AsyncRuntime.cpp - Async runtime reference implementation ----------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file implements basic Async runtime API for supporting Async dialect
+// to LLVM dialect lowering.
+//
+//===----------------------------------------------------------------------===//
+
+#include "mlir/ExecutionEngine/AsyncRuntime.h"
+
+#ifdef MLIR_ASYNCRUNTIME_DEFINE_FUNCTIONS
+
+#include <condition_variable>
+#include <functional>
+#include <iostream>
+#include <mutex>
+#include <thread>
+#include <vector>
+
+//===----------------------------------------------------------------------===//
+// Async runtime API.
+//===----------------------------------------------------------------------===//
+
+struct AsyncToken {
+  bool ready = false;
+  std::mutex mu;
+  std::condition_variable cv;
+  std::vector<std::function<void()>> awaiters;
+};
+
+// Create a new `async.token` in not-ready state.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT AsyncToken *mlirAsyncRuntimeCreateToken() {
+  AsyncToken *token = new AsyncToken;
+  return token;
+}
+
+// Switches `async.token` to ready state and runs all awaiters.
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeEmplaceToken(AsyncToken *token) {
+  std::unique_lock<std::mutex> lock(token->mu);
+  token->ready = true;
+  token->cv.notify_all();
+  for (auto &awaiter : token->awaiters)
+    awaiter();
+}
+
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeAwaitToken(AsyncToken *token) {
+  std::unique_lock<std::mutex> lock(token->mu);
+  if (!token->ready)
+    token->cv.wait(lock, [token] { return token->ready; });
+  delete token;
+}
+
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeExecute(CoroHandle handle, CoroResume resume) {
+  (*resume)(handle);
+}
+
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimeAwaitTokenAndExecute(AsyncToken *token, CoroHandle handle,
+                                     CoroResume resume) {
+  std::unique_lock<std::mutex> lock(token->mu);
+
+  auto execute = [token, handle, resume]() {
+    mlirAsyncRuntimeExecute(handle, resume);
+    delete token;
+  };
+
+  if (token->ready)
+    execute();
+  else
+    token->awaiters.push_back([execute]() { execute(); });
+}
+
+//===----------------------------------------------------------------------===//
+// Small async runtime support library for testing.
+//===----------------------------------------------------------------------===//
+
+extern "C" MLIR_ASYNCRUNTIME_EXPORT void
+mlirAsyncRuntimePrintCurrentThreadId() {
+  static thread_local std::thread::id thisId = std::this_thread::get_id();
+  std::cout << "Current thread id: " << thisId << "\n";
+}
+
+#endif // MLIR_ASYNCRUNTIME_DEFINE_FUNCTIONS
diff --git a/mlir/lib/ExecutionEngine/CMakeLists.txt b/mlir/lib/ExecutionEngine/CMakeLists.txt
index 373df9f1468dc..acbfd9d483068 100644
--- a/mlir/lib/ExecutionEngine/CMakeLists.txt
+++ b/mlir/lib/ExecutionEngine/CMakeLists.txt
@@ -2,6 +2,7 @@
 # is a big dependency which most don't need.
 
 set(LLVM_OPTIONAL_SOURCES
+  AsyncRuntime.cpp
   CRunnerUtils.cpp
   SparseUtils.cpp
   ExecutionEngine.cpp
@@ -96,3 +97,14 @@ add_mlir_library(mlir_runner_utils
   mlir_c_runner_utils_static
 )
 target_compile_definitions(mlir_runner_utils PRIVATE mlir_runner_utils_EXPORTS)
+
+add_mlir_library(mlir_async_runtime
+  SHARED
+  AsyncRuntime.cpp
+
+  EXCLUDE_FROM_LIBMLIR
+
+  LINK_LIBS PUBLIC
+  mlir_c_runner_utils_static
+)
+target_compile_definitions(mlir_async_runtime PRIVATE mlir_async_runtime_EXPORTS)
diff --git a/mlir/lib/ExecutionEngine/OptUtils.cpp b/mlir/lib/ExecutionEngine/OptUtils.cpp
index 773bd7cff14df..53ad7fd66a9af 100644
--- a/mlir/lib/ExecutionEngine/OptUtils.cpp
+++ b/mlir/lib/ExecutionEngine/OptUtils.cpp
@@ -25,6 +25,7 @@
 #include "llvm/Support/Error.h"
 #include "llvm/Support/StringSaver.h"
 #include "llvm/Target/TargetMachine.h"
+#include "llvm/Transforms/Coroutines.h"
 #include "llvm/Transforms/IPO.h"
 #include "llvm/Transforms/IPO/PassManagerBuilder.h"
 #include <climits>
@@ -56,6 +57,7 @@ void mlir::initializeLLVMPasses() {
   llvm::initializeAggressiveInstCombine(registry);
   llvm::initializeAnalysis(registry);
   llvm::initializeVectorization(registry);
+  llvm::initializeCoroutines(registry);
 }
 
 // Populate pass managers according to the optimization and size levels.
@@ -73,6 +75,9 @@ static void populatePassManagers(llvm::legacy::PassManager &modulePM,
   builder.SLPVectorize = optLevel > 1 && sizeLevel < 2;
   builder.DisableUnrollLoops = (optLevel == 0);
 
+  // Add all coroutine passes to the builder.
+  addCoroutinePassesToExtensionPoints(builder);
+
   if (targetMachine) {
     // Add pass to initialize TTI for this specific target. Otherwise, TTI will
     // be initialized to NoTTIImpl by default.
diff --git a/mlir/lib/IR/FunctionImplementation.cpp b/mlir/lib/IR/FunctionImplementation.cpp
index 13aee344bbdcd..56b2221fd44f5 100644
--- a/mlir/lib/IR/FunctionImplementation.cpp
+++ b/mlir/lib/IR/FunctionImplementation.cpp
@@ -13,11 +13,11 @@
 
 using namespace mlir;
 
-static ParseResult
-parseArgumentList(OpAsmParser &parser, bool allowVariadic,
-                  SmallVectorImpl<Type> &argTypes,
-                  SmallVectorImpl<OpAsmParser::OperandType> &argNames,
-                  SmallVectorImpl<NamedAttrList> &argAttrs, bool &isVariadic) {
+ParseResult mlir::impl::parseFunctionArgumentList(
+    OpAsmParser &parser, bool allowAttributes, bool allowVariadic,
+    SmallVectorImpl<OpAsmParser::OperandType> &argNames,
+    SmallVectorImpl<Type> &argTypes, SmallVectorImpl<NamedAttrList> &argAttrs,
+    bool &isVariadic) {
   if (parser.parseLParen())
     return failure();
 
@@ -56,6 +56,8 @@ parseArgumentList(OpAsmParser &parser, bool allowVariadic,
     NamedAttrList attrs;
     if (parser.parseOptionalAttrDict(attrs))
       return failure();
+    if (!allowAttributes && !attrs.empty())
+      return parser.emitError(loc, "expected arguments without attributes");
     argAttrs.push_back(attrs);
     return success();
   };
@@ -129,8 +131,9 @@ ParseResult mlir::impl::parseFunctionSignature(
     SmallVectorImpl<Type> &argTypes, SmallVectorImpl<NamedAttrList> &argAttrs,
     bool &isVariadic, SmallVectorImpl<Type> &resultTypes,
     SmallVectorImpl<NamedAttrList> &resultAttrs) {
-  if (parseArgumentList(parser, allowVariadic, argTypes, argNames, argAttrs,
-                        isVariadic))
+  bool allowArgAttrs = true;
+  if (parseFunctionArgumentList(parser, allowArgAttrs, allowVariadic, argNames,
+                                argTypes, argAttrs, isVariadic))
     return failure();
   if (succeeded(parser.parseOptionalArrow()))
     return parseFunctionResultList(parser, resultTypes, resultAttrs);
diff --git a/mlir/lib/Transforms/Bufferize.cpp b/mlir/lib/Transforms/Bufferize.cpp
index 682fd9ff6719f..26eabe2b89473 100644
--- a/mlir/lib/Transforms/Bufferize.cpp
+++ b/mlir/lib/Transforms/Bufferize.cpp
@@ -72,6 +72,10 @@ BufferizeTypeConverter::getResultConversionKind(Type origin, Type converted) {
   return KeepAsFunctionResult;
 }
 
+void mlir::populateBufferizeMaterializationLegality(ConversionTarget &target) {
+  target.addLegalOp<TensorLoadOp, TensorToMemrefOp>();
+};
+
 //===----------------------------------------------------------------------===//
 // BufferizeFuncOpConverter
 //===----------------------------------------------------------------------===//
diff --git a/mlir/test/Bindings/Python/ir_operation.py b/mlir/test/Bindings/Python/ir_operation.py
index 37b8305585283..84f303ca570b0 100644
--- a/mlir/test/Bindings/Python/ir_operation.py
+++ b/mlir/test/Bindings/Python/ir_operation.py
@@ -1,6 +1,7 @@
 # RUN: %PYTHON %s | FileCheck %s
 
 import gc
+import io
 import itertools
 import mlir
 
@@ -102,6 +103,35 @@ def walk_operations(indent, op):
 run(testTraverseOpRegionBlockIndices)
 
 
+# CHECK-LABEL: TEST: testBlockArgumentList
+def testBlockArgumentList():
+  ctx = mlir.ir.Context()
+  module = ctx.parse_module(r"""
+    func @f1(%arg0: i32, %arg1: f64, %arg2: index) {
+      return
+    }
+  """)
+  func = module.operation.regions[0].blocks[0].operations[0]
+  entry_block = func.regions[0].blocks[0]
+  assert len(entry_block.arguments) == 3
+  # CHECK: Argument 0, type i32
+  # CHECK: Argument 1, type f64
+  # CHECK: Argument 2, type index
+  for arg in entry_block.arguments:
+    print(f"Argument {arg.arg_number}, type {arg.type}")
+    new_type = mlir.ir.IntegerType.get_signless(ctx, 8 * (arg.arg_number + 1))
+    arg.set_type(new_type)
+
+  # CHECK: Argument 0, type i8
+  # CHECK: Argument 1, type i16
+  # CHECK: Argument 2, type i24
+  for arg in entry_block.arguments:
+    print(f"Argument {arg.arg_number}, type {arg.type}")
+
+
+run(testBlockArgumentList)
+
+
 # CHECK-LABEL: TEST: testDetachedOperation
 def testDetachedOperation():
   ctx = mlir.ir.Context()
@@ -196,3 +226,67 @@ def testOperationWithRegion():
   print(module)
 
 run(testOperationWithRegion)
+
+
+# CHECK-LABEL: TEST: testOperationResultList
+def testOperationResultList():
+  ctx = mlir.ir.Context()
+  module = ctx.parse_module(r"""
+    func @f1() {
+      %0:3 = call @f2() : () -> (i32, f64, index)
+      return
+    }
+    func @f2() -> (i32, f64, index)
+  """)
+  caller = module.operation.regions[0].blocks[0].operations[0]
+  call = caller.regions[0].blocks[0].operations[0]
+  assert len(call.results) == 3
+  # CHECK: Result 0, type i32
+  # CHECK: Result 1, type f64
+  # CHECK: Result 2, type index
+  for res in call.results:
+    print(f"Result {res.result_number}, type {res.type}")
+
+
+run(testOperationResultList)
+
+
+# CHECK-LABEL: TEST: testOperationPrint
+def testOperationPrint():
+  ctx = mlir.ir.Context()
+  module = ctx.parse_module(r"""
+    func @f1(%arg0: i32) -> i32 {
+      %0 = constant dense<[1, 2, 3, 4]> : tensor<4xi32>
+      return %arg0 : i32
+    }
+  """)
+
+  # Test print to stdout.
+  # CHECK: return %arg0 : i32
+  module.operation.print()
+
+  # Test print to text file.
+  f = io.StringIO()
+  # CHECK: <class 'str'>
+  # CHECK: return %arg0 : i32
+  module.operation.print(file=f)
+  str_value = f.getvalue()
+  print(str_value.__class__)
+  print(f.getvalue())
+
+  # Test print to binary file.
+  f = io.BytesIO()
+  # CHECK: <class 'bytes'>
+  # CHECK: return %arg0 : i32
+  module.operation.print(file=f, binary=True)
+  bytes_value = f.getvalue()
+  print(bytes_value.__class__)
+  print(bytes_value)
+
+  # Test get_asm with options.
+  # CHECK: value = opaque<"", "0xDEADBEEF"> : tensor<4xi32>
+  # CHECK: "std.return"(%arg0) : (i32) -> () -:4:7
+  module.operation.print(large_elements_limit=2, enable_debug_info=True,
+      pretty_debug_info=True, print_generic_op_form=True, use_local_scope=True)
+
+run(testOperationPrint)
diff --git a/mlir/test/CAPI/ir.c b/mlir/test/CAPI/ir.c
index 7c86f403b3391..fa9a6258a4720 100644
--- a/mlir/test/CAPI/ir.c
+++ b/mlir/test/CAPI/ir.c
@@ -10,9 +10,9 @@
 /* RUN: mlir-capi-ir-test 2>&1 | FileCheck %s
  */
 
+#include "mlir-c/IR.h"
 #include "mlir-c/AffineMap.h"
 #include "mlir-c/Diagnostics.h"
-#include "mlir-c/IR.h"
 #include "mlir-c/Registration.h"
 #include "mlir-c/StandardAttributes.h"
 #include "mlir-c/StandardDialect.h"
@@ -319,6 +319,25 @@ static void printFirstOfEach(MlirContext ctx, MlirOperation operation) {
   fprintf(stderr, "Removed attr is null: %d\n",
           mlirAttributeIsNull(
               mlirOperationGetAttributeByName(operation, "custom_attr")));
+
+  // Add a large attribute to verify printing flags.
+  int64_t eltsShape[] = {4};
+  int32_t eltsData[] = {1, 2, 3, 4};
+  mlirOperationSetAttributeByName(
+      operation, "elts",
+      mlirDenseElementsAttrInt32Get(
+          mlirRankedTensorTypeGet(1, eltsShape, mlirIntegerTypeGet(ctx, 32)), 4,
+          eltsData));
+  MlirOpPrintingFlags flags = mlirOpPrintingFlagsCreate();
+  mlirOpPrintingFlagsElideLargeElementsAttrs(flags, 2);
+  mlirOpPrintingFlagsPrintGenericOpForm(flags);
+  mlirOpPrintingFlagsEnableDebugInfo(flags, /*prettyForm=*/0);
+  mlirOpPrintingFlagsUseLocalScope(flags);
+  fprintf(stderr, "Op print with all flags: ");
+  mlirOperationPrintWithFlags(operation, flags, printToStderr, NULL);
+  fprintf(stderr, "\n");
+
+  mlirOpPrintingFlagsDestroy(flags);
 }
 
 /// Creates an operation with a region containing multiple blocks with
@@ -991,6 +1010,7 @@ int main() {
   // CHECK: Remove attr: 1
   // CHECK: Remove attr again: 0
   // CHECK: Removed attr is null: 1
+  // CHECK: Op print with all flags: %{{.*}} = "std.constant"() {elts = opaque<"", "0xDEADBEEF"> : tensor<4xi32>, value = 0 : index} : () -> index loc(unknown)
   // clang-format on
 
   mlirModuleDestroy(moduleOp);
diff --git a/mlir/test/CMakeLists.txt b/mlir/test/CMakeLists.txt
index 33211a180a9fb..cc4a9b1af10de 100644
--- a/mlir/test/CMakeLists.txt
+++ b/mlir/test/CMakeLists.txt
@@ -54,6 +54,7 @@ set(MLIR_TEST_DEPENDS
   mlir_test_cblas_interface
   mlir_runner_utils
   mlir_c_runner_utils
+  mlir_async_runtime
   )
 
 if(LLVM_BUILD_EXAMPLES)
diff --git a/mlir/test/Conversion/AsyncToLLVM/convert-to-llvm.mlir b/mlir/test/Conversion/AsyncToLLVM/convert-to-llvm.mlir
new file mode 100644
index 0000000000000..f838b8fb57dce
--- /dev/null
+++ b/mlir/test/Conversion/AsyncToLLVM/convert-to-llvm.mlir
@@ -0,0 +1,111 @@
+// RUN: mlir-opt %s -split-input-file -convert-async-to-llvm | FileCheck %s
+
+// CHECK-LABEL: execute_no_async_args
+func @execute_no_async_args(%arg0: f32, %arg1: memref<1xf32>) {
+  // CHECK: %[[TOKEN:.*]] = call @async_execute_fn(%arg0, %arg1)
+  %token = async.execute {
+    %c0 = constant 0 : index
+    store %arg0, %arg1[%c0] : memref<1xf32>
+    async.yield
+  }
+  // CHECK: call @mlirAsyncRuntimeAwaitToken(%[[TOKEN]])
+  // CHECK-NEXT: return
+  async.await %token : !async.token
+  return
+}
+
+// Function outlined from the async.execute operation.
+// CHECK: func @async_execute_fn(%arg0: f32, %arg1: memref<1xf32>)
+// CHECK-SAME: -> !llvm.ptr<i8> attributes {sym_visibility = "private"}
+
+// Create token for return op, and mark a function as a coroutine.
+// CHECK: %[[RET:.*]] = call @mlirAsyncRuntimeCreateToken()
+// CHECK: %[[HDL:.*]] = llvm.call @llvm.coro.begin
+
+// Pass a suspended coroutine to the async runtime.
+// CHECK: %[[RESUME:.*]] = llvm.mlir.addressof @__resume
+// CHECK: %[[STATE:.*]] = llvm.call @llvm.coro.save
+// CHECK: call @mlirAsyncRuntimeExecute(%[[HDL]], %[[RESUME]])
+// CHECK: %[[SUSPENDED:.*]] = llvm.call @llvm.coro.suspend(%[[STATE]]
+
+// Decide the next block based on the code returned from suspend.
+// CHECK: %[[ZERO:.*]] = llvm.mlir.constant(0 : i8)
+// CHECK: %[[NONE:.*]] = llvm.mlir.constant(-1 : i8)
+// CHECK: %[[IS_NONE:.*]] = llvm.icmp "eq" %[[SUSPENDED]], %[[NONE]]
+// CHECK: llvm.cond_br %[[IS_NONE]], ^[[SUSPEND:.*]], ^[[RESUME_OR_CLEANUP:.*]]
+
+// Decide if branch to resume or cleanup block.
+// CHECK: ^[[RESUME_OR_CLEANUP]]:
+// CHECK: %[[IS_ZERO:.*]] = llvm.icmp "eq" %[[SUSPENDED]], %[[ZERO]]
+// CHECK: llvm.cond_br %[[IS_ZERO]], ^[[RESUME:.*]], ^[[CLEANUP:.*]]
+
+// Resume coroutine after suspension.
+// CHECK: ^[[RESUME]]:
+// CHECK: store %arg0, %arg1[%c0] : memref<1xf32>
+// CHECK: call @mlirAsyncRuntimeEmplaceToken(%[[RET]])
+
+// Delete coroutine.
+// CHECK: ^[[CLEANUP]]:
+// CHECK: %[[MEM:.*]] = llvm.call @llvm.coro.free
+// CHECK: llvm.call @free(%[[MEM]])
+
+// Suspend coroutine, and also a return statement for ramp function.
+// CHECK: ^[[SUSPEND]]:
+// CHECK: llvm.call @llvm.coro.end
+// CHECK: return %[[RET]]
+
+// -----
+
+// CHECK-LABEL: nested_async_execute
+func @nested_async_execute(%arg0: f32, %arg1: f32, %arg2: memref<1xf32>) {
+  // CHECK: %[[TOKEN:.*]] = call @async_execute_fn_0(%arg0, %arg2, %arg1)
+  %token0 = async.execute {
+    %c0 = constant 0 : index
+
+    %token1 = async.execute {
+      %c1 = constant 1: index
+      store %arg0, %arg2[%c0] : memref<1xf32>
+      async.yield
+    }
+    async.await %token1 : !async.token
+
+    store %arg1, %arg2[%c0] : memref<1xf32>
+    async.yield
+  }
+  // CHECK: call @mlirAsyncRuntimeAwaitToken(%[[TOKEN]])
+  // CHECK-NEXT: return
+  async.await %token0 : !async.token
+  return
+}
+
+// Function outlined from the inner async.execute operation.
+// CHECK: func @async_execute_fn(%arg0: f32, %arg1: memref<1xf32>, %arg2: index)
+// CHECK-SAME: -> !llvm.ptr<i8> attributes {sym_visibility = "private"}
+// CHECK: %[[RET_0:.*]] = call @mlirAsyncRuntimeCreateToken()
+// CHECK: %[[HDL_0:.*]] = llvm.call @llvm.coro.begin
+// CHECK: call @mlirAsyncRuntimeExecute
+// CHECK: llvm.call @llvm.coro.suspend
+// CHECK: store %arg0, %arg1[%arg2] : memref<1xf32>
+// CHECK: call @mlirAsyncRuntimeEmplaceToken(%[[RET_0]])
+
+// Function outlined from the outer async.execute operation.
+// CHECK: func @async_execute_fn_0(%arg0: f32, %arg1: memref<1xf32>, %arg2: f32)
+// CHECK-SAME: -> !llvm.ptr<i8> attributes {sym_visibility = "private"}
+// CHECK: %[[RET_1:.*]] = call @mlirAsyncRuntimeCreateToken()
+// CHECK: %[[HDL_1:.*]] = llvm.call @llvm.coro.begin
+
+// Suspend coroutine in the beginning.
+// CHECK: call @mlirAsyncRuntimeExecute
+// CHECK: llvm.call @llvm.coro.suspend
+
+// Suspend coroutine second time waiting for the completion of inner execute op.
+// CHECK: %[[TOKEN_1:.*]] = call @async_execute_fn
+// CHECK: llvm.call @llvm.coro.save
+// CHECK: call @mlirAsyncRuntimeAwaitTokenAndExecute(%[[TOKEN_1]], %[[HDL_1]]
+// CHECK: llvm.call @llvm.coro.suspend
+
+// Emplace result token after second resumption.
+// CHECK: store %arg2, %arg1[%c0] : memref<1xf32>
+// CHECK: call @mlirAsyncRuntimeEmplaceToken(%[[RET_1]])
+
+
diff --git a/mlir/test/Conversion/GPUCommon/lower-launch-func-to-gpu-runtime-calls.mlir b/mlir/test/Conversion/GPUCommon/lower-launch-func-to-gpu-runtime-calls.mlir
index 8d0ba7cec1e78..908604e091f55 100644
--- a/mlir/test/Conversion/GPUCommon/lower-launch-func-to-gpu-runtime-calls.mlir
+++ b/mlir/test/Conversion/GPUCommon/lower-launch-func-to-gpu-runtime-calls.mlir
@@ -20,13 +20,14 @@ module attributes {gpu.container_module} {
   func @foo(%buffer: memref<?xf32>) {
     %c8 = constant 8 : index
     %c32 = constant 32 : i32
-    "gpu.launch_func"(%c8, %c8, %c8, %c8, %c8, %c8, %c32, %buffer) {
-      kernel = @kernel_module::@kernel
-    } : (index, index, index, index, index, index, i32, memref<?xf32>) -> ()
+    gpu.launch_func @kernel_module::@kernel
+        blocks in (%c8, %c8, %c8)
+        threads in (%c8, %c8, %c8)
+        args(%c32 : i32, %buffer : memref<?xf32>)
     return
   }
 
-  // CHECK: [[C8:%.*]] = llvm.mlir.constant(8 : index) : !llvm.i64   
+  // CHECK: [[C8:%.*]] = llvm.mlir.constant(8 : index) : !llvm.i64
   // CHECK: [[ADDRESSOF:%.*]] = llvm.mlir.addressof @[[GLOBAL]]
   // CHECK: [[C0:%.*]] = llvm.mlir.constant(0 : index)
   // CHECK: [[BINARY:%.*]] = llvm.getelementptr [[ADDRESSOF]]{{\[}}[[C0]], [[C0]]]
diff --git a/mlir/test/Conversion/GPUCommon/lower-wait-to-gpu-runtime-calls.mlir b/mlir/test/Conversion/GPUCommon/lower-wait-to-gpu-runtime-calls.mlir
new file mode 100644
index 0000000000000..b6eacfb969dde
--- /dev/null
+++ b/mlir/test/Conversion/GPUCommon/lower-wait-to-gpu-runtime-calls.mlir
@@ -0,0 +1,21 @@
+// RUN: mlir-opt -allow-unregistered-dialect %s --gpu-to-llvm | FileCheck %s
+
+module attributes {gpu.container_module} {
+
+  func @foo() {
+    // CHECK: %[[t0:.*]] = llvm.call @mgpuStreamCreate
+    // CHECK: %[[e0:.*]] = llvm.call @mgpuEventCreate
+    // CHECK: llvm.call @mgpuEventRecord(%[[e0]], %[[t0]])
+    %t0 = gpu.wait async
+    // CHECK: %[[t1:.*]] = llvm.call @mgpuStreamCreate
+    // CHECK: llvm.call @mgpuStreamWaitEvent(%[[t1]], %[[e0]])
+    // CHECK: llvm.call @mgpuEventDestroy(%[[e0]])
+    %t1 = gpu.wait async [%t0]
+    // CHECK: llvm.call @mgpuStreamSynchronize(%[[t0]])
+    // CHECK: llvm.call @mgpuStreamSynchronize(%[[t1]])
+    // CHECK: llvm.call @mgpuStreamDestroy(%[[t0]])
+    // CHECK: llvm.call @mgpuStreamDestroy(%[[t1]])
+    gpu.wait [%t0, %t1]
+    return
+  }
+}
diff --git a/mlir/test/Conversion/GPUToSPIRV/builtins.mlir b/mlir/test/Conversion/GPUToSPIRV/builtins.mlir
index b4f05fa271279..3281a60e34e37 100644
--- a/mlir/test/Conversion/GPUToSPIRV/builtins.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/builtins.mlir
@@ -3,7 +3,8 @@
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_workgroup_id_x} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_workgroup_id_x
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -26,7 +27,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_workgroup_id_y} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_workgroup_id_y
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -49,7 +51,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_workgroup_id_z} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_workgroup_id_z
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -72,7 +75,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_workgroup_size_x} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_workgroup_size_x
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -96,7 +100,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_workgroup_size_y} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_workgroup_size_y
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -117,7 +122,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_workgroup_size_z} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_workgroup_size_z
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -138,7 +144,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_local_id_x} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_local_id_x
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
@@ -161,7 +168,8 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @builtin() {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0) {kernel = @kernels::@builtin_num_workgroups_x} : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@builtin_num_workgroups_x
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
     return
   }
 
diff --git a/mlir/test/Conversion/GPUToSPIRV/if.mlir b/mlir/test/Conversion/GPUToSPIRV/if.mlir
index 2efde9af1537e..e650b9b2f427a 100644
--- a/mlir/test/Conversion/GPUToSPIRV/if.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/if.mlir
@@ -7,7 +7,9 @@ module attributes {
 } {
   func @main(%arg0 : memref<10xf32>, %arg1 : i1) {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0, %arg0, %arg1) { kernel = @kernels::@kernel_simple_selection} : (index, index, index, index, index, index, memref<10xf32>, i1) -> ()
+    gpu.launch_func @kernels::@kernel_simple_selection
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
+        args(%arg0 : memref<10xf32>, %arg1 : i1)
     return
   }
 
diff --git a/mlir/test/Conversion/GPUToSPIRV/load-store.mlir b/mlir/test/Conversion/GPUToSPIRV/load-store.mlir
index 4db58c0a0f946..3577329c27a8c 100644
--- a/mlir/test/Conversion/GPUToSPIRV/load-store.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/load-store.mlir
@@ -15,7 +15,10 @@ module attributes {
     %1 = subi %c4, %c0_0 : index
     %c1_1 = constant 1 : index
     %c1_2 = constant 1 : index
-    "gpu.launch_func"(%0, %c1_2, %c1_2, %1, %c1_2, %c1_2, %arg0, %arg1, %arg2, %c0, %c0_0, %c1, %c1_1) {kernel = @kernels::@load_store_kernel} : (index, index, index, index, index, index, memref<12x4xf32>, memref<12x4xf32>, memref<12x4xf32>, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@load_store_kernel
+        blocks in (%0, %c1_2, %c1_2) threads in (%1, %c1_2, %c1_2)
+        args(%arg0 : memref<12x4xf32>, %arg1 : memref<12x4xf32>, %arg2 : memref<12x4xf32>, 
+             %c0 : index, %c0_0 : index, %c1 : index, %c1_1 : index)
     return
   }
 
diff --git a/mlir/test/Conversion/GPUToSPIRV/loop.mlir b/mlir/test/Conversion/GPUToSPIRV/loop.mlir
index c181e1956f83a..2cfef2b9cf330 100644
--- a/mlir/test/Conversion/GPUToSPIRV/loop.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/loop.mlir
@@ -7,7 +7,9 @@ module attributes {
 } {
   func @loop(%arg0 : memref<10xf32>, %arg1 : memref<10xf32>) {
     %c0 = constant 1 : index
-    "gpu.launch_func"(%c0, %c0, %c0, %c0, %c0, %c0, %arg0, %arg1) { kernel = @kernels::@loop_kernel} : (index, index, index, index, index, index, memref<10xf32>, memref<10xf32>) -> ()
+    gpu.launch_func @kernels::@loop_kernel
+        blocks in (%c0, %c0, %c0) threads in (%c0, %c0, %c0)
+        args(%arg0 : memref<10xf32>, %arg1 : memref<10xf32>)
     return
   }
 
diff --git a/mlir/test/Conversion/GPUToSPIRV/module-structure-opencl.mlir b/mlir/test/Conversion/GPUToSPIRV/module-structure-opencl.mlir
index 6b188263ed8b4..3066e5bba34d2 100644
--- a/mlir/test/Conversion/GPUToSPIRV/module-structure-opencl.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/module-structure-opencl.mlir
@@ -22,8 +22,9 @@ module attributes {
     %0 = "op"() : () -> (f32)
     %1 = "op"() : () -> (memref<12xf32, 11>)
     %cst = constant 1 : index
-    "gpu.launch_func"(%cst, %cst, %cst, %cst, %cst, %cst, %0, %1) { kernel = @kernels::@basic_module_structure }
-        : (index, index, index, index, index, index, f32, memref<12xf32, 11>) -> ()
+    gpu.launch_func @kernels::@basic_module_structure
+        blocks in (%cst, %cst, %cst) threads in (%cst, %cst, %cst)
+        args(%0 : f32, %1 : memref<12xf32, 11>)
     return
   }
 }
diff --git a/mlir/test/Conversion/GPUToSPIRV/simple.mlir b/mlir/test/Conversion/GPUToSPIRV/simple.mlir
index 89e6f52e655db..a8c5c3543cc34 100644
--- a/mlir/test/Conversion/GPUToSPIRV/simple.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/simple.mlir
@@ -18,8 +18,9 @@ module attributes {gpu.container_module} {
     %0 = "op"() : () -> (f32)
     %1 = "op"() : () -> (memref<12xf32>)
     %cst = constant 1 : index
-    "gpu.launch_func"(%cst, %cst, %cst, %cst, %cst, %cst, %0, %1) { kernel = @kernels::@basic_module_structure }
-        : (index, index, index, index, index, index, f32, memref<12xf32>) -> ()
+    gpu.launch_func @kernels::@basic_module_structure
+        blocks in (%cst, %cst, %cst) threads in (%cst, %cst, %cst)
+        args(%0 : f32, %1 : memref<12xf32>)
     return
   }
 }
@@ -63,8 +64,9 @@ module attributes {gpu.container_module} {
     %0 = "op"() : () -> (f32)
     %1 = "op"() : () -> (memref<12xf32>)
     %cst = constant 1 : index
-    "gpu.launch_func"(%cst, %cst, %cst, %cst, %cst, %cst, %0, %1) { kernel = @kernels::@missing_entry_point_abi }
-        : (index, index, index, index, index, index, f32, memref<12xf32>) -> ()
+    gpu.launch_func @kernels::@missing_entry_point_abi
+        blocks in (%cst, %cst, %cst) threads in (%cst, %cst, %cst)
+        args(%0 : f32, %1 : memref<12xf32>)
     return
   }
 }
diff --git a/mlir/test/Conversion/GPUToVulkan/lower-gpu-launch-vulkan-launch.mlir b/mlir/test/Conversion/GPUToVulkan/lower-gpu-launch-vulkan-launch.mlir
index 4a592811f1de7..09054d59371e3 100644
--- a/mlir/test/Conversion/GPUToVulkan/lower-gpu-launch-vulkan-launch.mlir
+++ b/mlir/test/Conversion/GPUToVulkan/lower-gpu-launch-vulkan-launch.mlir
@@ -26,7 +26,10 @@ module attributes {gpu.container_module} {
   func @foo() {
     %0 = alloc() : memref<12xf32>
     %c1 = constant 1 : index
-    "gpu.launch_func"(%c1, %c1, %c1, %c1, %c1, %c1, %0) {kernel = @kernels::@kernel} : (index, index, index, index, index, index, memref<12xf32>) -> ()
+    gpu.launch_func @kernels::@kernel
+        blocks in(%c1, %c1, %c1)
+        threads in(%c1, %c1, %c1)
+        args(%0 : memref<12xf32>)
     return
   }
 }
diff --git a/mlir/test/Conversion/StandardToSPIRV/legalization.mlir b/mlir/test/Conversion/StandardToSPIRV/legalization.mlir
index acbda3540d22c..c5c59613b56e9 100644
--- a/mlir/test/Conversion/StandardToSPIRV/legalization.mlir
+++ b/mlir/test/Conversion/StandardToSPIRV/legalization.mlir
@@ -67,16 +67,17 @@ func @fold_dynamic_stride_subview_with_store(%arg0 : memref<12x32xf32>, %arg1 :
 // CHECK-SAME: [[ARG0:%.*]]: memref<12x32xf32>, [[ARG1:%.*]]: index, [[ARG2:%.*]]: index, [[ARG3:%.*]]: index, [[ARG4:%.*]]: index
 func @fold_static_stride_subview_with_transfer_read(%arg0 : memref<12x32xf32>, %arg1 : index, %arg2 : index, %arg3 : index, %arg4 : index) -> vector<4xf32> {
   // CHECK-NOT: subview
+  // CHECK: [[F1:%.*]] = constant 1.000000e+00 : f32
   // CHECK: [[C2:%.*]] = constant 2 : index
   // CHECK: [[C3:%.*]] = constant 3 : index
   // CHECK: [[STRIDE1:%.*]] = muli [[ARG3]], [[C2]] : index
   // CHECK: [[INDEX1:%.*]] = addi [[ARG1]], [[STRIDE1]] : index
   // CHECK: [[STRIDE2:%.*]] = muli [[ARG4]], [[C3]] : index
   // CHECK: [[INDEX2:%.*]] = addi [[ARG2]], [[STRIDE2]] : index
-  // CHECK: vector.transfer_read [[ARG0]]{{\[}}[[INDEX1]], [[INDEX2]]{{\]}}
-  %f0 = constant 0.0 : f32
+  // CHECK: vector.transfer_read [[ARG0]]{{\[}}[[INDEX1]], [[INDEX2]]{{\]}}, [[F1]] {masked = [false]}
+  %f1 = constant 1.0 : f32
   %0 = subview %arg0[%arg1, %arg2][4, 4][2, 3] : memref<12x32xf32> to memref<4x4xf32, offset:?, strides: [64, 3]>
-  %1 = vector.transfer_read %0[%arg3, %arg4], %f0 : memref<4x4xf32, offset:?, strides: [64, 3]>, vector<4xf32>
+  %1 = vector.transfer_read %0[%arg3, %arg4], %f1 {masked = [false]} : memref<4x4xf32, offset:?, strides: [64, 3]>, vector<4xf32>
   return %1 : vector<4xf32>
 }
 
@@ -90,9 +91,9 @@ func @fold_static_stride_subview_with_transfer_write(%arg0 : memref<12x32xf32>,
   // CHECK: [[INDEX1:%.*]] = addi [[ARG1]], [[STRIDE1]] : index
   // CHECK: [[STRIDE2:%.*]] = muli [[ARG4]], [[C3]] : index
   // CHECK: [[INDEX2:%.*]] = addi [[ARG2]], [[STRIDE2]] : index
-  // CHECK: vector.transfer_write [[ARG5]], [[ARG0]]{{\[}}[[INDEX1]], [[INDEX2]]{{\]}}
+  // CHECK: vector.transfer_write [[ARG5]], [[ARG0]]{{\[}}[[INDEX1]], [[INDEX2]]{{\]}} {masked = [false]}
   %0 = subview %arg0[%arg1, %arg2][4, 4][2, 3] :
     memref<12x32xf32> to memref<4x4xf32, offset:?, strides: [64, 3]>
-  vector.transfer_write %arg5, %0[%arg3, %arg4] : vector<4xf32>, memref<4x4xf32, offset:?, strides: [64, 3]>
+  vector.transfer_write %arg5, %0[%arg3, %arg4] {masked = [false]} : vector<4xf32>, memref<4x4xf32, offset:?, strides: [64, 3]>
   return
 }
diff --git a/mlir/test/Dialect/GPU/invalid.mlir b/mlir/test/Dialect/GPU/invalid.mlir
index cfdb06ac57023..3612b8e0dcc16 100644
--- a/mlir/test/Dialect/GPU/invalid.mlir
+++ b/mlir/test/Dialect/GPU/invalid.mlir
@@ -45,8 +45,7 @@ func @launch_func_too_few_operands(%sz : index) {
 
 func @launch_func_missing_parent_module_attribute(%sz : index) {
   // expected-error@+1 {{expected the closest surrounding module to have the 'gpu.container_module' attribute}}
-  "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz) {foo = "bar"}
-      : (index, index, index, index, index, index) -> ()
+  gpu.launch_func @foo::@bar blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz)
   return
 }
 
@@ -54,8 +53,8 @@ func @launch_func_missing_parent_module_attribute(%sz : index) {
 
 module attributes {gpu.container_module} {
   func @launch_func_missing_callee_attribute(%sz : index) {
-    // expected-error@+1 {{symbol reference attribute 'kernel' must be specified}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz) {foo = "bar"}
+    // expected-error@+1 {{'gpu.launch_func' op requires attribute 'kernel'}}
+    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz)
         : (index, index, index, index, index, index) -> ()
     return
   }
@@ -65,9 +64,8 @@ module attributes {gpu.container_module} {
 
 module attributes {gpu.container_module} {
   func @launch_func_no_function_attribute(%sz : index) {
-    // expected-error@+1 {{symbol reference attribute 'kernel' must be specified}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz) {kernel = 10}
-        : (index, index, index, index, index, index) -> ()
+    // expected-error@+1 {{custom op 'gpu.launch_func' invalid kind of attribute specified}}
+    gpu.launch_func "foo" blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz)
     return
   }
 }
@@ -77,9 +75,7 @@ module attributes {gpu.container_module} {
 module attributes {gpu.container_module} {
   func @launch_func_undefined_module(%sz : index) {
     // expected-error@+1 {{kernel module 'kernels' is undefined}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz)
-    { kernel = @kernels::@kernel_1 }
-        : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz)
     return
   }
 }
@@ -103,9 +99,7 @@ module attributes {gpu.container_module} {
 
   func @launch_func_missing_module_attribute(%sz : index) {
     // expected-error@+1 {{kernel module 'kernels' is undefined}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz)
-    { kernel = @kernels::@kernel_1 }
-        : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz)
     return
   }
 }
@@ -117,9 +111,7 @@ module attributes {gpu.container_module} {
 
   func @launch_func_undefined_function(%sz : index) {
     // expected-error@+1 {{kernel function '@kernels::@kernel_1' is undefined}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz)
-    { kernel = @kernels::@kernel_1 }
-        : (index, index, index, index, index, index) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz)
     return
   }
 }
@@ -135,9 +127,7 @@ module attributes {gpu.container_module} {
 
   func @launch_func_missing_kernel_attr(%sz : index, %arg : !llvm.ptr<float>) {
     // expected-error@+1 {{kernel module 'kernels' is undefined}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz, %arg)
-    {kernel = @kernels::@kernel_1}
-        : (index, index, index, index, index, index, !llvm.ptr<float>) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz) args(%arg : !llvm.ptr<float>)
     return
   }
 }
@@ -153,9 +143,7 @@ module attributes {gpu.container_module} {
 
   func @launch_func_missing_kernel_attr(%sz : index, %arg : !llvm.ptr<float>) {
     // expected-error@+1 {{kernel function is missing the 'gpu.kernel' attribute}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz, %arg)
-    {kernel = @kernels::@kernel_1}
-        : (index, index, index, index, index, index, !llvm.ptr<float>) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz) args(%arg : !llvm.ptr<float>)
     return
   }
 }
@@ -171,10 +159,7 @@ module attributes {gpu.container_module} {
 
   func @launch_func_kernel_operand_size(%sz : index, %arg : !llvm.ptr<float>) {
     // expected-error@+1 {{got 2 kernel operands but expected 1}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz, %arg, %arg)
-        {kernel = @kernels::@kernel_1}
-        : (index, index, index, index, index, index, !llvm.ptr<float>,
-           !llvm.ptr<float>) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz) args(%arg : !llvm.ptr<float>, %arg : !llvm.ptr<float>)
     return
   }
 }
@@ -190,9 +175,17 @@ module attributes {gpu.container_module} {
 
   func @launch_func_kernel_operand_types(%sz : index, %arg : f32) {
     // expected-err@+1 {{type of function argument 0 does not match}}
-    "gpu.launch_func"(%sz, %sz, %sz, %sz, %sz, %sz, %arg)
-        {kernel = @kernels::@kernel_1}
-        : (index, index, index, index, index, index, f32) -> ()
+    gpu.launch_func @kernels::@kernel_1 blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz) args(%arg : f32)
+    return
+  }
+}
+
+// -----
+
+module attributes {gpu.container_module} {
+  func @launch_func_kernel_operand_attr(%sz : index) {
+    // expected-error@+1 {{expected arguments without attributes}}
+    gpu.launch_func @foo::@bar blocks in (%sz, %sz, %sz) threads in (%sz, %sz, %sz) args(%sz : index {foo})
     return
   }
 }
diff --git a/mlir/test/Dialect/GPU/ops.mlir b/mlir/test/Dialect/GPU/ops.mlir
index 23cd6d5c7d0a9..e81b233abfbce 100644
--- a/mlir/test/Dialect/GPU/ops.mlir
+++ b/mlir/test/Dialect/GPU/ops.mlir
@@ -63,7 +63,7 @@ module attributes {gpu.container_module} {
       gpu.return
     }
 
-    gpu.func @kernel_2(%arg0: f32, %arg1: memref<?xf32, 1>) kernel {
+    gpu.func @kernel_2() kernel {
       gpu.return
     }
   }
@@ -74,15 +74,11 @@ module attributes {gpu.container_module} {
     // CHECK: %{{.*}} = constant 8
     %cst = constant 8 : index
 
-    // CHECK: "gpu.launch_func"(%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}) {kernel = @kernels::@kernel_1} : (index, index, index, index, index, index, f32, memref<?xf32, 1>) -> ()
-    "gpu.launch_func"(%cst, %cst, %cst, %cst, %cst, %cst, %0, %1)
-    { kernel = @kernels::@kernel_1}
-        : (index, index, index, index, index, index, f32, memref<?xf32, 1>) -> ()
+    // CHECK: gpu.launch_func @kernels::@kernel_1 blocks in (%{{.*}}, %{{.*}}, %{{.*}}) threads in (%{{.*}}, %{{.*}}, %{{.*}}) args(%{{.*}} : f32, %{{.*}} : memref<?xf32, 1>)
+    gpu.launch_func @kernels::@kernel_1 blocks in (%cst, %cst, %cst) threads in (%cst, %cst, %cst) args(%0 : f32, %1 : memref<?xf32, 1>)
 
-    // CHECK: "gpu.launch_func"(%{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}, %{{.*}}) {kernel = @kernels::@kernel_2} : (index, index, index, index, index, index, f32, memref<?xf32, 1>) -> ()
-    "gpu.launch_func"(%cst, %cst, %cst, %cst, %cst, %cst, %0, %1)
-    { kernel = @kernels::@kernel_2}
-        : (index, index, index, index, index, index, f32, memref<?xf32, 1>) -> ()
+    // CHECK: gpu.launch_func @kernels::@kernel_2 blocks in (%{{.*}}, %{{.*}}, %{{.*}}) threads in (%{{.*}}, %{{.*}}, %{{.*}})
+    gpu.launch_func @kernels::@kernel_2 blocks in (%cst, %cst, %cst) threads in (%cst, %cst, %cst)
 
     return
   }
diff --git a/mlir/test/Dialect/GPU/outlining.mlir b/mlir/test/Dialect/GPU/outlining.mlir
index 5fd8b6ce79cd5..e2f16fe96a08a 100644
--- a/mlir/test/Dialect/GPU/outlining.mlir
+++ b/mlir/test/Dialect/GPU/outlining.mlir
@@ -21,7 +21,7 @@ func @launch() {
   // CHECK: %[[BDIMZ:.*]] = constant 28
   %bDimZ = constant 28 : index
 
-  // CHECK: "gpu.launch_func"(%[[GDIMX]], %[[GDIMY]], %[[GDIMZ]], %[[BDIMX]], %[[BDIMY]], %[[BDIMZ]], %[[ARG0]], %[[ARG1]]) {kernel = @launch_kernel::@launch_kernel} : (index, index, index, index, index, index, f32, memref<?xf32, 1>) -> ()
+  // CHECK: gpu.launch_func @launch_kernel::@launch_kernel blocks in (%[[GDIMX]], %[[GDIMY]], %[[GDIMZ]]) threads in (%[[BDIMX]], %[[BDIMY]], %[[BDIMZ]]) args(%[[ARG0]] : f32, %[[ARG1]] : memref<?xf32, 1>)
   // CHECK-NOT: gpu.launch blocks
   gpu.launch blocks(%bx, %by, %bz) in (%grid_x = %gDimX, %grid_y = %gDimY,
                                        %grid_z = %gDimZ)
@@ -64,14 +64,14 @@ func @launch() {
 func @multiple_launches() {
   // CHECK: %[[CST:.*]] = constant 8 : index
   %cst = constant 8 : index
-  // CHECK: "gpu.launch_func"(%[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]]) {kernel = @multiple_launches_kernel::@multiple_launches_kernel} : (index, index, index, index, index, index) -> ()
+  // CHECK: gpu.launch_func @multiple_launches_kernel::@multiple_launches_kernel blocks in (%[[CST]], %[[CST]], %[[CST]]) threads in (%[[CST]], %[[CST]], %[[CST]])
   gpu.launch blocks(%bx, %by, %bz) in (%grid_x = %cst, %grid_y = %cst,
                                        %grid_z = %cst)
              threads(%tx, %ty, %tz) in (%block_x = %cst, %block_y = %cst,
                                         %block_z = %cst) {
     gpu.terminator
   }
-  // CHECK: "gpu.launch_func"(%[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]]) {kernel = @multiple_launches_kernel_0::@multiple_launches_kernel} : (index, index, index, index, index, index) -> ()
+  // CHECK: gpu.launch_func @multiple_launches_kernel_0::@multiple_launches_kernel blocks in (%[[CST]], %[[CST]], %[[CST]]) threads in (%[[CST]], %[[CST]], %[[CST]])
   gpu.launch blocks(%bx2, %by2, %bz2) in (%grid_x2 = %cst, %grid_y2 = %cst,
                                           %grid_z2 = %cst)
              threads(%tx2, %ty2, %tz2) in (%block_x2 = %cst, %block_y2 = %cst,
@@ -95,7 +95,7 @@ func @extra_constants_not_inlined(%arg0: memref<?xf32>) {
   %cst2 = constant 2 : index
   %c0 = constant 0 : index
   %cst3 = "secret_constant"() : () -> index
-  // CHECK: "gpu.launch_func"(%[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %{{.*}}, %{{.*}}) {kernel = @extra_constants_not_inlined_kernel::@extra_constants_not_inlined_kernel} : (index, index, index, index, index, index, memref<?xf32>, index) -> ()
+  // CHECK: gpu.launch_func @extra_constants_not_inlined_kernel::@extra_constants_not_inlined_kernel blocks in (%[[CST]], %[[CST]], %[[CST]]) threads in (%[[CST]], %[[CST]], %[[CST]]) args({{.*}} : memref<?xf32>, {{.*}} : index)
   gpu.launch blocks(%bx, %by, %bz) in (%grid_x = %cst, %grid_y = %cst,
                                        %grid_z = %cst)
              threads(%tx, %ty, %tz) in (%block_x = %cst, %block_y = %cst,
@@ -119,7 +119,7 @@ func @extra_constants(%arg0: memref<?xf32>) {
   %cst2 = constant 2 : index
   %c0 = constant 0 : index
   %cst3 = dim %arg0, %c0 : memref<?xf32>
-  // CHECK: "gpu.launch_func"(%[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[ARG0]]) {kernel = @extra_constants_kernel::@extra_constants_kernel} : (index, index, index, index, index, index, memref<?xf32>) -> ()
+  // CHECK: gpu.launch_func @extra_constants_kernel::@extra_constants_kernel blocks in (%[[CST]], %[[CST]], %[[CST]]) threads in (%[[CST]], %[[CST]], %[[CST]]) args(%[[ARG0]] : memref<?xf32>)
   gpu.launch blocks(%bx, %by, %bz) in (%grid_x = %cst, %grid_y = %cst,
                                        %grid_z = %cst)
              threads(%tx, %ty, %tz) in (%block_x = %cst, %block_y = %cst,
@@ -130,7 +130,7 @@ func @extra_constants(%arg0: memref<?xf32>) {
   return
 }
 
-// CHECK-LABEL: func @extra_constants_kernel
+// CHECK-LABEL: func @extra_constants_kernel(
 // CHECK-SAME: %[[KARG0:.*]]: memref<?xf32>
 // CHECK: constant 2
 // CHECK: constant 0
@@ -147,7 +147,7 @@ func @extra_constants_noarg(%arg0: memref<?xf32>, %arg1: memref<?xf32>) {
   %c0 = constant 0 : index
   // CHECK: dim %[[ARG1]]
   %cst3 = dim %arg1, %c0 : memref<?xf32>
-  // CHECK: "gpu.launch_func"(%[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[CST]], %[[ARG0]], %{{.*}}) {kernel = @extra_constants_noarg_kernel::@extra_constants_noarg_kernel} : (index, index, index, index, index, index, memref<?xf32>, index) -> ()
+  // CHECK: gpu.launch_func @extra_constants_noarg_kernel::@extra_constants_noarg_kernel blocks in (%[[CST]], %[[CST]], %[[CST]]) threads in (%[[CST]], %[[CST]], %[[CST]]) args(%[[ARG0]] : memref<?xf32>, {{.*}} : index)
   gpu.launch blocks(%bx, %by, %bz) in (%grid_x = %cst, %grid_y = %cst,
                                        %grid_z = %cst)
              threads(%tx, %ty, %tz) in (%block_x = %cst, %block_y = %cst,
@@ -158,7 +158,7 @@ func @extra_constants_noarg(%arg0: memref<?xf32>, %arg1: memref<?xf32>) {
   return
 }
 
-// CHECK-LABEL: func @extra_constants_noarg_kernel
+// CHECK-LABEL: func @extra_constants_noarg_kernel(
 // CHECK-SAME: %[[KARG0:.*]]: memref<?xf32>, %[[KARG1:.*]]: index
 // CHECK: %[[KCST:.*]] = constant 2
 // CHECK: "use"(%[[KCST]], %[[KARG0]], %[[KARG1]])
diff --git a/mlir/test/Dialect/SCF/bufferize.mlir b/mlir/test/Dialect/SCF/bufferize.mlir
new file mode 100644
index 0000000000000..01b353da83ed8
--- /dev/null
+++ b/mlir/test/Dialect/SCF/bufferize.mlir
@@ -0,0 +1,42 @@
+// RUN: mlir-opt %s -scf-bufferize | FileCheck %s
+
+// CHECK-LABEL:   func @if(
+// CHECK-SAME:             %[[PRED:.*]]: i1,
+// CHECK-SAME:             %[[TRUE_TENSOR:.*]]: tensor<?xf32>,
+// CHECK-SAME:             %[[FALSE_TENSOR:.*]]: tensor<?xf32>) -> tensor<?xf32> {
+// CHECK:           %[[RESULT_MEMREF:.*]] = scf.if %[[PRED]] -> (memref<?xf32>) {
+// CHECK:             %[[TRUE_MEMREF:.*]] = tensor_to_memref %[[TRUE_TENSOR]] : memref<?xf32>
+// CHECK:             scf.yield %[[TRUE_MEMREF]] : memref<?xf32>
+// CHECK:           } else {
+// CHECK:             %[[FALSE_MEMREF:.*]] = tensor_to_memref %[[FALSE_TENSOR]] : memref<?xf32>
+// CHECK:             scf.yield %[[FALSE_MEMREF]] : memref<?xf32>
+// CHECK:           }
+// CHECK:           %[[RESULT_TENSOR:.*]] = tensor_load %[[RESULT_MEMREF:.*]] : memref<?xf32>
+// CHECK:           return %[[RESULT_TENSOR]] : tensor<?xf32>
+// CHECK:         }
+func @if(%pred: i1, %true_val: tensor<?xf32>, %false_val: tensor<?xf32>) -> tensor<?xf32> {
+  %0 = scf.if %pred -> (tensor<?xf32>) {
+    scf.yield %true_val : tensor<?xf32>
+  } else {
+    scf.yield %false_val : tensor<?xf32>
+  }
+  return %0 : tensor<?xf32>
+}
+
+// CHECK-LABEL:   func @for(
+// CHECK-SAME:              %[[TENSOR:.*]]: tensor<f32>,
+// CHECK-SAME:              %[[LB:.*]]: index, %[[UB:.*]]: index,
+// CHECK-SAME:              %[[STEP:.*]]: index) -> tensor<f32> {
+// CHECK:           %[[MEMREF:.*]] = tensor_to_memref %[[TENSOR]] : memref<f32>
+// CHECK:           %[[RESULT_MEMREF:.*]] = scf.for %[[VAL_6:.*]] = %[[LB]] to %[[UB]] step %[[STEP]] iter_args(%[[ITER:.*]] = %[[MEMREF]]) -> (memref<f32>) {
+// CHECK:             scf.yield %[[ITER]] : memref<f32>
+// CHECK:           }
+// CHECK:           %[[VAL_8:.*]] = tensor_load %[[VAL_9:.*]] : memref<f32>
+// CHECK:           return %[[VAL_8]] : tensor<f32>
+// CHECK:         }
+func @for(%arg0: tensor<f32>, %lb: index, %ub: index, %step: index) -> tensor<f32> {
+  %ret = scf.for %iv = %lb to %ub step %step iter_args(%iter = %arg0) -> tensor<f32> {
+    scf.yield %iter : tensor<f32>
+  }
+  return %ret : tensor<f32>
+}
diff --git a/mlir/test/Dialect/Shape/bufferize.mlir b/mlir/test/Dialect/Shape/bufferize.mlir
index 7393de1014666..cb65a5d42d4b3 100644
--- a/mlir/test/Dialect/Shape/bufferize.mlir
+++ b/mlir/test/Dialect/Shape/bufferize.mlir
@@ -1,12 +1,20 @@
 // RUN: mlir-opt -split-input-file -shape-bufferize <%s | FileCheck %s
 
 // -----
-// Check that shape.assuming returns a memref.
-//
-// CHECK-LABEL: @shape_assuming_returns_memref
-func @shape_assuming_returns_memref() {
+
+// CHECK-LABEL:   func @shape_assuming() {
+// CHECK:           %[[WTRUE:.*]] = shape.const_witness true
+// CHECK:           %[[MEMREF:.*]] = shape.assuming %[[WTRUE]] -> (memref<2xf16>) {
+// CHECK:             %[[TENSOR_VAL:.*]] = "test.source"() : () -> tensor<2xf16>
+// CHECK:             %[[YIELDED_MEMREF:.*]] = tensor_to_memref %[[TENSOR_VAL]] : memref<2xf16>
+// CHECK:             shape.assuming_yield %[[YIELDED_MEMREF]] : memref<2xf16>
+// CHECK:           }
+// CHECK:           %[[TENSOR:.*]] = tensor_load %[[MEMREF:.*]] : memref<2xf16>
+// CHECK:           "test.sink"(%[[TENSOR]]) : (tensor<2xf16>) -> ()
+// CHECK:           return
+// CHECK:         }
+func @shape_assuming() {
   %0 = shape.const_witness true
-  // CHECK: shape.assuming %{{.*}} -> (memref<2xf16>) {
   %1 = shape.assuming %0 -> (tensor<2xf16>) {
     %2 = "test.source"() : () -> (tensor<2xf16>)
     shape.assuming_yield %2 : tensor<2xf16>
diff --git a/mlir/test/Dialect/Standard/invalid.mlir b/mlir/test/Dialect/Standard/invalid.mlir
index 72fe5c227578a..5bd94bd39d915 100644
--- a/mlir/test/Dialect/Standard/invalid.mlir
+++ b/mlir/test/Dialect/Standard/invalid.mlir
@@ -102,3 +102,132 @@ func @transpose_wrong_type(%v : memref<?x?xf32, affine_map<(i, j)[off, M]->(off
   // expected-error @+1 {{output type 'memref<?x?xf32, affine_map<(d0, d1)[s0, s1] -> (d0 * s1 + s0 + d1)>>' does not match transposed input type 'memref<?x?xf32, affine_map<(d0, d1)[s0, s1] -> (d0 * s1 + s0 + d1)>>'}}
   transpose %v (i, j) -> (j, i) : memref<?x?xf32, affine_map<(i, j)[off, M]->(off + M * i + j)>> to memref<?x?xf32, affine_map<(i, j)[off, M]->(off + M * i + j)>>
 }
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_too_many_offsets
+func @memref_reinterpret_cast_too_many_offsets(%in: memref<?xf32>) {
+  // expected-error @+1 {{expected 1 offset values}}
+  %out = memref_reinterpret_cast %in to
+           offset: [0, 0], sizes: [10, 10], strides: [10, 1]
+           : memref<?xf32> to memref<10x10xf32, offset: 0, strides: [10, 1]>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_incompatible_element_types
+func @memref_reinterpret_cast_incompatible_element_types(%in: memref<*xf32>) {
+  // expected-error @+1 {{different element types specified}}
+  %out = memref_reinterpret_cast %in to
+           offset: [0], sizes: [10], strides: [1]
+         : memref<*xf32> to memref<10xi32, offset: 0, strides: [1]>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_incompatible_memory_space
+func @memref_reinterpret_cast_incompatible_memory_space(%in: memref<*xf32>) {
+  // expected-error @+1 {{different memory spaces specified}}
+  %out = memref_reinterpret_cast %in to
+           offset: [0], sizes: [10], strides: [1]
+         : memref<*xf32> to memref<10xi32, offset: 0, strides: [1], 2>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_offset_mismatch
+func @memref_reinterpret_cast_offset_mismatch(%in: memref<?xf32>) {
+  // expected-error @+1 {{expected result type with offset = 0 instead of 1}}
+  %out = memref_reinterpret_cast %in to
+           offset: [1], sizes: [10], strides: [1]
+         : memref<?xf32> to memref<10xf32>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_size_mismatch
+func @memref_reinterpret_cast_size_mismatch(%in: memref<*xf32>) {
+  // expected-error @+1 {{expected result type with size = 10 instead of 1 in dim = 0}}
+  %out = memref_reinterpret_cast %in to
+           offset: [0], sizes: [10], strides: [1]
+         : memref<*xf32> to memref<1xf32, offset: 0, strides: [1]>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_stride_mismatch
+func @memref_reinterpret_cast_offset_mismatch(%in: memref<?xf32>) {
+  // expected-error @+1 {{expected result type with stride = 2 instead of 1 in dim = 0}}
+  %out = memref_reinterpret_cast %in to
+           offset: [0], sizes: [10], strides: [2]
+         : memref<?xf32> to memref<10xf32>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: func @memref_reinterpret_cast_dynamic_size_mismatch
+func @memref_reinterpret_cast_offset_mismatch(%in: memref<?xf32>) {
+  %c0 = constant 0 : index
+  %c10 = constant 10 : index
+  // expected-error @+1 {{expected result type with size = 10 instead of -1 in dim = 0}}
+  %out = memref_reinterpret_cast %in to
+           offset: [%c0], sizes: [10, %c10], strides: [%c10, 1]
+           : memref<?xf32> to memref<?x?xf32, offset: ?, strides: [?, 1]>
+  return
+}
+
+// -----
+
+// CHECK-LABEL: memref_reshape_element_type_mismatch
+func @memref_reshape_element_type_mismatch(
+       %buf: memref<*xf32>, %shape: memref<1xi32>) {
+  // expected-error @+1 {{element types of source and destination memref types should be the same}}
+  memref_reshape %buf(%shape) : (memref<*xf32>, memref<1xi32>) -> memref<?xi32>
+}
+
+// -----
+
+// CHECK-LABEL: memref_reshape_dst_ranked_shape_unranked
+func @memref_reshape_dst_ranked_shape_unranked(
+       %buf: memref<*xf32>, %shape: memref<?xi32>) {
+  // expected-error @+1 {{cannot use shape operand with dynamic length to reshape to statically-ranked memref type}}
+  memref_reshape %buf(%shape) : (memref<*xf32>, memref<?xi32>) -> memref<?xf32>
+}
+
+// -----
+
+// CHECK-LABEL: memref_reshape_dst_shape_rank_mismatch
+func @memref_reshape_dst_shape_rank_mismatch(
+       %buf: memref<*xf32>, %shape: memref<1xi32>) {
+  // expected-error @+1 {{length of shape operand differs from the result's memref rank}}
+  memref_reshape %buf(%shape)
+    : (memref<*xf32>, memref<1xi32>) -> memref<?x?xf32>
+}
+
+// -----
+
+// CHECK-LABEL: memref_reshape_src_affine_map_is_not_identity
+func @memref_reshape_src_affine_map_is_not_identity(
+        %buf: memref<4x4xf32, offset: 0, strides: [3, 2]>,
+        %shape: memref<1xi32>) {
+  // expected-error @+1 {{source memref type should have identity affine map}}
+  memref_reshape %buf(%shape)
+    : (memref<4x4xf32, offset: 0, strides: [3, 2]>, memref<1xi32>)
+    -> memref<8xf32>
+}
+
+// -----
+
+// CHECK-LABEL: memref_reshape_result_affine_map_is_not_identity
+func @memref_reshape_result_affine_map_is_not_identity(
+        %buf: memref<4x4xf32>, %shape: memref<1xi32>) {
+  // expected-error @+1 {{result memref type should have identity affine map}}
+  memref_reshape %buf(%shape)
+    : (memref<4x4xf32>, memref<1xi32>) -> memref<8xf32, offset: 0, strides: [2]>
+}
diff --git a/mlir/test/Dialect/Standard/ops.mlir b/mlir/test/Dialect/Standard/ops.mlir
index b11c9534cc2de..fcf2325fd2e8d 100644
--- a/mlir/test/Dialect/Standard/ops.mlir
+++ b/mlir/test/Dialect/Standard/ops.mlir
@@ -54,3 +54,26 @@ func @atan2(%arg0 : f32, %arg1 : f32) -> f32 {
   %result = atan2 %arg0, %arg1 : f32
   return %result : f32
 }
+
+// CHECK-LABEL: func @memref_reinterpret_cast
+func @memref_reinterpret_cast(%in: memref<?xf32>)
+    -> memref<10x?xf32, offset: ?, strides: [?, 1]> {
+  %c0 = constant 0 : index
+  %c10 = constant 10 : index
+  %out = memref_reinterpret_cast %in to
+           offset: [%c0], sizes: [10, %c10], strides: [%c10, 1]
+           : memref<?xf32> to memref<10x?xf32, offset: ?, strides: [?, 1]>
+  return %out : memref<10x?xf32, offset: ?, strides: [?, 1]>
+}
+
+// CHECK-LABEL: func @memref_reshape(
+func @memref_reshape(%unranked: memref<*xf32>, %shape1: memref<1xi32>,
+         %shape2: memref<2xi32>, %shape3: memref<?xi32>) -> memref<*xf32> {
+  %dyn_vec = memref_reshape %unranked(%shape1)
+               : (memref<*xf32>, memref<1xi32>) -> memref<?xf32>
+  %dyn_mat = memref_reshape %dyn_vec(%shape2)
+               : (memref<?xf32>, memref<2xi32>) -> memref<?x?xf32>
+  %new_unranked = memref_reshape %dyn_mat(%shape3)
+               : (memref<?x?xf32>, memref<?xi32>) -> memref<*xf32>
+  return %new_unranked : memref<*xf32>
+}
diff --git a/mlir/test/mlir-cpu-runner/async.mlir b/mlir/test/mlir-cpu-runner/async.mlir
new file mode 100644
index 0000000000000..ce9cdb14d4c72
--- /dev/null
+++ b/mlir/test/mlir-cpu-runner/async.mlir
@@ -0,0 +1,77 @@
+// RUN:   mlir-opt %s -convert-async-to-llvm                                   \
+// RUN:               -convert-linalg-to-loops                                 \
+// RUN:               -convert-linalg-to-llvm                                  \
+// RUN:               -convert-std-to-llvm                                     \
+// RUN: | mlir-cpu-runner                                                      \
+// RUN:     -e main -entry-point-result=void -O0                               \
+// RUN:     -shared-libs=%linalg_test_lib_dir/libmlir_c_runner_utils%shlibext  \
+// RUN:     -shared-libs=%linalg_test_lib_dir/libmlir_runner_utils%shlibext    \
+// RUN:     -shared-libs=%linalg_test_lib_dir/libmlir_async_runtime%shlibext   \
+// RUN: | FileCheck %s
+
+func @main() {
+  %i0 = constant 0 : index
+  %i1 = constant 1 : index
+  %i2 = constant 2 : index
+  %i3 = constant 3 : index
+
+  %c0 = constant 0.0 : f32
+  %c1 = constant 1.0 : f32
+  %c2 = constant 2.0 : f32
+  %c3 = constant 3.0 : f32
+  %c4 = constant 4.0 : f32
+
+  %A = alloc() : memref<4xf32>
+  linalg.fill(%A, %c0) : memref<4xf32>, f32
+
+  // CHECK: [0, 0, 0, 0]
+  %U = memref_cast %A :  memref<4xf32> to memref<*xf32>
+  call @print_memref_f32(%U): (memref<*xf32>) -> ()
+
+  // CHECK: Current thread id: [[MAIN:.*]]
+  // CHECK: [1, 0, 0, 0]
+  store %c1, %A[%i0]: memref<4xf32>
+  call @mlirAsyncRuntimePrintCurrentThreadId(): () -> ()
+  call @print_memref_f32(%U): (memref<*xf32>) -> ()
+
+  %outer = async.execute {
+    // CHECK: Current thread id: [[THREAD0:.*]]
+    // CHECK: [1, 2, 0, 0]
+    store %c2, %A[%i1]: memref<4xf32>
+    call @mlirAsyncRuntimePrintCurrentThreadId(): () -> ()
+    call @print_memref_f32(%U): (memref<*xf32>) -> ()
+
+    %inner = async.execute {
+      // CHECK: Current thread id: [[THREAD1:.*]]
+      // CHECK: [1, 2, 3, 0]
+      store %c3, %A[%i2]: memref<4xf32>
+      call @mlirAsyncRuntimePrintCurrentThreadId(): () -> ()
+      call @print_memref_f32(%U): (memref<*xf32>) -> ()
+
+      async.yield
+    }
+    async.await %inner : !async.token
+
+    // CHECK: Current thread id: [[THREAD2:.*]]
+    // CHECK: [1, 2, 3, 4]
+    store %c4, %A[%i3]: memref<4xf32>
+    call @mlirAsyncRuntimePrintCurrentThreadId(): () -> ()
+    call @print_memref_f32(%U): (memref<*xf32>) -> ()
+
+    async.yield
+  }
+  async.await %outer : !async.token
+
+  // CHECK: Current thread id: [[MAIN]]
+  // CHECK: [1, 2, 3, 4]
+  call @mlirAsyncRuntimePrintCurrentThreadId(): () -> ()
+  call @print_memref_f32(%U): (memref<*xf32>) -> ()
+
+  dealloc %A : memref<4xf32>
+
+  return
+}
+
+func @mlirAsyncRuntimePrintCurrentThreadId() -> ()
+
+func @print_memref_f32(memref<*xf32>) attributes { llvm.emit_c_interface }
diff --git a/mlir/test/mlir-tblgen/op-decl.td b/mlir/test/mlir-tblgen/op-decl.td
index 4ff77dc2c3f74..29438f1836a77 100644
--- a/mlir/test/mlir-tblgen/op-decl.td
+++ b/mlir/test/mlir-tblgen/op-decl.td
@@ -33,7 +33,9 @@ def NS_AOp : NS_Op<"a_op", [IsolatedFromAbove, IsolatedFromAbove]> {
     AnyRegion:$someRegion,
     VariadicRegion<AnyRegion>:$someRegions
   );
-  let builders = [OpBuilder<"Value val">];
+  let builders = [OpBuilderDAG<(ins "Value":$val)>,
+                  OpBuilderDAG<(ins CArg<"int", "0">:$integer)>,
+                  OpBuilder<"double deprecatedForm">];
   let parser = [{ foo }];
   let printer = [{ bar }];
   let verifier = [{ baz }];
@@ -81,6 +83,8 @@ def NS_AOp : NS_Op<"a_op", [IsolatedFromAbove, IsolatedFromAbove]> {
 // CHECK:   ::mlir::FloatAttr attr2Attr()
 // CHECK:   ::llvm::Optional< ::llvm::APFloat > attr2();
 // CHECK:   static void build(::mlir::OpBuilder &odsBuilder, ::mlir::OperationState &odsState, Value val);
+// CHECK:   static void build(::mlir::OpBuilder &odsBuilder, ::mlir::OperationState &odsState, int integer = 0);
+// CHECK:   static void build(::mlir::OpBuilder &odsBuilder, ::mlir::OperationState &odsState, double deprecatedForm);
 // CHECK:   static void build(::mlir::OpBuilder &odsBuilder, ::mlir::OperationState &odsState, ::mlir::Type r, ::mlir::TypeRange s, ::mlir::Value a, ::mlir::ValueRange b, ::mlir::IntegerAttr attr1, /*optional*/::mlir::FloatAttr attr2, unsigned someRegionsCount)
 // CHECK:   static void build(::mlir::OpBuilder &odsBuilder, ::mlir::OperationState &odsState, ::mlir::Type r, ::mlir::TypeRange s, ::mlir::Value a, ::mlir::ValueRange b, uint32_t attr1, /*optional*/::mlir::FloatAttr attr2, unsigned someRegionsCount)
 // CHECK:   static void build(::mlir::OpBuilder &, ::mlir::OperationState &odsState, ::mlir::TypeRange resultTypes, ::mlir::ValueRange operands, ::llvm::ArrayRef<::mlir::NamedAttribute> attributes, unsigned numRegions)
@@ -250,7 +254,7 @@ def NS_JOp : NS_Op<"op_with_InferTypeOpInterface_interface", [DeclareOpInterface
 
 def NS_SkipDefaultBuildersOp : NS_Op<"skip_default_builders", []> {
   let skipDefaultBuilders = 1;
-  let builders = [OpBuilder<"Value val">];
+  let builders = [OpBuilderDAG<(ins "Value":$val)>];
 }
 
 // CHECK-LABEL: NS::SkipDefaultBuildersOp declarations
diff --git a/mlir/test/mlir-tblgen/op-error.td b/mlir/test/mlir-tblgen/op-error.td
new file mode 100644
index 0000000000000..b5fea66287a97
--- /dev/null
+++ b/mlir/test/mlir-tblgen/op-error.td
@@ -0,0 +1,36 @@
+// RUN: not mlir-tblgen -gen-op-decls -I %S/../../include -DERROR1 %s 2>&1 | FileCheck --check-prefix=ERROR1 %s
+// RUN: not mlir-tblgen -gen-op-decls -I %S/../../include -DERROR2 %s 2>&1 | FileCheck --check-prefix=ERROR2 %s
+// RUN: not mlir-tblgen -gen-op-decls -I %S/../../include -DERROR3 %s 2>&1 | FileCheck --check-prefix=ERROR3 %s
+
+include "mlir/IR/OpBase.td"
+
+def Test_Dialect : Dialect {
+  let name = "test_dialect";
+}
+
+#ifdef ERROR1
+// ERROR1: error: expected 'ins'
+def OpInsMissing : Op<Test_Dialect, "ins_missing"> {
+  let builders = [
+    OpBuilderDAG<(outs)>
+  ];
+}
+#endif
+
+#ifdef ERROR2
+// ERROR2: error: expected an argument with default value after other arguments with default values
+def OpDefaultValueNotTrailing : Op<Test_Dialect, "default_value"> {
+  let builders = [
+    OpBuilderDAG<(ins CArg<"int", "42">, "int")>
+  ];
+}
+#endif
+
+#ifdef ERROR3
+// ERROR3: error: expected an argument with default value after other arguments with default values
+def OpDefaultValueNotTrailing : Op<Test_Dialect, "default_value"> {
+  let builders = [
+    OpBuilderDAG<(ins CArg<"int", "42">, CArg<"int">)>
+  ];
+}
+#endif
diff --git a/mlir/test/mlir-vulkan-runner/addf.mlir b/mlir/test/mlir-vulkan-runner/addf.mlir
index 6cb7cdec3442e..490ed2f20f258 100644
--- a/mlir/test/mlir-vulkan-runner/addf.mlir
+++ b/mlir/test/mlir-vulkan-runner/addf.mlir
@@ -37,8 +37,9 @@ module attributes {
 
     %cst1 = constant 1 : index
     %cst8 = constant 8 : index
-    "gpu.launch_func"(%cst8, %cst1, %cst1, %cst1, %cst1, %cst1, %arg0, %arg1, %arg2) { kernel = @kernels::@kernel_add }
-        : (index, index, index, index, index, index, memref<8xf32>, memref<8xf32>, memref<8xf32>) -> ()
+    gpu.launch_func @kernels::@kernel_add
+        blocks in (%cst8, %cst1, %cst1) threads in (%cst1, %cst1, %cst1)
+        args(%arg0 : memref<8xf32>, %arg1 : memref<8xf32>, %arg2 : memref<8xf32>)
     %arg6 = memref_cast %arg5 : memref<?xf32> to memref<*xf32>
     call @print_memref_f32(%arg6) : (memref<*xf32>) -> ()
     return
diff --git a/mlir/test/mlir-vulkan-runner/addi.mlir b/mlir/test/mlir-vulkan-runner/addi.mlir
index 696c5015565db..6b289cdb7290a 100644
--- a/mlir/test/mlir-vulkan-runner/addi.mlir
+++ b/mlir/test/mlir-vulkan-runner/addi.mlir
@@ -36,8 +36,9 @@ module attributes {
 
     %cst1 = constant 1 : index
     %cst8 = constant 8 : index
-    "gpu.launch_func"(%cst8, %cst8, %cst8, %cst1, %cst1, %cst1, %arg0, %arg1, %arg2) { kernel = @kernels::@kernel_addi }
-        : (index, index, index, index, index, index, memref<8xi32>, memref<8x8xi32>, memref<8x8x8xi32>) -> ()
+    gpu.launch_func @kernels::@kernel_addi
+        blocks in (%cst8, %cst8, %cst8) threads in (%cst1, %cst1, %cst1)
+        args(%arg0 : memref<8xi32>, %arg1 : memref<8x8xi32>, %arg2 : memref<8x8x8xi32>)
     %arg6 = memref_cast %arg5 : memref<?x?x?xi32> to memref<*xi32>
     call @print_memref_i32(%arg6) : (memref<*xi32>) -> ()
     return
diff --git a/mlir/test/mlir-vulkan-runner/addi8.mlir b/mlir/test/mlir-vulkan-runner/addi8.mlir
index eeb5222856961..27bfa7d02396c 100644
--- a/mlir/test/mlir-vulkan-runner/addi8.mlir
+++ b/mlir/test/mlir-vulkan-runner/addi8.mlir
@@ -37,8 +37,9 @@ module attributes {
 
     %cst1 = constant 1 : index
     %cst8 = constant 8 : index
-    "gpu.launch_func"(%cst8, %cst8, %cst8, %cst1, %cst1, %cst1, %arg0, %arg1, %arg2) { kernel = @kernels::@kernel_addi }
-        : (index, index, index, index, index, index, memref<8xi8>, memref<8x8xi8>, memref<8x8x8xi32>) -> ()
+    gpu.launch_func @kernels::@kernel_addi
+        blocks in (%cst8, %cst8, %cst8) threads in (%cst1, %cst1, %cst1)
+        args(%arg0 : memref<8xi8>, %arg1 : memref<8x8xi8>, %arg2 : memref<8x8x8xi32>)
     %arg6 = memref_cast %arg5 : memref<?x?x?xi32> to memref<*xi32>
     call @print_memref_i32(%arg6) : (memref<*xi32>) -> ()
     return
diff --git a/mlir/test/mlir-vulkan-runner/mulf.mlir b/mlir/test/mlir-vulkan-runner/mulf.mlir
index 0abcb53ebfe6b..a6b7b1b23e5dd 100644
--- a/mlir/test/mlir-vulkan-runner/mulf.mlir
+++ b/mlir/test/mlir-vulkan-runner/mulf.mlir
@@ -38,8 +38,9 @@ module attributes {
 
     %cst1 = constant 1 : index
     %cst4 = constant 4 : index
-    "gpu.launch_func"(%cst4, %cst4, %cst1, %cst1, %cst1, %cst1, %arg0, %arg1, %arg2) { kernel = @kernels::@kernel_mul }
-        : (index, index, index, index, index, index, memref<4x4xf32>, memref<4x4xf32>, memref<4x4xf32>) -> ()
+    gpu.launch_func @kernels::@kernel_mul
+        blocks in (%cst4, %cst4, %cst1) threads in(%cst1, %cst1, %cst1)
+        args(%arg0 : memref<4x4xf32>, %arg1 : memref<4x4xf32>, %arg2 : memref<4x4xf32>)
     %arg6 = memref_cast %arg5 : memref<?x?xf32> to memref<*xf32>
     call @print_memref_f32(%arg6) : (memref<*xf32>) -> ()
     return
diff --git a/mlir/test/mlir-vulkan-runner/subf.mlir b/mlir/test/mlir-vulkan-runner/subf.mlir
index 77c1f8841e8be..00940688eb163 100644
--- a/mlir/test/mlir-vulkan-runner/subf.mlir
+++ b/mlir/test/mlir-vulkan-runner/subf.mlir
@@ -40,8 +40,9 @@ module attributes {
     %cst1 = constant 1 : index
     %cst4 = constant 4 : index
     %cst8 = constant 8 : index
-    "gpu.launch_func"(%cst8, %cst4, %cst4, %cst1, %cst1, %cst1, %arg0, %arg1, %arg2) { kernel = @kernels::@kernel_sub }
-        : (index, index, index, index, index, index, memref<8x4x4xf32>, memref<4x4xf32>, memref<8x4x4xf32>) -> ()
+    gpu.launch_func @kernels::@kernel_sub
+        blocks in (%cst8, %cst4, %cst4) threads in (%cst1, %cst1, %cst1)
+        args(%arg0 : memref<8x4x4xf32>, %arg1 : memref<4x4xf32>, %arg2 : memref<8x4x4xf32>)
     %arg6 = memref_cast %arg5 : memref<?x?x?xf32> to memref<*xf32>
     call @print_memref_f32(%arg6) : (memref<*xf32>) -> ()
     return
diff --git a/mlir/test/mlir-vulkan-runner/time.mlir b/mlir/test/mlir-vulkan-runner/time.mlir
index 21b4b76d1df08..0bf92966cdec7 100644
--- a/mlir/test/mlir-vulkan-runner/time.mlir
+++ b/mlir/test/mlir-vulkan-runner/time.mlir
@@ -44,8 +44,9 @@ module attributes {
 
     %cst1 = constant 1 : index
     %cst128 = constant 128 : index
-    "gpu.launch_func"(%cst128, %cst1, %cst1, %cst128, %cst1, %cst1, %arg0, %arg1, %arg2) { kernel = @kernels::@kernel_add }
-        : (index, index, index, index, index, index, memref<16384xf32>, memref<16384xf32>, memref<16384xf32>) -> ()
+    gpu.launch_func @kernels::@kernel_add
+        blocks in (%cst128, %cst1, %cst1) threads in (%cst128, %cst1, %cst1)
+        args(%arg0 : memref<16384xf32>, %arg1 : memref<16384xf32>, %arg2 : memref<16384xf32>)
     %arg6 = memref_cast %arg5 : memref<?xf32> to memref<*xf32>
     return
   }
diff --git a/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp b/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp
index 3bcf021145556..f296b32849079 100644
--- a/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp
+++ b/mlir/tools/mlir-tblgen/OpDefinitionsGen.cpp
@@ -1144,6 +1144,82 @@ void OpEmitter::genUseAttrAsResultTypeBuilder() {
   body << "  }\n";
 }
 
+/// Returns a signature of the builder as defined by a dag-typed initializer.
+/// Updates the context `fctx` to enable replacement of $_builder and $_state
+/// in the body. Reports errors at `loc`.
+static std::string builderSignatureFromDAG(const DagInit *init,
+                                           ArrayRef<llvm::SMLoc> loc,
+                                           FmtContext &fctx) {
+  auto *defInit = dyn_cast<DefInit>(init->getOperator());
+  if (!defInit || !defInit->getDef()->getName().equals("ins"))
+    PrintFatalError(loc, "expected 'ins' in builders");
+
+  // Inject builder and state arguments.
+  llvm::SmallVector<std::string, 8> arguments;
+  arguments.reserve(init->getNumArgs() + 2);
+  arguments.push_back(llvm::formatv("::mlir::OpBuilder &{0}", builder).str());
+  arguments.push_back(
+      llvm::formatv("::mlir::OperationState &{0}", builderOpState).str());
+
+  // Accept either a StringInit or a DefInit with two string values as dag
+  // arguments. The former corresponds to the type, the latter to the type and
+  // the default value. Similarly to C++, once an argument with a default value
+  // is detected, the following arguments must have default values as well.
+  bool seenDefaultValue = false;
+  for (unsigned i = 0, e = init->getNumArgs(); i < e; ++i) {
+    // If no name is provided, generate one.
+    StringInit *argName = init->getArgName(i);
+    std::string name =
+        argName ? argName->getValue().str() : "odsArg" + std::to_string(i);
+
+    Init *argInit = init->getArg(i);
+    StringRef type;
+    std::string defaultValue;
+    if (StringInit *strType = dyn_cast<StringInit>(argInit)) {
+      type = strType->getValue();
+    } else {
+      const Record *typeAndDefaultValue = cast<DefInit>(argInit)->getDef();
+      type = typeAndDefaultValue->getValueAsString("type");
+      StringRef defaultValueRef =
+          typeAndDefaultValue->getValueAsString("defaultValue");
+      if (!defaultValueRef.empty()) {
+        seenDefaultValue = true;
+        defaultValue = llvm::formatv(" = {0}", defaultValueRef).str();
+      }
+    }
+    if (seenDefaultValue && defaultValue.empty())
+      PrintFatalError(loc,
+                      "expected an argument with default value after other "
+                      "arguments with default values");
+    arguments.push_back(
+        llvm::formatv("{0} {1}{2}", type, name, defaultValue).str());
+  }
+
+  fctx.withBuilder(builder);
+  fctx.addSubst("_state", builderOpState);
+
+  return llvm::join(arguments, ", ");
+}
+
+// Returns a signature fo the builder as defined by a string initializer,
+// optionally injecting the builder and state arguments.
+// TODO: to be removed after the transition is complete.
+static std::string builderSignatureFromString(StringRef params,
+                                              FmtContext &fctx) {
+  bool skipParamGen = params.startswith("OpBuilder") ||
+                      params.startswith("mlir::OpBuilder") ||
+                      params.startswith("::mlir::OpBuilder");
+  if (skipParamGen)
+    return params.str();
+
+  fctx.withBuilder(builder);
+  fctx.addSubst("_state", builderOpState);
+  return std::string(llvm::formatv("::mlir::OpBuilder &{0}, "
+                                   "::mlir::OperationState &{1}{2}{3}",
+                                   builder, builderOpState,
+                                   params.empty() ? "" : ", ", params));
+}
+
 void OpEmitter::genBuilder() {
   // Handle custom builders if provided.
   // TODO: Create wrapper class for OpBuilder to hide the native
@@ -1153,35 +1229,23 @@ void OpEmitter::genBuilder() {
     if (listInit) {
       for (Init *init : listInit->getValues()) {
         Record *builderDef = cast<DefInit>(init)->getDef();
-        StringRef params = builderDef->getValueAsString("params").trim();
-        // TODO: Remove this and just generate the builder/state always.
-        bool skipParamGen = params.startswith("OpBuilder") ||
-                            params.startswith("mlir::OpBuilder") ||
-                            params.startswith("::mlir::OpBuilder");
+        llvm::Optional<StringRef> params =
+            builderDef->getValueAsOptionalString("params");
+        FmtContext fctx;
+        std::string paramStr =
+            params.hasValue() ? builderSignatureFromString(params->trim(), fctx)
+                              : builderSignatureFromDAG(
+                                    builderDef->getValueAsDag("dagParams"),
+                                    op.getLoc(), fctx);
+
         StringRef body = builderDef->getValueAsString("body");
         bool hasBody = !body.empty();
-
         OpMethod::Property properties =
             hasBody ? OpMethod::MP_Static : OpMethod::MP_StaticDeclaration;
-        std::string paramStr =
-            skipParamGen ? params.str()
-                         : llvm::formatv("::mlir::OpBuilder &{0}, "
-                                         "::mlir::OperationState &{1}{2}{3}",
-                                         builder, builderOpState,
-                                         params.empty() ? "" : ", ", params)
-                               .str();
         auto *method =
             opClass.addMethodAndPrune("void", "build", properties, paramStr);
-        if (hasBody) {
-          if (skipParamGen) {
-            method->body() << body;
-          } else {
-            FmtContext fctx;
-            fctx.withBuilder(builder);
-            fctx.addSubst("_state", builderOpState);
-            method->body() << tgfmt(body, &fctx);
-          }
-        }
+        if (hasBody)
+          method->body() << tgfmt(body, &fctx);
       }
     }
     if (op.skipDefaultBuilders()) {
diff --git a/openmp/CMakeLists.txt b/openmp/CMakeLists.txt
index 51f496eff1bff..6cc36d9b75773 100644
--- a/openmp/CMakeLists.txt
+++ b/openmp/CMakeLists.txt
@@ -66,6 +66,20 @@ if (APPLE OR WIN32 OR NOT OPENMP_HAVE_STD_CPP14_FLAG)
   set(ENABLE_LIBOMPTARGET OFF)
 endif()
 
+# Attempt to locate LLVM source, required by libomptarget
+if (NOT LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR)
+  if (LLVM_MAIN_INCLUDE_DIR)
+    set(LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR ${LLVM_MAIN_INCLUDE_DIR})
+  elseif (EXISTS ${CMAKE_CURRENT_SOURCE_DIR}/../llvm/include)
+    set(LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR ${CMAKE_CURRENT_SOURCE_DIR}/../llvm/include)
+  endif()
+endif()
+
+if (NOT LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR)
+  message(STATUS "Missing definition for LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR, disabling libomptarget")
+  set(ENABLE_LIBOMPTARGET OFF)
+endif()
+
 option(OPENMP_ENABLE_LIBOMPTARGET "Enable building libomptarget for offloading."
        ${ENABLE_LIBOMPTARGET})
 if (OPENMP_ENABLE_LIBOMPTARGET)
diff --git a/openmp/libomptarget/CMakeLists.txt b/openmp/libomptarget/CMakeLists.txt
index c1bc29faaf45d..7ef0bafdf3c67 100644
--- a/openmp/libomptarget/CMakeLists.txt
+++ b/openmp/libomptarget/CMakeLists.txt
@@ -29,6 +29,11 @@ include(LibomptargetUtils)
 # Get dependencies for the different components of the project.
 include(LibomptargetGetDependencies)
 
+# LLVM source tree is required at build time for libomptarget
+if (NOT LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR)
+  message(FATAL_ERROR "Missing definition for LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR")
+endif()
+
 # This is a list of all the targets that are supported/tested right now.
 set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} aarch64-unknown-linux-gnu")
 set (LIBOMPTARGET_ALL_TARGETS "${LIBOMPTARGET_ALL_TARGETS} powerpc64le-ibm-linux-gnu")
diff --git a/openmp/libomptarget/plugins/amdgpu/CMakeLists.txt b/openmp/libomptarget/plugins/amdgpu/CMakeLists.txt
index 3882b777f5b1e..0c50ffdf2fa6e 100644
--- a/openmp/libomptarget/plugins/amdgpu/CMakeLists.txt
+++ b/openmp/libomptarget/plugins/amdgpu/CMakeLists.txt
@@ -30,8 +30,8 @@ if(NOT CMAKE_SYSTEM_PROCESSOR MATCHES "(x86_64)|(ppc64le)|(aarch64)$" AND CMAKE_
   return()
 endif()
 
-if (NOT LLVM_MAIN_INCLUDE_DIR)
-  libomptarget_say("Not building AMDGPU plugin: Missing definition for LLVM_MAIN_INCLUDE_DIR")
+if (NOT LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR)
+  libomptarget_say("Not building AMDGPU plugin: Missing definition for LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR")
   return()
 endif()
 
@@ -50,7 +50,7 @@ endif()
 
 include_directories(
   ${CMAKE_CURRENT_SOURCE_DIR}/impl
-  ${LLVM_MAIN_INCLUDE_DIR}
+  ${LIBOMPTARGET_LLVM_MAIN_INCLUDE_DIR}
 )
 
 add_library(omptarget.rtl.amdgpu SHARED
diff --git a/openmp/libomptarget/plugins/amdgpu/impl/atmi.cpp b/openmp/libomptarget/plugins/amdgpu/impl/atmi.cpp
index 0586cd358c25c..285dc2dbe7639 100644
--- a/openmp/libomptarget/plugins/amdgpu/impl/atmi.cpp
+++ b/openmp/libomptarget/plugins/amdgpu/impl/atmi.cpp
@@ -3,7 +3,13 @@
  *
  * This file is distributed under the MIT License. See LICENSE.txt for details.
  *===------------------------------------------------------------------------*/
+#include "atmi_runtime.h"
+#include "internal.h"
 #include "rt.h"
+#include <hsa.h>
+#include <hsa_ext_amd.h>
+#include <memory>
+
 /*
  * Initialize/Finalize
  */
@@ -33,9 +39,44 @@ atmi_status_t atmi_module_register_from_memory_to_place(
 /*
  * Data
  */
-atmi_status_t atmi_memcpy(hsa_signal_t sig, void *dest, const void *src,
-                          size_t size) {
-  hsa_status_t rc = hsa_memory_copy(dest, src, size);
+
+static hsa_status_t invoke_hsa_copy(hsa_signal_t sig, void *dest,
+                                    const void *src, size_t size,
+                                    hsa_agent_t agent) {
+  const hsa_signal_value_t init = 1;
+  const hsa_signal_value_t success = 0;
+  hsa_signal_store_screlease(sig, init);
+
+  hsa_status_t err =
+      hsa_amd_memory_async_copy(dest, agent, src, agent, size, 0, NULL, sig);
+  if (err != HSA_STATUS_SUCCESS) {
+    return err;
+  }
+
+  // async_copy reports success by decrementing and failure by setting to < 0
+  hsa_signal_value_t got = init;
+  while (got == init) {
+    got = hsa_signal_wait_scacquire(sig, HSA_SIGNAL_CONDITION_NE, init,
+                                    UINT64_MAX, ATMI_WAIT_STATE);
+  }
+
+  if (got != success) {
+    return HSA_STATUS_ERROR;
+  }
+
+  return err;
+}
+
+struct atmiFreePtrDeletor {
+  void operator()(void *p) {
+    atmi_free(p); // ignore failure to free
+  }
+};
+
+atmi_status_t atmi_memcpy_h2d(hsa_signal_t signal, void *deviceDest,
+                              const void *hostSrc, size_t size,
+                              hsa_agent_t agent) {
+  hsa_status_t rc = hsa_memory_copy(deviceDest, hostSrc, size);
 
   // hsa_memory_copy sometimes fails in situations where
   // allocate + copy succeeds. Looks like it might be related to
@@ -44,17 +85,53 @@ atmi_status_t atmi_memcpy(hsa_signal_t sig, void *dest, const void *src,
     return ATMI_STATUS_SUCCESS;
   }
 
-  return core::Runtime::Memcpy(sig, dest, src, size);
-}
+  void *tempHostPtr;
+  atmi_mem_place_t CPU = ATMI_MEM_PLACE_CPU_MEM(0, 0, 0);
+  atmi_status_t ret = atmi_malloc(&tempHostPtr, size, CPU);
+  if (ret != ATMI_STATUS_SUCCESS) {
+    DEBUG_PRINT("atmi_malloc: Unable to alloc %d bytes for temp scratch\n",
+                size);
+    return ret;
+  }
+  std::unique_ptr<void, atmiFreePtrDeletor> del(tempHostPtr);
+  memcpy(tempHostPtr, hostSrc, size);
 
-atmi_status_t atmi_memcpy_h2d(hsa_signal_t sig, void *device_dest,
-                              const void *host_src, size_t size) {
-  return atmi_memcpy(sig, device_dest, host_src, size);
+  if (invoke_hsa_copy(signal, deviceDest, tempHostPtr, size, agent) !=
+      HSA_STATUS_SUCCESS) {
+    return ATMI_STATUS_ERROR;
+  }
+  return ATMI_STATUS_SUCCESS;
 }
 
-atmi_status_t atmi_memcpy_d2h(hsa_signal_t sig, void *host_dest,
-                              const void *device_src, size_t size) {
-  return atmi_memcpy(sig, host_dest, device_src, size);
+atmi_status_t atmi_memcpy_d2h(hsa_signal_t signal, void *dest,
+                              const void *deviceSrc, size_t size,
+                              hsa_agent_t agent) {
+  hsa_status_t rc = hsa_memory_copy(dest, deviceSrc, size);
+
+  // hsa_memory_copy sometimes fails in situations where
+  // allocate + copy succeeds. Looks like it might be related to
+  // locking part of a read only segment. Fall back for now.
+  if (rc == HSA_STATUS_SUCCESS) {
+    return ATMI_STATUS_SUCCESS;
+  }
+
+  void *tempHostPtr;
+  atmi_mem_place_t CPU = ATMI_MEM_PLACE_CPU_MEM(0, 0, 0);
+  atmi_status_t ret = atmi_malloc(&tempHostPtr, size, CPU);
+  if (ret != ATMI_STATUS_SUCCESS) {
+    DEBUG_PRINT("atmi_malloc: Unable to alloc %d bytes for temp scratch\n",
+                size);
+    return ret;
+  }
+  std::unique_ptr<void, atmiFreePtrDeletor> del(tempHostPtr);
+
+  if (invoke_hsa_copy(signal, tempHostPtr, deviceSrc, size, agent) !=
+      HSA_STATUS_SUCCESS) {
+    return ATMI_STATUS_ERROR;
+  }
+
+  memcpy(dest, tempHostPtr, size);
+  return ATMI_STATUS_SUCCESS;
 }
 
 atmi_status_t atmi_free(void *ptr) { return core::Runtime::Memfree(ptr); }
diff --git a/openmp/libomptarget/plugins/amdgpu/impl/atmi_runtime.h b/openmp/libomptarget/plugins/amdgpu/impl/atmi_runtime.h
index a935b6ad4b759..47022f7f5dea3 100644
--- a/openmp/libomptarget/plugins/amdgpu/impl/atmi_runtime.h
+++ b/openmp/libomptarget/plugins/amdgpu/impl/atmi_runtime.h
@@ -155,11 +155,13 @@ atmi_status_t atmi_malloc(void **ptr, size_t size, atmi_mem_place_t place);
  */
 atmi_status_t atmi_free(void *ptr);
 
-atmi_status_t atmi_memcpy_h2d(hsa_signal_t sig, void *device_dest,
-                              const void *host_src, size_t size);
+atmi_status_t atmi_memcpy_h2d(hsa_signal_t signal, void *deviceDest,
+                              const void *hostSrc, size_t size,
+                              hsa_agent_t agent);
 
-atmi_status_t atmi_memcpy_d2h(hsa_signal_t sig, void *host_dest,
-                              const void *device_src, size_t size);
+atmi_status_t atmi_memcpy_d2h(hsa_signal_t sig, void *hostDest,
+                              const void *deviceSrc, size_t size,
+                              hsa_agent_t agent);
 
 /** @} */
 
diff --git a/openmp/libomptarget/plugins/amdgpu/impl/data.cpp b/openmp/libomptarget/plugins/amdgpu/impl/data.cpp
index 91627bd8d4b31..39546fbae4b3b 100644
--- a/openmp/libomptarget/plugins/amdgpu/impl/data.cpp
+++ b/openmp/libomptarget/plugins/amdgpu/impl/data.cpp
@@ -3,7 +3,6 @@
  *
  * This file is distributed under the MIT License. See LICENSE.txt for details.
  *===------------------------------------------------------------------------*/
-#include "data.h"
 #include "atmi_runtime.h"
 #include "internal.h"
 #include "machine.h"
@@ -21,7 +20,6 @@ using core::TaskImpl;
 extern ATLMachine g_atl_machine;
 
 namespace core {
-ATLPointerTracker g_data_map; // Track all am pointer allocations.
 void allow_access_to_all_gpu_agents(void *ptr);
 
 const char *getPlaceStr(atmi_devtype_t type) {
@@ -35,37 +33,6 @@ const char *getPlaceStr(atmi_devtype_t type) {
   }
 }
 
-std::ostream &operator<<(std::ostream &os, const ATLData *ap) {
-  atmi_mem_place_t place = ap->place();
-  os << " devicePointer:" << ap->ptr() << " sizeBytes:" << ap->size()
-     << " place:(" << getPlaceStr(place.dev_type) << ", " << place.dev_id
-     << ", " << place.mem_id << ")";
-  return os;
-}
-
-void ATLPointerTracker::insert(void *pointer, ATLData *p) {
-  std::lock_guard<std::mutex> l(mutex_);
-
-  DEBUG_PRINT("insert: %p + %zu\n", pointer, p->size());
-  tracker_.insert(std::make_pair(ATLMemoryRange(pointer, p->size()), p));
-}
-
-void ATLPointerTracker::remove(void *pointer) {
-  std::lock_guard<std::mutex> l(mutex_);
-  DEBUG_PRINT("remove: %p\n", pointer);
-  tracker_.erase(ATLMemoryRange(pointer, 1));
-}
-
-ATLData *ATLPointerTracker::find(const void *pointer) {
-  std::lock_guard<std::mutex> l(mutex_);
-  ATLData *ret = NULL;
-  auto iter = tracker_.find(ATLMemoryRange(pointer, 1));
-  DEBUG_PRINT("find: %p\n", pointer);
-  if (iter != tracker_.end()) // found
-    ret = iter->second;
-  return ret;
-}
-
 ATLProcessor &get_processor_by_mem_place(atmi_mem_place_t place) {
   int dev_id = place.dev_id;
   switch (place.dev_type) {
@@ -76,18 +43,12 @@ ATLProcessor &get_processor_by_mem_place(atmi_mem_place_t place) {
   }
 }
 
-static hsa_agent_t get_mem_agent(atmi_mem_place_t place) {
-  return get_processor_by_mem_place(place).agent();
-}
-
 hsa_amd_memory_pool_t get_memory_pool_by_mem_place(atmi_mem_place_t place) {
   ATLProcessor &proc = get_processor_by_mem_place(place);
   return get_memory_pool(proc, place.mem_id);
 }
 
 void register_allocation(void *ptr, size_t size, atmi_mem_place_t place) {
-  ATLData *data = new ATLData(ptr, size, place);
-  g_data_map.insert(ptr, data);
   if (place.dev_type == ATMI_DEVTYPE_CPU)
     allow_access_to_all_gpu_agents(ptr);
   // TODO(ashwinma): what if one GPU wants to access another GPU?
@@ -112,103 +73,13 @@ atmi_status_t Runtime::Malloc(void **ptr, size_t size, atmi_mem_place_t place) {
 atmi_status_t Runtime::Memfree(void *ptr) {
   atmi_status_t ret = ATMI_STATUS_SUCCESS;
   hsa_status_t err;
-  ATLData *data = g_data_map.find(ptr);
-  if (!data)
-    ErrorCheck(Checking pointer info userData,
-               HSA_STATUS_ERROR_INVALID_ALLOCATION);
-
-  g_data_map.remove(ptr);
-  delete data;
-
   err = hsa_amd_memory_pool_free(ptr);
   ErrorCheck(atmi_free, err);
   DEBUG_PRINT("Freed %p\n", ptr);
 
-  if (err != HSA_STATUS_SUCCESS || !data)
+  if (err != HSA_STATUS_SUCCESS)
     ret = ATMI_STATUS_ERROR;
   return ret;
 }
 
-static hsa_status_t invoke_hsa_copy(hsa_signal_t sig, void *dest,
-                                    const void *src, size_t size,
-                                    hsa_agent_t agent) {
-  const hsa_signal_value_t init = 1;
-  const hsa_signal_value_t success = 0;
-  hsa_signal_store_screlease(sig, init);
-
-  hsa_status_t err =
-      hsa_amd_memory_async_copy(dest, agent, src, agent, size, 0, NULL, sig);
-  if (err != HSA_STATUS_SUCCESS) {
-    return err;
-  }
-
-  // async_copy reports success by decrementing and failure by setting to < 0
-  hsa_signal_value_t got = init;
-  while (got == init) {
-    got = hsa_signal_wait_scacquire(sig, HSA_SIGNAL_CONDITION_NE, init,
-                                    UINT64_MAX, ATMI_WAIT_STATE);
-  }
-
-  if (got != success) {
-    return HSA_STATUS_ERROR;
-  }
-
-  return err;
-}
-
-struct atmiFreePtrDeletor {
-  void operator()(void *p) {
-    atmi_free(p); // ignore failure to free
-  }
-};
-
-atmi_status_t Runtime::Memcpy(hsa_signal_t sig, void *dest, const void *src,
-                              size_t size) {
-  ATLData *src_data = g_data_map.find(src);
-  ATLData *dest_data = g_data_map.find(dest);
-  atmi_mem_place_t cpu = ATMI_MEM_PLACE_CPU_MEM(0, 0, 0);
-
-  void *temp_host_ptr;
-  atmi_status_t ret = atmi_malloc(&temp_host_ptr, size, cpu);
-  if (ret != ATMI_STATUS_SUCCESS) {
-    return ret;
-  }
-  std::unique_ptr<void, atmiFreePtrDeletor> del(temp_host_ptr);
-
-  if (src_data && !dest_data) {
-    // Copy from device to scratch to host
-    hsa_agent_t agent = get_mem_agent(src_data->place());
-    DEBUG_PRINT("Memcpy D2H device agent: %lu\n", agent.handle);
-
-    if (invoke_hsa_copy(sig, temp_host_ptr, src, size, agent) !=
-        HSA_STATUS_SUCCESS) {
-      return ATMI_STATUS_ERROR;
-    }
-
-    memcpy(dest, temp_host_ptr, size);
-
-  } else if (!src_data && dest_data) {
-    // Copy from host to scratch to device
-    hsa_agent_t agent = get_mem_agent(dest_data->place());
-    DEBUG_PRINT("Memcpy H2D device agent: %lu\n", agent.handle);
-
-    memcpy(temp_host_ptr, src, size);
-
-    if (invoke_hsa_copy(sig, dest, temp_host_ptr, size, agent) !=
-        HSA_STATUS_SUCCESS) {
-      return ATMI_STATUS_ERROR;
-    }
-
-  } else if (!src_data && !dest_data) {
-    // would be host to host, just call memcpy, or missing metadata
-    DEBUG_PRINT("atmi_memcpy invoked without metadata\n");
-    return ATMI_STATUS_ERROR;
-  } else {
-    DEBUG_PRINT("atmi_memcpy unimplemented device to device copy\n");
-    return ATMI_STATUS_ERROR;
-  }
-
-  return ATMI_STATUS_SUCCESS;
-}
-
 } // namespace core
diff --git a/openmp/libomptarget/plugins/amdgpu/impl/data.h b/openmp/libomptarget/plugins/amdgpu/impl/data.h
deleted file mode 100644
index fa9e7380bf673..0000000000000
--- a/openmp/libomptarget/plugins/amdgpu/impl/data.h
+++ /dev/null
@@ -1,83 +0,0 @@
-/*===--------------------------------------------------------------------------
- *              ATMI (Asynchronous Task and Memory Interface)
- *
- * This file is distributed under the MIT License. See LICENSE.txt for details.
- *===------------------------------------------------------------------------*/
-#ifndef SRC_RUNTIME_INCLUDE_DATA_H_
-#define SRC_RUNTIME_INCLUDE_DATA_H_
-#include "atmi.h"
-#include <hsa.h>
-#include <map>
-#include <mutex>
-#include <stdio.h>
-#include <stdlib.h>
-// we maintain our own mapping of device addr to a user specified data object
-// in order to work around a (possibly historic) bug in ROCr's
-// hsa_amd_pointer_info_set_userdata for variable symbols
-// this is expected to be temporary
-
-namespace core {
-// Internal representation of any data that is created and managed by ATMI.
-// Data can be located on any device memory or host memory.
-class ATLData {
-public:
-  ATLData(void *ptr, size_t size, atmi_mem_place_t place)
-      : ptr_(ptr), size_(size), place_(place) {}
-
-  void *ptr() const { return ptr_; }
-  size_t size() const { return size_; }
-  atmi_mem_place_t place() const { return place_; }
-
-private:
-  void *ptr_;
-  size_t size_;
-  atmi_mem_place_t place_;
-};
-
-//---
-struct ATLMemoryRange {
-  const void *base_pointer;
-  const void *end_pointer;
-  ATLMemoryRange(const void *bp, size_t size_bytes)
-      : base_pointer(bp),
-        end_pointer(reinterpret_cast<const unsigned char *>(bp) + size_bytes -
-                    1) {}
-};
-
-// Functor to compare ranges:
-struct ATLMemoryRangeCompare {
-  // Return true is LHS range is less than RHS - used to order the ranges
-  bool operator()(const ATLMemoryRange &lhs, const ATLMemoryRange &rhs) const {
-    return lhs.end_pointer < rhs.base_pointer;
-  }
-};
-
-//-------------------------------------------------------------------------------------------------
-// This structure tracks information for each pointer.
-// Uses memory-range-based lookups - so pointers that exist anywhere in the
-// range of hostPtr + size
-// will find the associated ATLPointerInfo.
-// The insertions and lookups use a self-balancing binary tree and should
-// support O(logN) lookup speed.
-// The structure is thread-safe - writers obtain a mutex before modifying the
-// tree.  Multiple simulatenous readers are supported.
-class ATLPointerTracker {
-  typedef std::map<ATLMemoryRange, ATLData *, ATLMemoryRangeCompare>
-      MapTrackerType;
-
-public:
-  void insert(void *pointer, ATLData *data);
-  void remove(void *pointer);
-  ATLData *find(const void *pointer);
-
-private:
-  MapTrackerType tracker_;
-  std::mutex mutex_;
-};
-
-extern ATLPointerTracker g_data_map; // Track all am pointer allocations.
-
-enum class Direction { ATMI_H2D, ATMI_D2H, ATMI_D2D, ATMI_H2H };
-
-} // namespace core
-#endif // SRC_RUNTIME_INCLUDE_DATA_H_
diff --git a/openmp/libomptarget/plugins/amdgpu/src/rtl.cpp b/openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
index 8fd7b7e34155f..dc3a288903f03 100644
--- a/openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
+++ b/openmp/libomptarget/plugins/amdgpu/src/rtl.cpp
@@ -335,27 +335,28 @@ class RTLDeviceInfoTy {
   static const int Default_WG_Size =
       llvm::omp::AMDGPUGpuGridValues[llvm::omp::GVIDX::GV_Default_WG_Size];
 
-  using MemcpyFunc = atmi_status_t(hsa_signal_t, void *, const void *,
-                                   size_t size);
+  using MemcpyFunc = atmi_status_t (*)(hsa_signal_t, void *, const void *,
+                                       size_t size, hsa_agent_t);
   atmi_status_t freesignalpool_memcpy(void *dest, const void *src, size_t size,
-                                      MemcpyFunc Func) {
+                                      MemcpyFunc Func, int32_t deviceId) {
+    hsa_agent_t agent = HSAAgents[deviceId];
     hsa_signal_t s = FreeSignalPool.pop();
     if (s.handle == 0) {
       return ATMI_STATUS_ERROR;
     }
-    atmi_status_t r = Func(s, dest, src, size);
+    atmi_status_t r = Func(s, dest, src, size, agent);
     FreeSignalPool.push(s);
     return r;
   }
 
   atmi_status_t freesignalpool_memcpy_d2h(void *dest, const void *src,
-                                          size_t size) {
-    return freesignalpool_memcpy(dest, src, size, atmi_memcpy_d2h);
+                                          size_t size, int32_t deviceId) {
+    return freesignalpool_memcpy(dest, src, size, atmi_memcpy_d2h, deviceId);
   }
 
   atmi_status_t freesignalpool_memcpy_h2d(void *dest, const void *src,
-                                          size_t size) {
-    return freesignalpool_memcpy(dest, src, size, atmi_memcpy_h2d);
+                                          size_t size, int32_t deviceId) {
+    return freesignalpool_memcpy(dest, src, size, atmi_memcpy_h2d, deviceId);
   }
 
   // Record entry point associated with device
@@ -550,7 +551,8 @@ int32_t dataRetrieve(int32_t DeviceId, void *HstPtr, void *TgtPtr, int64_t Size,
      (long long unsigned)(Elf64_Addr)TgtPtr,
      (long long unsigned)(Elf64_Addr)HstPtr);
 
-  err = DeviceInfo.freesignalpool_memcpy_d2h(HstPtr, TgtPtr, (size_t)Size);
+  err = DeviceInfo.freesignalpool_memcpy_d2h(HstPtr, TgtPtr, (size_t)Size,
+                                             DeviceId);
 
   if (err != ATMI_STATUS_SUCCESS) {
     DP("Error when copying data from device to host. Pointers: "
@@ -576,7 +578,8 @@ int32_t dataSubmit(int32_t DeviceId, void *TgtPtr, void *HstPtr, int64_t Size,
   DP("Submit data %ld bytes, (hst:%016llx) -> (tgt:%016llx).\n", Size,
      (long long unsigned)(Elf64_Addr)HstPtr,
      (long long unsigned)(Elf64_Addr)TgtPtr);
-  err = DeviceInfo.freesignalpool_memcpy_h2d(TgtPtr, HstPtr, (size_t)Size);
+  err = DeviceInfo.freesignalpool_memcpy_h2d(TgtPtr, HstPtr, (size_t)Size,
+                                             DeviceId);
   if (err != ATMI_STATUS_SUCCESS) {
     DP("Error when copying data from host to device. Pointers: "
        "host = 0x%016lx, device = 0x%016lx, size = %lld\n",
@@ -1033,7 +1036,8 @@ __tgt_target_table *__tgt_rtl_load_binary_locked(int32_t device_id,
     }
 
     // write ptr to device memory so it can be used by later kernels
-    err = DeviceInfo.freesignalpool_memcpy_h2d(state_ptr, &ptr, sizeof(void *));
+    err = DeviceInfo.freesignalpool_memcpy_h2d(state_ptr, &ptr, sizeof(void *),
+                                               device_id);
     if (err != ATMI_STATUS_SUCCESS) {
       fprintf(stderr, "memcpy install of state_ptr failed\n");
       return NULL;
@@ -1103,7 +1107,7 @@ __tgt_target_table *__tgt_rtl_load_binary_locked(int32_t device_id,
         // can access host addresses directly. There is no longer a
         // need for device copies.
         err = DeviceInfo.freesignalpool_memcpy_h2d(varptr, e->addr,
-                                                   sizeof(void *));
+                                                   sizeof(void *), device_id);
         if (err != ATMI_STATUS_SUCCESS)
           DP("Error when copying USM\n");
         DP("Copy linked variable host address (" DPxMOD ")"
@@ -1532,7 +1536,7 @@ static void *AllocateNestedParallelCallMemory(int MaxParLevel, int NumGroups,
   atmi_status_t err =
       atmi_malloc(&TgtPtr, NestedMemSize, get_gpu_mem_place(device_id));
   err = DeviceInfo.freesignalpool_memcpy_h2d(CallStackAddr, &TgtPtr,
-                                             sizeof(void *));
+                                             sizeof(void *), device_id);
   if (print_kernel_trace > 2)
     fprintf(stderr, "CallSck %lx TgtPtr %lx *TgtPtr %lx \n",
             (long)CallStackAddr, (long)&TgtPtr, (long)TgtPtr);
diff --git a/polly/CMakeLists.txt b/polly/CMakeLists.txt
index fe7f6b78b4792..ca7c04c565bb6 100644
--- a/polly/CMakeLists.txt
+++ b/polly/CMakeLists.txt
@@ -125,7 +125,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     DESTINATION include
     FILES_MATCHING
     PATTERN "*.h"
-    PATTERN ".svn" EXCLUDE
     )
 
   install(DIRECTORY ${POLLY_BINARY_DIR}/include/
@@ -133,7 +132,6 @@ if (NOT LLVM_INSTALL_TOOLCHAIN_ONLY)
     FILES_MATCHING
     PATTERN "*.h"
     PATTERN "CMakeFiles" EXCLUDE
-    PATTERN ".svn" EXCLUDE
     )
 endif()
 
diff --git a/polly/lib/External/CMakeLists.txt b/polly/lib/External/CMakeLists.txt
index c953ea48475d7..8991094d92c7b 100644
--- a/polly/lib/External/CMakeLists.txt
+++ b/polly/lib/External/CMakeLists.txt
@@ -279,7 +279,6 @@ if (POLLY_BUNDLED_ISL)
       FILES_MATCHING
       PATTERN "*.h"
       PATTERN "CMakeFiles" EXCLUDE
-      PATTERN ".svn" EXCLUDE
       )
   endif()
 
diff --git a/polly/lib/Transform/ForwardOpTree.cpp b/polly/lib/Transform/ForwardOpTree.cpp
index 56c5fefeab39a..141ac26b714d8 100644
--- a/polly/lib/Transform/ForwardOpTree.cpp
+++ b/polly/lib/Transform/ForwardOpTree.cpp
@@ -696,7 +696,7 @@ class ForwardOpTreeImpl : ZoneAlgorithm {
       }
     }
 
-    auto ExecAction = [this, TargetStmt, UseInst, DefStmt]() {
+    auto ExecAction = [this, TargetStmt, UseInst]() {
       // To ensure the right order, prepend this instruction before its
       // operands. This ensures that its operands are inserted before the
       // instruction using them.