AMDGPU: Disable sincos fold for constant inputs #134579
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
The use list for ConstantData should not be used. We could do better by just constant folding here.
|
@llvm/pr-subscribers-backend-amdgpu Author: Matt Arsenault (arsenm) ChangesThe use list for ConstantData should not be used. We could do better Full diff: https://github.com/llvm/llvm-project/pull/134579.diff 3 Files Affected:
diff --git a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
index cf8b416d23e50..b65b4d67b3f8c 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPULibCalls.cpp
@@ -1361,6 +1361,11 @@ bool AMDGPULibCalls::fold_sincos(FPMathOperator *FPOp, IRBuilder<> &B,
bool const isSin = fInfo.getId() == AMDGPULibFunc::EI_SIN;
Value *CArgVal = FPOp->getOperand(0);
+
+ // TODO: Constant fold the call
+ if (isa<ConstantData>(CArgVal))
+ return false;
+
CallInst *CI = cast<CallInst>(FPOp);
Function *F = B.GetInsertBlock()->getParent();
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.defined.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.defined.ll
index ad7e913e6d22c..c8f45fe11390c 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.defined.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.defined.ll
@@ -125,12 +125,10 @@ define void @sincos_f32_value_is_same_constantfp(ptr addrspace(1) nocapture writ
; CHECK-LABEL: define void @sincos_f32_value_is_same_constantfp
; CHECK-SAME: (ptr addrspace(1) writeonly captures(none) [[SIN_OUT:%.*]], ptr addrspace(1) writeonly captures(none) [[COS_OUT:%.*]]) {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[__SINCOS_:%.*]] = alloca float, align 4, addrspace(5)
-; CHECK-NEXT: [[TMP0:%.*]] = call contract float @_Z6sincosfPU3AS5f(float 4.200000e+01, ptr addrspace(5) [[__SINCOS_]])
-; CHECK-NEXT: [[TMP1:%.*]] = load float, ptr addrspace(5) [[__SINCOS_]], align 4
-; CHECK-NEXT: store float [[TMP0]], ptr addrspace(1) [[SIN_OUT]], align 4
+; CHECK-NEXT: [[CALL:%.*]] = tail call contract float @_Z3sinf(float 4.200000e+01)
+; CHECK-NEXT: store float [[CALL]], ptr addrspace(1) [[SIN_OUT]], align 4
; CHECK-NEXT: [[CALL1:%.*]] = tail call contract float @_Z3cosf(float 4.200000e+01)
-; CHECK-NEXT: store float [[TMP1]], ptr addrspace(1) [[COS_OUT]], align 4
+; CHECK-NEXT: store float [[CALL1]], ptr addrspace(1) [[COS_OUT]], align 4
; CHECK-NEXT: ret void
;
entry:
diff --git a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll
index da5686972a86b..03b7d1646eb86 100644
--- a/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll
+++ b/llvm/test/CodeGen/AMDGPU/amdgpu-simplify-libcall-sincos.ll
@@ -745,13 +745,12 @@ entry:
define void @sincos_f32_value_is_same_constantfp(ptr addrspace(1) nocapture writeonly %sin_out, ptr addrspace(1) nocapture writeonly %cos_out) {
; CHECK-LABEL: define void @sincos_f32_value_is_same_constantfp
-; CHECK-SAME: (ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[SIN_OUT:%.*]], ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[COS_OUT:%.*]]) local_unnamed_addr #[[ATTR3]] {
+; CHECK-SAME: (ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[SIN_OUT:%.*]], ptr addrspace(1) writeonly captures(none) initializes((0, 4)) [[COS_OUT:%.*]]) local_unnamed_addr #[[ATTR2]] {
; CHECK-NEXT: entry:
-; CHECK-NEXT: [[__SINCOS_:%.*]] = alloca float, align 4, addrspace(5)
-; CHECK-NEXT: [[TMP0:%.*]] = call contract float @_Z6sincosfPU3AS5f(float 4.200000e+01, ptr addrspace(5) [[__SINCOS_]])
-; CHECK-NEXT: [[TMP1:%.*]] = load float, ptr addrspace(5) [[__SINCOS_]], align 4
-; CHECK-NEXT: store float [[TMP0]], ptr addrspace(1) [[SIN_OUT]], align 4
-; CHECK-NEXT: store float [[TMP1]], ptr addrspace(1) [[COS_OUT]], align 4
+; CHECK-NEXT: [[CALL:%.*]] = tail call contract float @_Z3sinf(float 4.200000e+01)
+; CHECK-NEXT: store float [[CALL]], ptr addrspace(1) [[SIN_OUT]], align 4
+; CHECK-NEXT: [[CALL1:%.*]] = tail call contract float @_Z3cosf(float 4.200000e+01)
+; CHECK-NEXT: store float [[CALL1]], ptr addrspace(1) [[COS_OUT]], align 4
; CHECK-NEXT: ret void
;
entry:
|
Pierre-vh
approved these changes
Apr 7, 2025
arsenm
deleted the
users/arsenm/amdgpu-skip-sincos-fold-for-constant-arguments
branch
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The use list for ConstantData should not be used. We could do better
by just constant folding here.