AMDGPU: Cost model for minimumnum/maximumnum #141946

arsenm · 2025-05-29T13:37:08Z

No description provided.

arsenm · 2025-05-29T13:37:23Z

This stack of pull requests is managed by Graphite. Learn more about stacking.

llvmbot · 2025-05-29T17:06:49Z

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

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

Patch is 183.80 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/141946.diff

4 Files Affected:

(modified) llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp (+19)
(modified) llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll (+318-264)
(modified) llvm/test/Analysis/CostModel/AMDGPU/minimumnum.ll (+318-264)
(modified) llvm/test/Transforms/SLPVectorizer/AMDGPU/slp-v2f16.ll (+40)

diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
index 563c46f57dfa5..357f8c5cfcd02 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
@@ -685,6 +685,8 @@ static bool intrinsicHasPackedVectorBenefit(Intrinsic::ID ID) {
   case Intrinsic::fma:
   case Intrinsic::fmuladd:
   case Intrinsic::copysign:
+  case Intrinsic::minimumnum:
+  case Intrinsic::maximumnum:
   case Intrinsic::canonicalize:
   // There's a small benefit to using vector ops in the legalized code.
   case Intrinsic::round:
@@ -742,6 +744,23 @@ GCNTTIImpl::getIntrinsicInstrCost(const IntrinsicCostAttributes &ICA,
     break;
   case Intrinsic::copysign:
     return NElts * getFullRateInstrCost();
+  case Intrinsic::minimumnum:
+  case Intrinsic::maximumnum: {
+    // Instruction + 2 canonicalizes. For cases that need type promotion, we the
+    // promotion takes the place of the canonicalize.
+    unsigned NumOps = 3;
+    if (const IntrinsicInst *II = ICA.getInst()) {
+      // Directly legal with ieee=0
+      // TODO: Not directly legal with strictfp
+      if (fpenvIEEEMode(*II) == KnownIEEEMode::Off)
+        NumOps = 1;
+    }
+
+    unsigned BaseRate =
+        SLT == MVT::f64 ? get64BitInstrCost(CostKind) : getFullRateInstrCost();
+    InstRate = BaseRate * NumOps;
+    break;
+  }
   case Intrinsic::canonicalize: {
     InstRate =
         SLT == MVT::f64 ? get64BitInstrCost(CostKind) : getFullRateInstrCost();
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll b/llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll
index 5b158e3d8d674..a81cb63f0c51f 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/maximumnum.ll
@@ -11,75 +11,75 @@
 
 define void @maximumnum_f16() {
 ; GFX7-LABEL: 'maximumnum_f16'
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX7-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX7-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX7-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX7-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX7-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX7-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX7-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX8-LABEL: 'maximumnum_f16'
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX8-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX8-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX9-LABEL: 'maximumnum_f16'
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX9-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX9-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX9-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX12-LABEL: 'maximumnum_f16'
-; GFX12-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX12-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX12-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX12-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX12-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX12-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX12-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX12-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX12-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX12-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX12-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX12-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX12-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
 ;
 ; GFX7-SIZE-LABEL: 'maximumnum_f16'
-; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX7-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX8-SIZE-LABEL: 'maximumnum_f16'
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 48 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX8-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX9-SIZE-LABEL: 'maximumnum_f16'
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX9-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
 ; GFX12-SIZE-LABEL: 'maximumnum_f16'
-; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
-; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
-; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
-; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
-; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
-; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
+; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
+; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %v2f16 = call <2 x half> @llvm.maximumnum.v2f16(<2 x half> poison, <2 x half> poison)
+; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v3f16 = call <3 x half> @llvm.maximumnum.v3f16(<3 x half> poison, <3 x half> poison)
+; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %v4f16 = call <4 x half> @llvm.maximumnum.v4f16(<4 x half> poison, <4 x half> poison)
+; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %v8f16 = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> poison, <8 x half> poison)
+; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %v16f16 = call <16 x half> @llvm.maximumnum.v16f16(<16 x half> poison, <16 x half> poison)
 ; GFX12-SIZE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: ret void
 ;
   %f16 = call half @llvm.maximumnum.f16(half poison, half poison)
@@ -92,77 +92,23 @@ define void @maximumnum_f16() {
 }
 
 define void @maximumnum_bf16() {
-; GFX7-LABEL: 'maximumnum_bf16'
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bf16 = call bfloat @llvm.maximumnum.bf16(bfloat poison, bfloat poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %v2bf16 = call <2 x bfloat> @llvm.maximumnum.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v3bf16 = call <3 x bfloat> @llvm.maximumnum.v3bf16(<3 x bfloat> poison, <3 x bfloat> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %v4bf16 = call <4 x bfloat> @llvm.maximumnum.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %v8bf16 = call <8 x bfloat> @llvm.maximumnum.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v16bf16 = call <16 x bfloat> @llvm.maximumnum.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
-; GFX7-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: ret void
-;
-; GFX8-LABEL: 'maximumnum_bf16'
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %bf16 = call bfloat @llvm.maximumnum.bf16(bfloat poison, bfloat poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 21 for instruction: %v2bf16 = call <2 x bfloat> @llvm.maximumnum.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %v3bf16 = call <3 x bfloat> @llvm.maximumnum.v3bf16(<3 x bfloat> poison, <3 x bfloat> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 43 for instruction: %v4bf16 = call <4 x bfloat> @llvm.maximumnum.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 87 for instruction: %v8bf16 = call <8 x bfloat> @llvm.maximumnum.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated cost of 175 for instruction: %v16bf16 = call <16 x bfloat> @llvm.maximumnum.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
-; GFX8-NEXT:  Cost Model: Found an estimated...
[truncated]

Pierre-vh · 2025-06-02T07:36:11Z

llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp

+    // Instruction + 2 canonicalizes. For cases that need type promotion, we the
+    // promotion takes the place of the canonicalize.


Suggested change

// Instruction + 2 canonicalizes. For cases that need type promotion, we the

// promotion takes the place of the canonicalize.

// Instruction + 2 canonicalizes. For cases that need type promotion, the

// promotion takes the place of the canonicalize.

arsenm · 2025-06-17T22:54:38Z

Merge activity

Jun 17, 10:54 PM UTC: A user started a stack merge that includes this pull request via Graphite.
Jun 17, 11:14 PM UTC: Graphite rebased this pull request as part of a merge.
Jun 17, 11:16 PM UTC: Graphite rebased this pull request as part of a merge.
Jun 17, 11:19 PM UTC: @arsenm merged this pull request with Graphite.

This was referenced May 29, 2025

AMDGPU: Reduce cost of f64 copysign #141944

Merged

AMDGPU: Add baseline cost model tests for special argument intrinsics #141947

Merged

AMDGPU: Report special input intrinsics as free #141948

Merged

arsenm added the backend:AMDGPU label May 29, 2025 — with Graphite App

arsenm requested review from jayfoad, jrbyrnes, mbrkusanin, Pierre-vh, shiltian and Sisyph May 29, 2025 17:06

arsenm marked this pull request as ready for review May 29, 2025 17:06

llvmbot added llvm:analysis Includes value tracking, cost tables and constant folding llvm:transforms labels May 29, 2025

Pierre-vh approved these changes Jun 2, 2025

View reviewed changes

arsenm force-pushed the users/arsenm/amdgpu/move-fpenvIEEEMode-tti branch from dc0d249 to fff3465 Compare June 17, 2025 13:21

arsenm force-pushed the users/arsenm/amdgpu/better-costs-minimumnum-maximumnum branch from 628e1c2 to 1725fdc Compare June 17, 2025 13:21

arsenm force-pushed the users/arsenm/amdgpu/move-fpenvIEEEMode-tti branch from fff3465 to 63d221a Compare June 17, 2025 15:32

arsenm force-pushed the users/arsenm/amdgpu/better-costs-minimumnum-maximumnum branch from 1725fdc to 06cc9b7 Compare June 17, 2025 15:32

arsenm force-pushed the users/arsenm/amdgpu/move-fpenvIEEEMode-tti branch from 63d221a to 9916237 Compare June 17, 2025 15:33

arsenm force-pushed the users/arsenm/amdgpu/better-costs-minimumnum-maximumnum branch from 06cc9b7 to 5839fa0 Compare June 17, 2025 15:33

arsenm force-pushed the users/arsenm/amdgpu/move-fpenvIEEEMode-tti branch from 9916237 to c9bbcd5 Compare June 17, 2025 23:11

Base automatically changed from users/arsenm/amdgpu/move-fpenvIEEEMode-tti to main June 17, 2025 23:13

arsenm force-pushed the users/arsenm/amdgpu/better-costs-minimumnum-maximumnum branch from 5839fa0 to 21915fa Compare June 17, 2025 23:14

AMDGPU: Cost model for minimumnum/maximumnum

a1b854b

arsenm force-pushed the users/arsenm/amdgpu/better-costs-minimumnum-maximumnum branch from 21915fa to a1b854b Compare June 17, 2025 23:16

arsenm merged commit 54015f3 into main Jun 17, 2025
5 of 7 checks passed

arsenm deleted the users/arsenm/amdgpu/better-costs-minimumnum-maximumnum branch June 17, 2025 23:19

fschlimb pushed a commit to fschlimb/llvm-project that referenced this pull request Jun 18, 2025

AMDGPU: Cost model for minimumnum/maximumnum (llvm#141946)

0b5393f

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

AMDGPU: Cost model for minimumnum/maximumnum #141946

AMDGPU: Cost model for minimumnum/maximumnum #141946

Uh oh!

arsenm commented May 29, 2025

Uh oh!

arsenm commented May 29, 2025 •

edited

Loading

Uh oh!

llvmbot commented May 29, 2025 •

edited

Loading

Uh oh!

Pierre-vh Jun 2, 2025

Uh oh!

arsenm commented Jun 17, 2025 •

edited

Loading

Uh oh!

Uh oh!

Uh oh!

		// Instruction + 2 canonicalizes. For cases that need type promotion, we the
		// promotion takes the place of the canonicalize.

AMDGPU: Cost model for minimumnum/maximumnum #141946

AMDGPU: Cost model for minimumnum/maximumnum #141946

Uh oh!

Conversation

arsenm commented May 29, 2025

Uh oh!

arsenm commented May 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

llvmbot commented May 29, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Pierre-vh Jun 2, 2025

Choose a reason for hiding this comment

Uh oh!

arsenm commented Jun 17, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Merge activity

Uh oh!

Uh oh!

Uh oh!

arsenm commented May 29, 2025 •

edited

Loading

llvmbot commented May 29, 2025 •

edited

Loading

arsenm commented Jun 17, 2025 •

edited

Loading