[llvm] e7630a0 - AMDGPU: Improve cost handling of canonicalize (#101479)

via llvm-commits llvm-commits at lists.llvm.org
Thu Aug 1 08:02:23 PDT 2024


Author: Matt Arsenault
Date: 2024-08-01T19:02:20+04:00
New Revision: e7630a0d60821dc13bb0be4e50b49fba5f90471f

URL: https://github.com/llvm/llvm-project/commit/e7630a0d60821dc13bb0be4e50b49fba5f90471f
DIFF: https://github.com/llvm/llvm-project/commit/e7630a0d60821dc13bb0be4e50b49fba5f90471f.diff

LOG: AMDGPU: Improve cost handling of canonicalize (#101479)

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
    llvm/test/Analysis/CostModel/AMDGPU/canonicalize.ll
    llvm/test/Transforms/SLPVectorizer/AMDGPU/packed-math.ll
    llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 5191fb0673cc7..8d4ff64ac5adc 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -689,6 +689,7 @@ static bool intrinsicHasPackedVectorBenefit(Intrinsic::ID ID) {
   case Intrinsic::fma:
   case Intrinsic::fmuladd:
   case Intrinsic::copysign:
+  case Intrinsic::canonicalize:
   // There's a small benefit to using vector ops in the legalized code.
   case Intrinsic::round:
   case Intrinsic::uadd_sat:
@@ -742,15 +743,23 @@ GCNTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     break;
   case Intrinsic::copysign:
     return NElts * getFullRateInstrCost();
+  case Intrinsic::canonicalize: {
+    InstRate =
+        SLT == MVT::f64 ? get64BitInstrCost(CostKind) : getFullRateInstrCost();
+    break;
+  }
   case Intrinsic::uadd_sat:
   case Intrinsic::usub_sat:
   case Intrinsic::sadd_sat:
-  case Intrinsic::ssub_sat:
+  case Intrinsic::ssub_sat: {
     static const auto ValidSatTys = {MVT::v2i16, MVT::v4i16};
     if (any_of(ValidSatTys, [&LT](MVT M) { return M == LT.second; }))
       NElts = 1;
     break;
   }
+  default:
+    break;
+  }
 
   return LT.first * NElts * InstRate;
 }

diff  --git a/llvm/test/Analysis/CostModel/AMDGPU/canonicalize.ll b/llvm/test/Analysis/CostModel/AMDGPU/canonicalize.ll
index e980c91048050..e162edbf611e2 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/canonicalize.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/canonicalize.ll
@@ -12,22 +12,22 @@
 define void @canonicalize_f16() {
 ; BASE-LABEL: 'canonicalize_f16'
 ; BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %f16 = call half @llvm.canonicalize.f16(half undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; BASE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX8-LABEL: 'canonicalize_f16'
 ; GFX8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %f16 = call half @llvm.canonicalize.f16(half undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; GFX8-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-LABEL: 'canonicalize_f16'
@@ -35,9 +35,9 @@ define void @canonicalize_f16() {
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX10-LABEL: 'canonicalize_f16'
@@ -45,29 +45,29 @@ define void @canonicalize_f16() {
 ; GFX10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
 ; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
 ; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; GFX10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; BASE-SIZE-LABEL: 'canonicalize_f16'
 ; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %f16 = call half @llvm.canonicalize.f16(half undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX8-SIZE-LABEL: 'canonicalize_f16'
 ; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %f16 = call half @llvm.canonicalize.f16(half undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX9-SIZE-LABEL: 'canonicalize_f16'
@@ -75,9 +75,9 @@ define void @canonicalize_f16() {
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX10-SIZE-LABEL: 'canonicalize_f16'
@@ -85,9 +85,9 @@ define void @canonicalize_f16() {
 ; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %v2f16 = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> undef)
 ; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.canonicalize.v3f16(<3 x half> undef)
 ; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.canonicalize.v4f16(<4 x half> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v5f16 = call <5 x half> @llvm.canonicalize.v5f16(<5 x half> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v16f16 = call <16 x half> @llvm.canonicalize.v16f16(<16 x half> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v17f16 = call <17 x half> @llvm.canonicalize.v17f16(<17 x half> undef)
 ; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %f16 = call half @llvm.canonicalize.f16(half undef) #1
@@ -103,82 +103,82 @@ define void @canonicalize_f16() {
 define void @canonicalize_bf16() {
 ; BASE-LABEL: 'canonicalize_bf16'
 ; BASE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; BASE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; BASE-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; BASE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX8-LABEL: 'canonicalize_bf16'
 ; GFX8-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; GFX8-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-LABEL: 'canonicalize_bf16'
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX10-LABEL: 'canonicalize_bf16'
 ; GFX10-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; GFX10-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; GFX10-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; GFX10-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; BASE-SIZE-LABEL: 'canonicalize_bf16'
 ; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; BASE-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX8-SIZE-LABEL: 'canonicalize_bf16'
 ; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX9-SIZE-LABEL: 'canonicalize_bf16'
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX10-SIZE-LABEL: 'canonicalize_bf16'
 ; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
-; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 33 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2bf16 = call <2 x bfloat> @llvm.canonicalize.v2bf16(<2 x bfloat> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v3bf16 = call <3 x bfloat> @llvm.canonicalize.v3bf16(<3 x bfloat> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4bf16 = call <4 x bfloat> @llvm.canonicalize.v4bf16(<4 x bfloat> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v5bf16 = call <5 x bfloat> @llvm.canonicalize.v5bf16(<5 x bfloat> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16bf16 = call <16 x bfloat> @llvm.canonicalize.v16bf16(<16 x bfloat> undef)
+; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v17bf16 = call <17 x bfloat> @llvm.canonicalize.v17bf16(<17 x bfloat> undef)
 ; GFX10-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %bf16 = call bfloat @llvm.canonicalize.bf16(bfloat undef) #1
@@ -199,8 +199,8 @@ define void @canonicalize_f32() {
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4f32 = call <4 x float> @llvm.canonicalize.v4f32(<4 x float> undef)
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v5f32 = call <5 x float> @llvm.canonicalize.v5f32(<5 x float> undef)
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8f32 = call <8 x float> @llvm.canonicalize.v8f32(<8 x float> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v9f32 = call <9 x float> @llvm.canonicalize.v9f32(<9 x float> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16f32 = call <16 x float> @llvm.canonicalize.v16f32(<16 x float> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %v9f32 = call <9 x float> @llvm.canonicalize.v9f32(<9 x float> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v16f32 = call <16 x float> @llvm.canonicalize.v16f32(<16 x float> undef)
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'canonicalize_f32'
@@ -210,8 +210,8 @@ define void @canonicalize_f32() {
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4f32 = call <4 x float> @llvm.canonicalize.v4f32(<4 x float> undef)
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v5f32 = call <5 x float> @llvm.canonicalize.v5f32(<5 x float> undef)
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8f32 = call <8 x float> @llvm.canonicalize.v8f32(<8 x float> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %v9f32 = call <9 x float> @llvm.canonicalize.v9f32(<9 x float> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16f32 = call <16 x float> @llvm.canonicalize.v16f32(<16 x float> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 27 for instruction: %v9f32 = call <9 x float> @llvm.canonicalize.v9f32(<9 x float> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v16f32 = call <16 x float> @llvm.canonicalize.v16f32(<16 x float> undef)
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %f32 = call float @llvm.canonicalize.f32(float undef) #1
@@ -227,23 +227,23 @@ define void @canonicalize_f32() {
 
 define void @canonicalize_f64() {
 ; ALL-LABEL: 'canonicalize_f64'
-; ALL-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %f64 = call double @llvm.canonicalize.f64(double undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f64 = call <2 x double> @llvm.canonicalize.v2f64(<2 x double> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v3f64 = call <3 x double> @llvm.canonicalize.v3f64(<3 x double> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4f64 = call <4 x double> @llvm.canonicalize.v4f64(<4 x double> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v5f64 = call <5 x double> @llvm.canonicalize.v5f64(<5 x double> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8f64 = call <8 x double> @llvm.canonicalize.v8f64(<8 x double> undef)
-; ALL-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16f64 = call <16 x double> @llvm.canonicalize.v16f64(<16 x double> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %f64 = call double @llvm.canonicalize.f64(double undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v2f64 = call <2 x double> @llvm.canonicalize.v2f64(<2 x double> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v3f64 = call <3 x double> @llvm.canonicalize.v3f64(<3 x double> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v4f64 = call <4 x double> @llvm.canonicalize.v4f64(<4 x double> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v5f64 = call <5 x double> @llvm.canonicalize.v5f64(<5 x double> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 96 for instruction: %v8f64 = call <8 x double> @llvm.canonicalize.v8f64(<8 x double> undef)
+; ALL-NEXT:  Cost Model: Found an estimated cost of 320 for instruction: %v16f64 = call <16 x double> @llvm.canonicalize.v16f64(<16 x double> undef)
 ; ALL-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; ALL-SIZE-LABEL: 'canonicalize_f64'
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %f64 = call double @llvm.canonicalize.f64(double undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f64 = call <2 x double> @llvm.canonicalize.v2f64(<2 x double> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v3f64 = call <3 x double> @llvm.canonicalize.v3f64(<3 x double> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v4f64 = call <4 x double> @llvm.canonicalize.v4f64(<4 x double> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v5f64 = call <5 x double> @llvm.canonicalize.v5f64(<5 x double> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v8f64 = call <8 x double> @llvm.canonicalize.v8f64(<8 x double> undef)
-; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v16f64 = call <16 x double> @llvm.canonicalize.v16f64(<16 x double> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f64 = call double @llvm.canonicalize.f64(double undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2f64 = call <2 x double> @llvm.canonicalize.v2f64(<2 x double> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v3f64 = call <3 x double> @llvm.canonicalize.v3f64(<3 x double> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4f64 = call <4 x double> @llvm.canonicalize.v4f64(<4 x double> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v5f64 = call <5 x double> @llvm.canonicalize.v5f64(<5 x double> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v8f64 = call <8 x double> @llvm.canonicalize.v8f64(<8 x double> undef)
+; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 160 for instruction: %v16f64 = call <16 x double> @llvm.canonicalize.v16f64(<16 x double> undef)
 ; ALL-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %f64 = call double @llvm.canonicalize.f64(double undef) #1

diff  --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/packed-math.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/packed-math.ll
index 7152115220b00..429640ea7ff16 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/packed-math.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/packed-math.ll
@@ -7,10 +7,10 @@
 ; Simple 3-pair chain with loads and stores
 define amdgpu_kernel void @test1_as_3_3_3_v2f16(ptr addrspace(3) %a, ptr addrspace(3) %b, ptr addrspace(3) %c) {
 ; GCN-LABEL: @test1_as_3_3_3_v2f16(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP4:%.*]] = load <2 x half>, ptr addrspace(3) [[B:%.*]], align 2
-; GCN-NEXT:    [[TMP5:%.*]] = fmul <2 x half> [[TMP2]], [[TMP4]]
-; GCN-NEXT:    store <2 x half> [[TMP5]], ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[B:%.*]], align 2
+; GCN-NEXT:    [[TMP3:%.*]] = fmul <2 x half> [[TMP1]], [[TMP2]]
+; GCN-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(3) [[C:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
@@ -29,10 +29,10 @@ define amdgpu_kernel void @test1_as_3_3_3_v2f16(ptr addrspace(3) %a, ptr addrspa
 
 define amdgpu_kernel void @test1_as_3_0_0(ptr addrspace(3) %a, ptr %b, ptr %c) {
 ; GCN-LABEL: @test1_as_3_0_0(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP4:%.*]] = load <2 x half>, ptr [[B:%.*]], align 2
-; GCN-NEXT:    [[TMP5:%.*]] = fmul <2 x half> [[TMP2]], [[TMP4]]
-; GCN-NEXT:    store <2 x half> [[TMP5]], ptr [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr [[B:%.*]], align 2
+; GCN-NEXT:    [[TMP3:%.*]] = fmul <2 x half> [[TMP1]], [[TMP2]]
+; GCN-NEXT:    store <2 x half> [[TMP3]], ptr [[C:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
@@ -51,10 +51,10 @@ define amdgpu_kernel void @test1_as_3_0_0(ptr addrspace(3) %a, ptr %b, ptr %c) {
 
 define amdgpu_kernel void @test1_as_0_0_3_v2f16(ptr %a, ptr %b, ptr addrspace(3) %c) {
 ; GCN-LABEL: @test1_as_0_0_3_v2f16(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP4:%.*]] = load <2 x half>, ptr [[B:%.*]], align 2
-; GCN-NEXT:    [[TMP5:%.*]] = fmul <2 x half> [[TMP2]], [[TMP4]]
-; GCN-NEXT:    store <2 x half> [[TMP5]], ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr [[B:%.*]], align 2
+; GCN-NEXT:    [[TMP3:%.*]] = fmul <2 x half> [[TMP1]], [[TMP2]]
+; GCN-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(3) [[C:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr %a, align 2
@@ -73,11 +73,11 @@ define amdgpu_kernel void @test1_as_0_0_3_v2f16(ptr %a, ptr %b, ptr addrspace(3)
 
 define amdgpu_kernel void @test1_fma_v2f16(ptr addrspace(3) %a, ptr addrspace(3) %b, ptr addrspace(3) %c, ptr addrspace(3) %d) {
 ; GCN-LABEL: @test1_fma_v2f16(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP4:%.*]] = load <2 x half>, ptr addrspace(3) [[B:%.*]], align 2
-; GCN-NEXT:    [[TMP6:%.*]] = load <2 x half>, ptr addrspace(3) [[C:%.*]], align 2
-; GCN-NEXT:    [[TMP7:%.*]] = call <2 x half> @llvm.fma.v2f16(<2 x half> [[TMP2]], <2 x half> [[TMP4]], <2 x half> [[TMP6]])
-; GCN-NEXT:    store <2 x half> [[TMP7]], ptr addrspace(3) [[D:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[B:%.*]], align 2
+; GCN-NEXT:    [[TMP3:%.*]] = load <2 x half>, ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP4:%.*]] = call <2 x half> @llvm.fma.v2f16(<2 x half> [[TMP1]], <2 x half> [[TMP2]], <2 x half> [[TMP3]])
+; GCN-NEXT:    store <2 x half> [[TMP4]], ptr addrspace(3) [[D:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
@@ -99,10 +99,10 @@ define amdgpu_kernel void @test1_fma_v2f16(ptr addrspace(3) %a, ptr addrspace(3)
 
 define amdgpu_kernel void @mul_scalar_v2f16(ptr addrspace(3) %a, half %scalar, ptr addrspace(3) %c) {
 ; GCN-LABEL: @mul_scalar_v2f16(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP3:%.*]] = insertelement <2 x half> poison, half [[SCALAR:%.*]], i32 0
-; GCN-NEXT:    [[SHUFFLE:%.*]] = shufflevector <2 x half> [[TMP3]], <2 x half> poison, <2 x i32> zeroinitializer
-; GCN-NEXT:    [[TMP4:%.*]] = fmul <2 x half> [[TMP2]], [[SHUFFLE]]
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = insertelement <2 x half> poison, half [[SCALAR:%.*]], i32 0
+; GCN-NEXT:    [[TMP3:%.*]] = shufflevector <2 x half> [[TMP2]], <2 x half> poison, <2 x i32> zeroinitializer
+; GCN-NEXT:    [[TMP4:%.*]] = fmul <2 x half> [[TMP1]], [[TMP3]]
 ; GCN-NEXT:    store <2 x half> [[TMP4]], ptr addrspace(3) [[C:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
@@ -119,9 +119,9 @@ define amdgpu_kernel void @mul_scalar_v2f16(ptr addrspace(3) %a, half %scalar, p
 
 define amdgpu_kernel void @fabs_v2f16(ptr addrspace(3) %a, ptr addrspace(3) %c) {
 ; GCN-LABEL: @fabs_v2f16(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP3:%.*]] = call <2 x half> @llvm.fabs.v2f16(<2 x half> [[TMP2]])
-; GCN-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = call <2 x half> @llvm.fabs.v2f16(<2 x half> [[TMP1]])
+; GCN-NEXT:    store <2 x half> [[TMP2]], ptr addrspace(3) [[C:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
@@ -137,12 +137,12 @@ define amdgpu_kernel void @fabs_v2f16(ptr addrspace(3) %a, ptr addrspace(3) %c)
 
 define amdgpu_kernel void @test1_fabs_fma_v2f16(ptr addrspace(3) %a, ptr addrspace(3) %b, ptr addrspace(3) %c, ptr addrspace(3) %d) {
 ; GCN-LABEL: @test1_fabs_fma_v2f16(
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP4:%.*]] = load <2 x half>, ptr addrspace(3) [[B:%.*]], align 2
-; GCN-NEXT:    [[TMP6:%.*]] = load <2 x half>, ptr addrspace(3) [[C:%.*]], align 2
-; GCN-NEXT:    [[TMP7:%.*]] = call <2 x half> @llvm.fabs.v2f16(<2 x half> [[TMP2]])
-; GCN-NEXT:    [[TMP8:%.*]] = call <2 x half> @llvm.fma.v2f16(<2 x half> [[TMP7]], <2 x half> [[TMP4]], <2 x half> [[TMP6]])
-; GCN-NEXT:    store <2 x half> [[TMP8]], ptr addrspace(3) [[D:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[B:%.*]], align 2
+; GCN-NEXT:    [[TMP3:%.*]] = load <2 x half>, ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP4:%.*]] = call <2 x half> @llvm.fabs.v2f16(<2 x half> [[TMP1]])
+; GCN-NEXT:    [[TMP5:%.*]] = call <2 x half> @llvm.fma.v2f16(<2 x half> [[TMP4]], <2 x half> [[TMP2]], <2 x half> [[TMP3]])
+; GCN-NEXT:    store <2 x half> [[TMP5]], ptr addrspace(3) [[D:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
@@ -172,12 +172,12 @@ define amdgpu_kernel void @test1_fabs_scalar_fma_v2f16(ptr addrspace(3) %a, ptr
 ; GCN-NEXT:    [[I1_FABS:%.*]] = call half @llvm.fabs.f16(half [[I1]])
 ; GCN-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds half, ptr addrspace(3) [[B]], i64 1
 ; GCN-NEXT:    [[I4:%.*]] = load half, ptr addrspace(3) [[ARRAYIDX4]], align 2
-; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GCN-NEXT:    [[TMP4:%.*]] = load <2 x half>, ptr addrspace(3) [[C:%.*]], align 2
-; GCN-NEXT:    [[TMP5:%.*]] = insertelement <2 x half> poison, half [[I1_FABS]], i32 0
-; GCN-NEXT:    [[TMP6:%.*]] = insertelement <2 x half> [[TMP5]], half [[I4]], i32 1
-; GCN-NEXT:    [[TMP7:%.*]] = call <2 x half> @llvm.fma.v2f16(<2 x half> [[TMP2]], <2 x half> [[TMP6]], <2 x half> [[TMP4]])
-; GCN-NEXT:    store <2 x half> [[TMP7]], ptr addrspace(3) [[D:%.*]], align 2
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    [[TMP3:%.*]] = insertelement <2 x half> poison, half [[I1_FABS]], i32 0
+; GCN-NEXT:    [[TMP4:%.*]] = insertelement <2 x half> [[TMP3]], half [[I4]], i32 1
+; GCN-NEXT:    [[TMP5:%.*]] = call <2 x half> @llvm.fma.v2f16(<2 x half> [[TMP1]], <2 x half> [[TMP4]], <2 x half> [[TMP2]])
+; GCN-NEXT:    store <2 x half> [[TMP5]], ptr addrspace(3) [[D:%.*]], align 2
 ; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
@@ -200,22 +200,11 @@ define amdgpu_kernel void @test1_fabs_scalar_fma_v2f16(ptr addrspace(3) %a, ptr
 }
 
 define amdgpu_kernel void @canonicalize_v2f16(ptr addrspace(3) %a, ptr addrspace(3) %c) {
-; GFX9-LABEL: @canonicalize_v2f16(
-; GFX9-NEXT:    [[TMP2:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
-; GFX9-NEXT:    [[TMP3:%.*]] = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> [[TMP2]])
-; GFX9-NEXT:    store <2 x half> [[TMP3]], ptr addrspace(3) [[C:%.*]], align 2
-; GFX9-NEXT:    ret void
-;
-; VI-LABEL: @canonicalize_v2f16(
-; VI-NEXT:    [[I0:%.*]] = load half, ptr addrspace(3) [[A:%.*]], align 2
-; VI-NEXT:    [[CANONICALIZE0:%.*]] = call half @llvm.canonicalize.f16(half [[I0]])
-; VI-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds half, ptr addrspace(3) [[A]], i64 1
-; VI-NEXT:    [[I3:%.*]] = load half, ptr addrspace(3) [[ARRAYIDX3]], align 2
-; VI-NEXT:    [[CANONICALIZE1:%.*]] = call half @llvm.canonicalize.f16(half [[I3]])
-; VI-NEXT:    store half [[CANONICALIZE0]], ptr addrspace(3) [[C:%.*]], align 2
-; VI-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds half, ptr addrspace(3) [[C]], i64 1
-; VI-NEXT:    store half [[CANONICALIZE1]], ptr addrspace(3) [[ARRAYIDX5]], align 2
-; VI-NEXT:    ret void
+; GCN-LABEL: @canonicalize_v2f16(
+; GCN-NEXT:    [[TMP1:%.*]] = load <2 x half>, ptr addrspace(3) [[A:%.*]], align 2
+; GCN-NEXT:    [[TMP2:%.*]] = call <2 x half> @llvm.canonicalize.v2f16(<2 x half> [[TMP1]])
+; GCN-NEXT:    store <2 x half> [[TMP2]], ptr addrspace(3) [[C:%.*]], align 2
+; GCN-NEXT:    ret void
 ;
   %i0 = load half, ptr addrspace(3) %a, align 2
   %canonicalize0 = call half @llvm.canonicalize.f16(half %i0)
@@ -234,3 +223,6 @@ declare half @llvm.canonicalize.f16(half) #1
 
 attributes #0 = { nounwind }
 attributes #1 = { nounwind readnone }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX9: {{.*}}
+; VI: {{.*}}

diff  --git a/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll b/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll
index c51bdedd5be3f..c79fa9c84d1c3 100644
--- a/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll
+++ b/llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll
@@ -217,10 +217,7 @@ bb:
 }
 
 ; GCN-LABEL: @canonicalize_combine_v2f16
-; GFX8: call half @llvm.canonicalize.f16(
-; GFX8: call half @llvm.canonicalize.f16(
-
-; GFX9: call <2 x half> @llvm.canonicalize.v2f16(
+; GCN: call <2 x half> @llvm.canonicalize.v2f16(
 define void @canonicalize_combine_v2f16(ptr addrspace(1) %arg) {
 bb:
   %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()
@@ -330,11 +327,8 @@ bb:
 }
 
 ; GCN-LABEL: @canonicalize_combine_v4f16
-; GFX8: call half @llvm.canonicalize.f16(
-; GFX8: call half @llvm.canonicalize.f16(
-
-; GFX9: call <2 x half> @llvm.canonicalize.v2f16(
-; GFX9: call <2 x half> @llvm.canonicalize.v2f16(
+; GCN: call <2 x half> @llvm.canonicalize.v2f16(
+; GCN: call <2 x half> @llvm.canonicalize.v2f16(
 define void @canonicalize_combine_v4f16(ptr addrspace(1) %arg) {
 bb:
   %tmp = tail call i32 @llvm.amdgcn.workitem.id.x()


        


More information about the llvm-commits mailing list