[WebAssembly][TTI] Avoid crash when costing scalable vector shifts - #212759
Merged
Conversation
llvmorg-github-actions
Bot
added
backend:WebAssembly
llvm:analysis
Member
Author
|
Basically We should restrict the WebAssembly-specific shift cost calculation to |
|
@llvm/pr-subscribers-backend-webassembly @llvm/pr-subscribers-llvm-analysis Author: Anutosh Bhat (anutosh491) ChangesI see the following Full diff: https://github.com/llvm/llvm-project/pull/212759.diff 2 Files Affected:
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
index 7d1136bb9beba..75c631055ae2d 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyTargetTransformInfo.cpp
@@ -82,7 +82,7 @@ InstructionCost WebAssemblyTTIImpl::getArithmeticInstrCost(
BasicTTIImplBase<WebAssemblyTTIImpl>::getArithmeticInstrCost(
Opcode, Ty, CostKind, Op1Info, Op2Info);
- if (auto *VTy = dyn_cast<VectorType>(Ty)) {
+ if (auto *VTy = dyn_cast<FixedVectorType>(Ty)) {
switch (Opcode) {
case Instruction::LShr:
case Instruction::AShr:
@@ -92,7 +92,7 @@ InstructionCost WebAssemblyTTIImpl::getArithmeticInstrCost(
// approximation.
if (!Op2Info.isUniform())
Cost =
- cast<FixedVectorType>(VTy)->getNumElements() *
+ VTy->getNumElements() *
(TargetTransformInfo::TCC_Basic +
getArithmeticInstrCost(Opcode, VTy->getElementType(), CostKind) +
TargetTransformInfo::TCC_Basic);
diff --git a/llvm/test/Analysis/CostModel/WebAssembly/vector-shift.ll b/llvm/test/Analysis/CostModel/WebAssembly/vector-shift.ll
new file mode 100644
index 0000000000000..97f2a94b7b460
--- /dev/null
+++ b/llvm/test/Analysis/CostModel/WebAssembly/vector-shift.ll
@@ -0,0 +1,42 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 6
+; RUN: opt -mtriple=wasm32-unknown-unknown -mattr=+simd128 \
+; RUN: -passes='print<cost-model>' -disable-output < %s 2>&1 | FileCheck %s
+
+define void @fixed_non_uniform(<4 x i32> %x, <4 x i32> %amount) {
+; CHECK-LABEL: 'fixed_non_uniform'
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %shl = shl <4 x i32> %x, %amount
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %lshr = lshr <4 x i32> %x, %amount
+; CHECK-NEXT: Cost Model: Found an estimated cost of 12 for instruction: %ashr = ashr <4 x i32> %x, %amount
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+ %shl = shl <4 x i32> %x, %amount
+ %lshr = lshr <4 x i32> %x, %amount
+ %ashr = ashr <4 x i32> %x, %amount
+ ret void
+}
+
+define void @scalable_uniform(<vscale x 4 x i32> %x) {
+; CHECK-LABEL: 'scalable_uniform'
+; CHECK-NEXT: Cost Model: Invalid cost for instruction: %shl = shl <vscale x 4 x i32> %x, splat (i32 1)
+; CHECK-NEXT: Cost Model: Invalid cost for instruction: %lshr = lshr <vscale x 4 x i32> %x, splat (i32 1)
+; CHECK-NEXT: Cost Model: Invalid cost for instruction: %ashr = ashr <vscale x 4 x i32> %x, splat (i32 1)
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+ %shl = shl <vscale x 4 x i32> %x, splat (i32 1)
+ %lshr = lshr <vscale x 4 x i32> %x, splat (i32 1)
+ %ashr = ashr <vscale x 4 x i32> %x, splat (i32 1)
+ ret void
+}
+
+define void @scalable_non_uniform(<vscale x 4 x i32> %x, <vscale x 4 x i32> %amount) {
+; CHECK-LABEL: 'scalable_non_uniform'
+; CHECK-NEXT: Cost Model: Invalid cost for instruction: %shl = shl <vscale x 4 x i32> %x, %amount
+; CHECK-NEXT: Cost Model: Invalid cost for instruction: %lshr = lshr <vscale x 4 x i32> %x, %amount
+; CHECK-NEXT: Cost Model: Invalid cost for instruction: %ashr = ashr <vscale x 4 x i32> %x, %amount
+; CHECK-NEXT: Cost Model: Found an estimated cost of 1 for instruction: ret void
+;
+ %shl = shl <vscale x 4 x i32> %x, %amount
+ %lshr = lshr <vscale x 4 x i32> %x, %amount
+ %ashr = ashr <vscale x 4 x i32> %x, %amount
+ ret void
+}
|
aheejin
approved these changes
Jul 29, 2026
aheejin
left a comment
aheejin
left a comment
Member
There was a problem hiding this comment.
Not crashing here sounds fine, but we don't support compile scalable vectors anyway, no?
Member
Author
Yess ! Just a corner case I came across. Would help exit the scalable case gracefully. |
Member
Author
|
Thanks for the approval. Enabling auto merge. |
anutosh491
enabled auto-merge (squash)
July 30, 2026 04:01
anutosh491
disabled auto-merge
July 30, 2026 07:45
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
I see the following