[DAG] Generalize setcc(setcc) fold to use known bits.

If we have a `SETCC (SETCC), 0, NE` and ZeroOrOneBooleanContent, we can remove the outer setcc as it will produce the same value as the inner. This can be generalized to anything where the top bits are known to be 0, as the value will remain as 1 or 0.
pradt2 · Feb 6, 2024 · 2e3de99 · 2e3de99
1 parent 29fa64f
commit 2e3de99
Show file tree

Hide file tree

Showing 6 changed files with 52 additions and 48 deletions.
diff --git a/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp b/llvm/lib/CodeGen/SelectionDAG/TargetLowering.cpp
@@ -4718,21 +4718,25 @@ SDValue TargetLowering::SimplifySetCC(EVT VT, SDValue N0, SDValue N1,
                           DAG.getConstant(C1 & Imm, dl, ExtDstTy), Cond);
     } else if ((N1C->isZero() || N1C->isOne()) &&
                (Cond == ISD::SETEQ || Cond == ISD::SETNE)) {
-      // SETCC (SETCC), [0|1], [EQ|NE]  -> SETCC
-      if (N0.getOpcode() == ISD::SETCC &&
+      // SETCC (X), [0|1], [EQ|NE]  -> X if X is known 0/1. i1 types are
+      // excluded as they are handled below whilst checking for foldBooleans.
+      if ((N0.getOpcode() == ISD::SETCC || VT.getScalarType() != MVT::i1) &&
           isTypeLegal(VT) && VT.bitsLE(N0.getValueType()) &&
           (N0.getValueType() == MVT::i1 ||
-           getBooleanContents(N0.getOperand(0).getValueType()) ==
-                       ZeroOrOneBooleanContent)) {
+           getBooleanContents(N0.getValueType()) == ZeroOrOneBooleanContent) &&
+          DAG.MaskedValueIsZero(
+              N0, APInt::getBitsSetFrom(N0.getValueSizeInBits(), 1))) {
         bool TrueWhenTrue = (Cond == ISD::SETEQ) ^ (!N1C->isOne());
         if (TrueWhenTrue)
           return DAG.getNode(ISD::TRUNCATE, dl, VT, N0);
         // Invert the condition.
-        ISD::CondCode CC = cast<CondCodeSDNode>(N0.getOperand(2))->get();
-        CC = ISD::getSetCCInverse(CC, N0.getOperand(0).getValueType());
-        if (DCI.isBeforeLegalizeOps() ||
-            isCondCodeLegal(CC, N0.getOperand(0).getSimpleValueType()))
-          return DAG.getSetCC(dl, VT, N0.getOperand(0), N0.getOperand(1), CC);
+        if (N0.getOpcode() == ISD::SETCC) {
+          ISD::CondCode CC = cast<CondCodeSDNode>(N0.getOperand(2))->get();
+          CC = ISD::getSetCCInverse(CC, N0.getOperand(0).getValueType());
+          if (DCI.isBeforeLegalizeOps() ||
+              isCondCodeLegal(CC, N0.getOperand(0).getSimpleValueType()))
+            return DAG.getSetCC(dl, VT, N0.getOperand(0), N0.getOperand(1), CC);
+        }
       }
 
       if ((N0.getOpcode() == ISD::XOR ||

diff --git a/llvm/test/CodeGen/AArch64/setcc_knownbits.ll b/llvm/test/CodeGen/AArch64/setcc_knownbits.ll
@@ -49,9 +49,8 @@ define i1 @lshr_ctlz_undef_cmpeq_one_i64(i64 %in) {
 ; CHECK-LABEL: lshr_ctlz_undef_cmpeq_one_i64:
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    clz x8, x0
-; CHECK-NEXT:    lsr x8, x8, #6
-; CHECK-NEXT:    cmp x8, #1
-; CHECK-NEXT:    cset w0, eq
+; CHECK-NEXT:    lsr x0, x8, #6
+; CHECK-NEXT:    // kill: def $w0 killed $w0 killed $x0
 ; CHECK-NEXT:    ret
   %ctlz = call i64 @llvm.ctlz.i64(i64 %in, i1 -1)
   %lshr = lshr i64 %ctlz, 6

diff --git a/llvm/test/CodeGen/WebAssembly/xor_reassociate.ll b/llvm/test/CodeGen/WebAssembly/xor_reassociate.ll
@@ -17,7 +17,7 @@ define i32 @reassociate_xor(float %x, float %y) {
 ; CHECK-NEXT:    local.get 0
 ; CHECK-NEXT:    f32.const 0x1p-23
 ; CHECK-NEXT:    f32.gt
-; CHECK-NEXT:    i32.ne
+; CHECK-NEXT:    i32.xor
 ; CHECK-NEXT:    br_if 0 # 0: down to label0
 ; CHECK-NEXT:  # %bb.1: # %if.then.i
 ; CHECK-NEXT:    i32.const 0

diff --git a/llvm/test/CodeGen/X86/lzcnt-cmp.ll b/llvm/test/CodeGen/X86/lzcnt-cmp.ll
@@ -50,35 +50,33 @@ define i1 @lshr_ctlz_undef_cmpeq_one_i64(i64 %in) nounwind {
 ; X86-BSR-NEXT:    xorl $31, %eax
 ; X86-BSR-NEXT:    addl $32, %eax
 ; X86-BSR-NEXT:  .LBB1_2:
-; X86-BSR-NEXT:    testl $-64, %eax
-; X86-BSR-NEXT:    setne %al
+; X86-BSR-NEXT:    shrl $6, %eax
+; X86-BSR-NEXT:    # kill: def $al killed $al killed $eax
 ; X86-BSR-NEXT:    retl
 ;
 ; X86-LZCNT-LABEL: lshr_ctlz_undef_cmpeq_one_i64:
 ; X86-LZCNT:       # %bb.0:
-; X86-LZCNT-NEXT:    lzcntl {{[0-9]+}}(%esp), %eax
-; X86-LZCNT-NEXT:    addl $32, %eax
-; X86-LZCNT-NEXT:    xorl %ecx, %ecx
+; X86-LZCNT-NEXT:    lzcntl {{[0-9]+}}(%esp), %ecx
+; X86-LZCNT-NEXT:    addl $32, %ecx
+; X86-LZCNT-NEXT:    xorl %eax, %eax
 ; X86-LZCNT-NEXT:    cmpl $0, {{[0-9]+}}(%esp)
-; X86-LZCNT-NEXT:    cmovel %eax, %ecx
-; X86-LZCNT-NEXT:    testb $64, %cl
-; X86-LZCNT-NEXT:    setne %al
+; X86-LZCNT-NEXT:    cmovel %ecx, %eax
+; X86-LZCNT-NEXT:    shrl $6, %eax
+; X86-LZCNT-NEXT:    # kill: def $al killed $al killed $eax
 ; X86-LZCNT-NEXT:    retl
 ;
 ; X64-BSR-LABEL: lshr_ctlz_undef_cmpeq_one_i64:
 ; X64-BSR:       # %bb.0:
 ; X64-BSR-NEXT:    bsrq %rdi, %rax
 ; X64-BSR-NEXT:    shrl $6, %eax
-; X64-BSR-NEXT:    cmpl $1, %eax
-; X64-BSR-NEXT:    sete %al
+; X64-BSR-NEXT:    # kill: def $al killed $al killed $rax
 ; X64-BSR-NEXT:    retq
 ;
 ; X64-LZCNT-LABEL: lshr_ctlz_undef_cmpeq_one_i64:
 ; X64-LZCNT:       # %bb.0:
 ; X64-LZCNT-NEXT:    lzcntq %rdi, %rax
 ; X64-LZCNT-NEXT:    shrl $6, %eax
-; X64-LZCNT-NEXT:    cmpl $1, %eax
-; X64-LZCNT-NEXT:    sete %al
+; X64-LZCNT-NEXT:    # kill: def $al killed $al killed $rax
 ; X64-LZCNT-NEXT:    retq
   %ctlz = call i64 @llvm.ctlz.i64(i64 %in, i1 -1)
   %lshr = lshr i64 %ctlz, 6
@@ -131,33 +129,33 @@ define i1 @lshr_ctlz_undef_cmpne_zero_i64(i64 %in) nounwind {
 ; X86-BSR-NEXT:    xorl $31, %eax
 ; X86-BSR-NEXT:    addl $32, %eax
 ; X86-BSR-NEXT:  .LBB3_2:
-; X86-BSR-NEXT:    testl $-64, %eax
-; X86-BSR-NEXT:    setne %al
+; X86-BSR-NEXT:    shrl $6, %eax
+; X86-BSR-NEXT:    # kill: def $al killed $al killed $eax
 ; X86-BSR-NEXT:    retl
 ;
 ; X86-LZCNT-LABEL: lshr_ctlz_undef_cmpne_zero_i64:
 ; X86-LZCNT:       # %bb.0:
-; X86-LZCNT-NEXT:    lzcntl {{[0-9]+}}(%esp), %eax
-; X86-LZCNT-NEXT:    addl $32, %eax
-; X86-LZCNT-NEXT:    xorl %ecx, %ecx
+; X86-LZCNT-NEXT:    lzcntl {{[0-9]+}}(%esp), %ecx
+; X86-LZCNT-NEXT:    addl $32, %ecx
+; X86-LZCNT-NEXT:    xorl %eax, %eax
 ; X86-LZCNT-NEXT:    cmpl $0, {{[0-9]+}}(%esp)
-; X86-LZCNT-NEXT:    cmovel %eax, %ecx
-; X86-LZCNT-NEXT:    testb $64, %cl
-; X86-LZCNT-NEXT:    setne %al
+; X86-LZCNT-NEXT:    cmovel %ecx, %eax
+; X86-LZCNT-NEXT:    shrl $6, %eax
+; X86-LZCNT-NEXT:    # kill: def $al killed $al killed $eax
 ; X86-LZCNT-NEXT:    retl
 ;
 ; X64-BSR-LABEL: lshr_ctlz_undef_cmpne_zero_i64:
 ; X64-BSR:       # %bb.0:
 ; X64-BSR-NEXT:    bsrq %rdi, %rax
-; X64-BSR-NEXT:    testl $-64, %eax
-; X64-BSR-NEXT:    setne %al
+; X64-BSR-NEXT:    shrl $6, %eax
+; X64-BSR-NEXT:    # kill: def $al killed $al killed $rax
 ; X64-BSR-NEXT:    retq
 ;
 ; X64-LZCNT-LABEL: lshr_ctlz_undef_cmpne_zero_i64:
 ; X64-LZCNT:       # %bb.0:
 ; X64-LZCNT-NEXT:    lzcntq %rdi, %rax
-; X64-LZCNT-NEXT:    testb $64, %al
-; X64-LZCNT-NEXT:    setne %al
+; X64-LZCNT-NEXT:    shrl $6, %eax
+; X64-LZCNT-NEXT:    # kill: def $al killed $al killed $rax
 ; X64-LZCNT-NEXT:    retq
   %ctlz = call i64 @llvm.ctlz.i64(i64 %in, i1 -1)
   %lshr = lshr i64 %ctlz, 6

diff --git a/llvm/test/CodeGen/X86/umul_fix_sat.ll b/llvm/test/CodeGen/X86/umul_fix_sat.ll
@@ -517,15 +517,13 @@ define i64 @func8(i64 %x, i64 %y) nounwind {
 ; X86-NEXT:    adcl $0, %ecx
 ; X86-NEXT:    addl %ebp, %edx
 ; X86-NEXT:    adcl $0, %ecx
+; X86-NEXT:    shldl $1, %edx, %ecx
 ; X86-NEXT:    shrdl $31, %edx, %eax
-; X86-NEXT:    movl %edx, %esi
-; X86-NEXT:    shrl $31, %esi
-; X86-NEXT:    xorl %edi, %edi
-; X86-NEXT:    negl %esi
-; X86-NEXT:    sbbl %edi, %edi
-; X86-NEXT:    orl %edi, %eax
-; X86-NEXT:    shrdl $31, %ecx, %edx
-; X86-NEXT:    orl %edi, %edx
+; X86-NEXT:    testl $-2147483648, %edx # imm = 0x80000000
+; X86-NEXT:    movl $-1, %edx
+; X86-NEXT:    cmovnel %edx, %eax
+; X86-NEXT:    cmovnel %edx, %ecx
+; X86-NEXT:    movl %ecx, %edx
 ; X86-NEXT:    popl %esi
 ; X86-NEXT:    popl %edi
 ; X86-NEXT:    popl %ebx

diff --git a/llvm/test/CodeGen/X86/xor.ll b/llvm/test/CodeGen/X86/xor.ll
@@ -403,14 +403,19 @@ define i32 @PR17487(i1 %tobool) {
 ;
 ; X64-LIN-LABEL: PR17487:
 ; X64-LIN:       # %bb.0:
-; X64-LIN-NEXT:    movl %edi, %eax
-; X64-LIN-NEXT:    andl $1, %eax
+; X64-LIN-NEXT:    movd %edi, %xmm0
+; X64-LIN-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; X64-LIN-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; X64-LIN-NEXT:    pextrw $4, %xmm0, %eax
 ; X64-LIN-NEXT:    retq
 ;
 ; X64-WIN-LABEL: PR17487:
 ; X64-WIN:       # %bb.0:
-; X64-WIN-NEXT:    andb $1, %cl
 ; X64-WIN-NEXT:    movzbl %cl, %eax
+; X64-WIN-NEXT:    movd %eax, %xmm0
+; X64-WIN-NEXT:    pshufd {{.*#+}} xmm0 = xmm0[0,1,0,1]
+; X64-WIN-NEXT:    pand __xmm@00000000000000010000000000000001(%rip), %xmm0
+; X64-WIN-NEXT:    pextrw $4, %xmm0, %eax
 ; X64-WIN-NEXT:    retq
   %tmp = insertelement <2 x i1> undef, i1 %tobool, i32 1
   %tmp1 = zext <2 x i1> %tmp to <2 x i64>