[llvm] r326884 - [LoadStoreVectorizer] Differentiate between <1 x T> and T
Sven van Haastregt via llvm-commits
llvm-commits at lists.llvm.org
Wed Mar 7 02:29:29 PST 2018
Author: svenvh
Date: Wed Mar 7 02:29:28 2018
New Revision: 326884
URL: http://llvm.org/viewvc/llvm-project?rev=326884&view=rev
Log:
[LoadStoreVectorizer] Differentiate between <1 x T> and T
The LoadStoreVectorizer thought that <1 x T> and T were the same types
when merging stores, leading to a crash later.
Patch by Erik Hogeman.
Differential Revision: https://reviews.llvm.org/D44014
Added:
llvm/trunk/test/Transforms/LoadStoreVectorizer/X86/vector-scalar.ll
Modified:
llvm/trunk/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp
Modified: llvm/trunk/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp?rev=326884&r1=326883&r2=326884&view=diff
==============================================================================
--- llvm/trunk/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp (original)
+++ llvm/trunk/lib/Transforms/Vectorize/LoadStoreVectorizer.cpp Wed Mar 7 02:29:28 2018
@@ -316,6 +316,7 @@ bool Vectorizer::isConsecutiveAccess(Val
Type *PtrATy = PtrA->getType()->getPointerElementType();
Type *PtrBTy = PtrB->getType()->getPointerElementType();
if (PtrA == PtrB ||
+ PtrATy->isVectorTy() != PtrBTy->isVectorTy() ||
DL.getTypeStoreSize(PtrATy) != DL.getTypeStoreSize(PtrBTy) ||
DL.getTypeStoreSize(PtrATy->getScalarType()) !=
DL.getTypeStoreSize(PtrBTy->getScalarType()))
Added: llvm/trunk/test/Transforms/LoadStoreVectorizer/X86/vector-scalar.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/Transforms/LoadStoreVectorizer/X86/vector-scalar.ll?rev=326884&view=auto
==============================================================================
--- llvm/trunk/test/Transforms/LoadStoreVectorizer/X86/vector-scalar.ll (added)
+++ llvm/trunk/test/Transforms/LoadStoreVectorizer/X86/vector-scalar.ll Wed Mar 7 02:29:28 2018
@@ -0,0 +1,14 @@
+; RUN: opt -mtriple=x86_64-unknown-linux-gnu -load-store-vectorizer -mcpu haswell -S -o - %s | FileCheck %s
+
+; Check that the LoadStoreVectorizer does not crash due to not differentiating <1 x T> and T.
+
+; CHECK-LABEL: @vector_scalar(
+; CHECK: store double
+; CHECK: store <1 x double>
+define void @vector_scalar(double* %ptr, double %a, <1 x double> %b) {
+ %1 = bitcast double* %ptr to <1 x double>*
+ %2 = getelementptr <1 x double>, <1 x double>* %1, i32 1
+ store double %a, double* %ptr, align 8
+ store <1 x double> %b, <1 x double>* %2, align 8
+ ret void
+}
More information about the llvm-commits
mailing list