Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

[AArch64] Add ZIP and UZP shuffle costs. #88150

Merged
merged 1 commit into from
Apr 11, 2024
Merged

[AArch64] Add ZIP and UZP shuffle costs. #88150

merged 1 commit into from
Apr 11, 2024

Conversation

davemgreen
Copy link
Collaborator

This adds some costs for the shuffle instructions that should be lowered to zip1/zip2/uzp1/uzp2 instructions.

@davemgreen
[AArch64] Add ZIP and UZP shuffle costs.
e046146 
This adds some costs for the shuffle instructions that should be lowered to
zip1/zip2/uzp1/uzp2 instructions.
@llvmbot
Copy link
Member

llvmbot commented Apr 9, 2024 •
edited
Loading

@llvm/pr-subscribers-llvm-analysis
@llvm/pr-subscribers-llvm-transforms

@llvm/pr-subscribers-backend-aarch64

Author: David Green (davemgreen)

Changes

This adds some costs for the shuffle instructions that should be lowered to zip1/zip2/uzp1/uzp2 instructions.

Patch is 30.82 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/88150.diff

6 Files Affected:

  • (modified) llvm/lib/Target/AArch64/AArch64ISelLowering.cpp (-29)
  • (modified) llvm/lib/Target/AArch64/AArch64PerfectShuffle.h (+33)
  • (modified) llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp (+10)
  • (modified) llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll (+33-33)
  • (modified) llvm/test/Analysis/CostModel/AArch64/shuffle-store.ll (+7-7)
  • (modified) llvm/test/Transforms/VectorCombine/AArch64/vecreduce-shuffle.ll (+1-1)
diff --git a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
index 819e8ccd5c33f0..43a0f960577bd2 100644
--- a/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
+++ b/llvm/lib/Target/AArch64/AArch64ISelLowering.cpp
@@ -11851,35 +11851,6 @@ static bool isREVMask(ArrayRef<int> M, EVT VT, unsigned BlockSize) {
   return true;
 }
 
-static bool isZIPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
-  unsigned NumElts = VT.getVectorNumElements();
-  if (NumElts % 2 != 0)
-    return false;
-  WhichResult = (M[0] == 0 ? 0 : 1);
-  unsigned Idx = WhichResult * NumElts / 2;
-  for (unsigned i = 0; i != NumElts; i += 2) {
-    if ((M[i] >= 0 && (unsigned)M[i] != Idx) ||
-        (M[i + 1] >= 0 && (unsigned)M[i + 1] != Idx + NumElts))
-      return false;
-    Idx += 1;
-  }
-
-  return true;
-}
-
-static bool isUZPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
-  unsigned NumElts = VT.getVectorNumElements();
-  WhichResult = (M[0] == 0 ? 0 : 1);
-  for (unsigned i = 0; i != NumElts; ++i) {
-    if (M[i] < 0)
-      continue; // ignore UNDEF indices
-    if ((unsigned)M[i] != 2 * i + WhichResult)
-      return false;
-  }
-
-  return true;
-}
-
 static bool isTRNMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
   unsigned NumElts = VT.getVectorNumElements();
   if (NumElts % 2 != 0)
diff --git a/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h b/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h
index 5846fd454b6548..7abaead694d114 100644
--- a/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h
+++ b/llvm/lib/Target/AArch64/AArch64PerfectShuffle.h
@@ -16,6 +16,8 @@
 
 #include "llvm/ADT/ArrayRef.h"
 
+namespace llvm {
+
 // 31 entries have cost 0
 // 756 entries have cost 1
 // 3690 entries have cost 2
@@ -6618,4 +6620,35 @@ static unsigned getPerfectShuffleCost(llvm::ArrayRef<int> M) {
   return (PFEntry >> 30) + 1;
 }
 
+inline bool isZIPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
+  unsigned NumElts = VT.getVectorNumElements();
+  if (NumElts % 2 != 0)
+    return false;
+  WhichResult = (M[0] == 0 ? 0 : 1);
+  unsigned Idx = WhichResult * NumElts / 2;
+  for (unsigned i = 0; i != NumElts; i += 2) {
+    if ((M[i] >= 0 && (unsigned)M[i] != Idx) ||
+        (M[i + 1] >= 0 && (unsigned)M[i + 1] != Idx + NumElts))
+      return false;
+    Idx += 1;
+  }
+
+  return true;
+}
+
+inline bool isUZPMask(ArrayRef<int> M, EVT VT, unsigned &WhichResult) {
+  unsigned NumElts = VT.getVectorNumElements();
+  WhichResult = (M[0] == 0 ? 0 : 1);
+  for (unsigned i = 0; i != NumElts; ++i) {
+    if (M[i] < 0)
+      continue; // ignore UNDEF indices
+    if ((unsigned)M[i] != 2 * i + WhichResult)
+      return false;
+  }
+
+  return true;
+}
+
+} // namespace llvm
+
 #endif
diff --git a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
index 20150d7386753f..bd943de06b4b2a 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetTransformInfo.cpp
@@ -3932,6 +3932,16 @@ InstructionCost AArch64TTIImpl::getShuffleCost(
       }))
     return 0;
 
+  // Check for other shuffles that are not SK_ kinds but we have native
+  // instructions for, for example ZIP and UZP.
+  unsigned Unused;
+  if (LT.second.isFixedLengthVector() &&
+      LT.second.getVectorNumElements() == Mask.size() &&
+      (Kind == TTI::SK_PermuteTwoSrc || Kind == TTI::SK_PermuteSingleSrc) &&
+      (isZIPMask(Mask, LT.second, Unused) ||
+       isUZPMask(Mask, LT.second, Unused)))
+    return 1;
+
   if (Kind == TTI::SK_Broadcast || Kind == TTI::SK_Transpose ||
       Kind == TTI::SK_Select || Kind == TTI::SK_PermuteSingleSrc ||
       Kind == TTI::SK_Reverse || Kind == TTI::SK_Splice) {
diff --git a/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll b/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
index d469ce63059322..6c45ebcb69f400 100644
--- a/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
+++ b/llvm/test/Analysis/CostModel/AArch64/shuffle-other.ll
@@ -171,28 +171,28 @@ define void @zip() {
 ; CHECK-LABEL: 'zip'
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 0, i32 2>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <2 x i32> <i32 1, i32 3>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %zipv2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %zip1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %zip2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %zipv4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %zipv8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %zip1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %zip2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %zipv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv2i8 = shufflevector <2 x i8> undef, <2 x i8> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zipv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v2i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 0, i32 2>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v2i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <2 x i32> <i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv2i16 = shufflevector <2 x i16> undef, <2 x i16> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %zipv4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %zip2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %zipv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %zip1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %zip2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %zipv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zipv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 8, i32 1, i32 9, i32 2, i32 10, i32 3, i32 11, i32 4, i32 12, i32 5, i32 13, i32 6, i32 14, i32 7, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zip1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %zip2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %zipv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 16, i32 1, i32 17, i32 2, i32 18, i32 3, i32 19, i32 4, i32 20, i32 5, i32 21, i32 6, i32 22, i32 7, i32 23, i32 8, i32 24, i32 9, i32 25, i32 10, i32 26, i32 11, i32 27, i32 12, i32 28, i32 13, i32 29, i32 14, i32 30, i32 15, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip1v2i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 0, i32 2>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zip2v2i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <2 x i32> <i32 1, i32 3>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %zipv2i32 = shufflevector <2 x i32> undef, <2 x i32> undef, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
@@ -276,24 +276,24 @@ define void @zip() {
 
 define void @uzp() {
 ; CHECK-LABEL: 'uzp'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %uzp1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %uzp2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %uzpv4i8 = shufflevector <4 x i8> undef, <4 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %uzpv8i8 = shufflevector <8 x i8> undef, <8 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %uzp1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 60 for instruction: %uzp2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 120 for instruction: %uzpv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %uzpv16i8 = shufflevector <16 x i8> undef, <16 x i8> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %uzpv4i16 = shufflevector <4 x i16> undef, <4 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 1, i32 3, i32 5, i32 7>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 28 for instruction: %uzp2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %uzpv8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %uzp1v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 56 for instruction: %uzp2v16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 112 for instruction: %uzpv16i16 = shufflevector <16 x i16> undef, <16 x i16> undef, <32 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 16, i32 18, i32 20, i32 22, i32 24, i32 26, i32 28, i32 30, i32 1, i32 3, i32 5, i32 7, i32 9, i32 11, i32 13, i32 15, i32 17, i32 19, i32 21, i32 23, i32 25, i32 27, i32 29, i32 31>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp1v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %uzp2v8i16 = shufflevector <8 x i16> undef, <8 x i16> undef, <8 x i32> <...
[truncated]

SamTebbs33
SamTebbs33 approved these changes Apr 10, 2024
View reviewed changes
Copy link
Collaborator

@SamTebbs33 SamTebbs33 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

@davemgreen
Copy link
Collaborator Author

Thanks

@davemgreen davemgreen merged commit a536743 into llvm:main Apr 11, 2024
8 checks passed
@davemgreen davemgreen deleted the gh-a64-shufflecost-zip branch April 11, 2024 07:45
@sjoerdmeijer
Copy link
Collaborator

Just a question about performance: did you measure any perf improvements/regressions?

@davemgreen
Copy link
Collaborator Author

Hello. Yes it fixed some regressions from #87510, in a similar way to #87934. Do you have any cases where it is causing problems?

@sjoerdmeijer
Copy link
Collaborator

Hello. Yes it fixed some regressions from #87510, in a similar way to #87934. Do you have any cases where it is causing problems?

Ok, that's great. I don't have any specific cases, it was just a curiousity because shuffle vectors has caused us some pain and was wondering if making them cheaper (at least some variants) would lead to any surprises.

This also LGTM, let's see how this behaves.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@SamTebbs33 SamTebbs33 SamTebbs33 approved these changes

@stuij stuij Awaiting requested review from stuij

@labrinea labrinea Awaiting requested review from labrinea

@sjoerdmeijer sjoerdmeijer Awaiting requested review from sjoerdmeijer

Assignees
No one assigned
Labels
backend:AArch64 llvm:analysis llvm:transforms
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@davemgreen @llvmbot @sjoerdmeijer @SamTebbs33