Skip to content

TTI: Check legalization cost of abs nodes #100523

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
Merged
merged 1 commit into from
Aug 9, 2024
Merged

TTI: Check legalization cost of abs nodes #100523

merged 1 commit into from
Aug 9, 2024

Conversation

@arsenm
Copy link
Contributor

@arsenm arsenm commented Jul 25, 2024

No description provided.

This was referenced Jul 25, 2024
Merged
Merged
Merged
Merged
Merged
@arsenm Graphite App
Copy link
Contributor Author

arsenm commented Jul 25, 2024 •
edited
Loading

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

Join @arsenm and the rest of your teammates on Graphite Graphite

This was referenced Jul 25, 2024
Merged
Merged
Merged
@arsenm arsenm added backend:AMDGPU vectorizers llvm:analysis Includes value tracking, cost tables and constant folding llvm:transforms labels Jul 25, 2024 — with Graphite App
@arsenm arsenm marked this pull request as ready for review July 25, 2024 07:05
@llvmbot
Copy link
Member

llvmbot commented Jul 25, 2024 •
edited
Loading

@llvm/pr-subscribers-llvm-transforms

@llvm/pr-subscribers-llvm-analysis

Author: Matt Arsenault (arsenm)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/100523.diff

2 Files Affected:

  • (modified) llvm/include/llvm/CodeGen/BasicTTIImpl.h (+18-14)
  • (modified) llvm/test/Analysis/CostModel/AMDGPU/abs.ll (+20-20)
diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index ba70498bfb731..65f929369c1f0 100644
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -2116,20 +2116,9 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
     case Intrinsic::vector_reduce_fminimum:
       return thisT()->getMinMaxReductionCost(getMinMaxReductionIntrinsicOp(IID),
                                              VecOpTy, ICA.getFlags(), CostKind);
-    case Intrinsic::abs: {
-      // abs(X) = select(icmp(X,0),X,sub(0,X))
-      Type *CondTy = RetTy->getWithNewBitWidth(1);
-      CmpInst::Predicate Pred = CmpInst::ICMP_SGT;
-      InstructionCost Cost = 0;
-      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::ICmp, RetTy, CondTy,
-                                          Pred, CostKind);
-      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::Select, RetTy, CondTy,
-                                          Pred, CostKind);
-      // TODO: Should we add an OperandValueProperties::OP_Zero property?
-      Cost += thisT()->getArithmeticInstrCost(
-         BinaryOperator::Sub, RetTy, CostKind, {TTI::OK_UniformConstantValue, TTI::OP_None});
-      return Cost;
-    }
+    case Intrinsic::abs:
+      ISD = ISD::ABS;
+      break;
     case Intrinsic::smax:
       ISD = ISD::SMAX;
       break;
@@ -2398,6 +2387,21 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
       Cost += thisT()->getArithmeticInstrCost(Instruction::Or, RetTy, CostKind);
       return Cost;
     }
+    case Intrinsic::abs: {
+      // abs(X) = select(icmp(X,0),X,sub(0,X))
+      Type *CondTy = RetTy->getWithNewBitWidth(1);
+      CmpInst::Predicate Pred = CmpInst::ICMP_SGT;
+      InstructionCost Cost = 0;
+      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::ICmp, RetTy, CondTy,
+                                          Pred, CostKind);
+      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::Select, RetTy, CondTy,
+                                          Pred, CostKind);
+      // TODO: Should we add an OperandValueProperties::OP_Zero property?
+      Cost += thisT()->getArithmeticInstrCost(
+          BinaryOperator::Sub, RetTy, CostKind,
+          {TTI::OK_UniformConstantValue, TTI::OP_None});
+      return Cost;
+    }
     case Intrinsic::fptosi_sat:
     case Intrinsic::fptoui_sat: {
       if (Tys.empty())
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/abs.ll b/llvm/test/Analysis/CostModel/AMDGPU/abs.ll
index 133b95609bc15..623e02eb8239d 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/abs.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/abs.ll
@@ -54,11 +54,11 @@ define i32 @abs_nonpoison(i32 %arg) {
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 174 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 false)
@@ -112,11 +112,11 @@ define i32 @abs_nonpoison(i32 %arg) {
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 false)
@@ -204,11 +204,11 @@ define i32 @abs_poison(i32 %arg) {
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 174 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 true)
@@ -262,11 +262,11 @@ define i32 @abs_poison(i32 %arg) {
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 true)

@llvmbot
Copy link
Member

llvmbot commented Jul 25, 2024

@llvm/pr-subscribers-backend-amdgpu

Author: Matt Arsenault (arsenm)

Changes

Full diff: https://github.com/llvm/llvm-project/pull/100523.diff

2 Files Affected:

  • (modified) llvm/include/llvm/CodeGen/BasicTTIImpl.h (+18-14)
  • (modified) llvm/test/Analysis/CostModel/AMDGPU/abs.ll (+20-20)
diff --git a/llvm/include/llvm/CodeGen/BasicTTIImpl.h b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
index ba70498bfb731..65f929369c1f0 100644
--- a/llvm/include/llvm/CodeGen/BasicTTIImpl.h
+++ b/llvm/include/llvm/CodeGen/BasicTTIImpl.h
@@ -2116,20 +2116,9 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
     case Intrinsic::vector_reduce_fminimum:
       return thisT()->getMinMaxReductionCost(getMinMaxReductionIntrinsicOp(IID),
                                              VecOpTy, ICA.getFlags(), CostKind);
-    case Intrinsic::abs: {
-      // abs(X) = select(icmp(X,0),X,sub(0,X))
-      Type *CondTy = RetTy->getWithNewBitWidth(1);
-      CmpInst::Predicate Pred = CmpInst::ICMP_SGT;
-      InstructionCost Cost = 0;
-      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::ICmp, RetTy, CondTy,
-                                          Pred, CostKind);
-      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::Select, RetTy, CondTy,
-                                          Pred, CostKind);
-      // TODO: Should we add an OperandValueProperties::OP_Zero property?
-      Cost += thisT()->getArithmeticInstrCost(
-         BinaryOperator::Sub, RetTy, CostKind, {TTI::OK_UniformConstantValue, TTI::OP_None});
-      return Cost;
-    }
+    case Intrinsic::abs:
+      ISD = ISD::ABS;
+      break;
     case Intrinsic::smax:
       ISD = ISD::SMAX;
       break;
@@ -2398,6 +2387,21 @@ class BasicTTIImplBase : public TargetTransformInfoImplCRTPBase<T> {
       Cost += thisT()->getArithmeticInstrCost(Instruction::Or, RetTy, CostKind);
       return Cost;
     }
+    case Intrinsic::abs: {
+      // abs(X) = select(icmp(X,0),X,sub(0,X))
+      Type *CondTy = RetTy->getWithNewBitWidth(1);
+      CmpInst::Predicate Pred = CmpInst::ICMP_SGT;
+      InstructionCost Cost = 0;
+      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::ICmp, RetTy, CondTy,
+                                          Pred, CostKind);
+      Cost += thisT()->getCmpSelInstrCost(BinaryOperator::Select, RetTy, CondTy,
+                                          Pred, CostKind);
+      // TODO: Should we add an OperandValueProperties::OP_Zero property?
+      Cost += thisT()->getArithmeticInstrCost(
+          BinaryOperator::Sub, RetTy, CostKind,
+          {TTI::OK_UniformConstantValue, TTI::OP_None});
+      return Cost;
+    }
     case Intrinsic::fptosi_sat:
     case Intrinsic::fptoui_sat: {
       if (Tys.empty())
diff --git a/llvm/test/Analysis/CostModel/AMDGPU/abs.ll b/llvm/test/Analysis/CostModel/AMDGPU/abs.ll
index 133b95609bc15..623e02eb8239d 100644
--- a/llvm/test/Analysis/CostModel/AMDGPU/abs.ll
+++ b/llvm/test/Analysis/CostModel/AMDGPU/abs.ll
@@ -54,11 +54,11 @@ define i32 @abs_nonpoison(i32 %arg) {
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 174 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 false)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 false)
@@ -112,11 +112,11 @@ define i32 @abs_nonpoison(i32 %arg) {
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 false)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 false)
@@ -204,11 +204,11 @@ define i32 @abs_poison(i32 %arg) {
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 80 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 7 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 34 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 70 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 114 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
-; FAST-NEXT:  Cost Model: Found an estimated cost of 174 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
+; FAST-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 true)
 ; FAST-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 true)
@@ -262,11 +262,11 @@ define i32 @abs_poison(i32 %arg) {
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V16I32 = call <16 x i32> @llvm.abs.v16i32(<16 x i32> undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I16 = call i16 @llvm.abs.i16(i16 undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2I16 = call <2 x i16> @llvm.abs.v2i16(<2 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
-; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 50 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4I16 = call <4 x i16> @llvm.abs.v4i16(<4 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 true)
+; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %I8 = call i8 @llvm.abs.i8(i8 undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2I8 = call <2 x i8> @llvm.abs.v2i8(<2 x i8> undef, i1 true)
 ; FAST-SIZE-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V4I8 = call <4 x i8> @llvm.abs.v4i8(<4 x i8> undef, i1 true)

jayfoad
jayfoad reviewed Jul 25, 2024
View reviewed changes
llvm/test/Analysis/CostModel/AMDGPU/abs.ll Outdated
; FAST-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V8I16 = call <8 x i16> @llvm.abs.v8i16(<8 x i16> undef, i1 false)
; FAST-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V16I16 = call <16 x i16> @llvm.abs.v16i16(<16 x i16> undef, i1 false)
; FAST-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V17I16 = call <17 x i16> @llvm.abs.v17i16(<17 x i16> undef, i1 false)
; FAST-NEXT: Cost Model: Found an estimated cost of 2 for instruction: %V32I16 = call <32 x i16> @llvm.abs.v32i16(<32 x i16> undef, i1 false)
Copy link
Contributor

@jayfoad jayfoad Jul 25, 2024

Choose a reason for hiding this comment

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

What is this demonstrating? 2 does not seem like the right cost for any VALU/SALU operation on v32i16.

Copy link
Collaborator

@RKSimon RKSimon Jul 25, 2024

Choose a reason for hiding this comment

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

ISD::ABS v32i16 is set to CUSTOM lowering in SIISelLowering.cpp so the cost used is LT.first * 2 - this assumption appears to be the cause of most regression in this patch series

Copy link
Contributor Author

@arsenm arsenm Jul 25, 2024

Choose a reason for hiding this comment

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

Most of the custom lowerings are working around the default legalization actions being to scalarize, rather than decompose into 2 x vectors

@arsenm arsenm force-pushed the users/arsenm/amdgpu-add-baseline-tti-cost-abs branch from 330c0e2 to df2b6b7 Compare July 25, 2024 17:27
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from ca78bfb to 85c14e0 Compare July 25, 2024 17:28
@arsenm arsenm changed the base branch from users/arsenm/amdgpu-add-baseline-tti-cost-abs to users/arsenm/tti-check-fptoi-sat-legalize-costs July 25, 2024 17:28
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 85c14e0 to 949edfe Compare July 25, 2024 17:40
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 8c88a8f to 8180759 Compare July 25, 2024 20:56
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 949edfe to 49db2b2 Compare July 25, 2024 20:57
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 8180759 to 45aed42 Compare July 26, 2024 20:00
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 49db2b2 to b448d7d Compare July 26, 2024 20:00
@arsenm arsenm mentioned this pull request Jul 26, 2024
Merged
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 45aed42 to 19f7331 Compare July 28, 2024 13:48
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from b448d7d to 6a73464 Compare July 28, 2024 13:48
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 19f7331 to a2900f1 Compare August 2, 2024 16:25
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 6a73464 to 5bb2cad Compare August 2, 2024 16:25
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from a2900f1 to 2746c43 Compare August 5, 2024 21:07
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 5bb2cad to 574affb Compare August 5, 2024 21:07
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 2746c43 to 88fe51a Compare August 6, 2024 18:27
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 574affb to bbd9d3d Compare August 6, 2024 18:27
RKSimon
RKSimon approved these changes Aug 7, 2024
View reviewed changes
Copy link
Collaborator

@RKSimon RKSimon left a comment

Choose a reason for hiding this comment

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

LGTM

@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 88fe51a to d428b60 Compare August 8, 2024 13:10
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from bbd9d3d to 65c4d58 Compare August 8, 2024 13:10
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from d428b60 to 50af0bb Compare August 8, 2024 19:56
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 65c4d58 to 9c21872 Compare August 8, 2024 19:56
@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from 50af0bb to b5f3942 Compare August 9, 2024 07:15
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 9c21872 to ef1f347 Compare August 9, 2024 07:15
@arsenm Graphite App
Copy link
Contributor Author

arsenm commented Aug 9, 2024 •
edited
Loading

Merge activity

  • Aug 9, 4:27 AM EDT: @arsenm started a stack merge that includes this pull request via Graphite.
  • Aug 9, 4:32 AM EDT: Graphite rebased this pull request as part of a merge.
  • Aug 9, 4:37 AM EDT: Graphite rebased this pull request as part of a merge.
  • Aug 9, 4:40 AM EDT: Graphite rebased this pull request as part of a merge.
  • Aug 9, 4:43 AM EDT: Graphite rebased this pull request as part of a merge.
  • Aug 9, 4:47 AM EDT: Graphite rebased this pull request as part of a merge.
  • Aug 9, 4:51 AM EDT: @arsenm merged this pull request with Graphite.

@arsenm arsenm force-pushed the users/arsenm/tti-check-fptoi-sat-legalize-costs branch from b5f3942 to cdac929 Compare August 9, 2024 08:28
Base automatically changed from users/arsenm/tti-check-fptoi-sat-legalize-costs to main August 9, 2024 08:32
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch 4 times, most recently from c653e64 to 7c7e93f Compare August 9, 2024 08:43
@arsenm
TTI: Check legalization cost of abs nodes
ea74c99 
Also adjust the AMDGPU cost.
@arsenm arsenm force-pushed the users/arsenm/tti-check-abs-legalize-costs branch from 7c7e93f to ea74c99 Compare August 9, 2024 08:47
@arsenm arsenm merged commit d7824fa into main Aug 9, 2024
@arsenm arsenm deleted the users/arsenm/tti-check-abs-legalize-costs branch August 9, 2024 08:51
shiltian pushed a commit that referenced this pull request Aug 9, 2024
@arsenm @shiltian
TTI: Check legalization cost of abs nodes (#100523)
39e4b4a
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@jayfoad jayfoad jayfoad left review comments

@RKSimon RKSimon RKSimon approved these changes

@alexey-bataev alexey-bataev Awaiting requested review from alexey-bataev

@nikic nikic Awaiting requested review from nikic

@rotateright rotateright Awaiting requested review from rotateright

@sdesmalen-arm sdesmalen-arm Awaiting requested review from sdesmalen-arm

Assignees

No one assigned

Labels

backend:AMDGPU llvm:analysis Includes value tracking, cost tables and constant folding llvm:transforms vectorizers

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

5 participants

@arsenm @llvmbot @jayfoad @RKSimon