mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-11-26 04:32:44 +01:00
f14a122eeb
Fix PR39417, PR39497 The loop vectorizer may generate runtime SCEV checks for overflow and stride==1 cases, leading to execution of original scalar loop. The latter is forbidden when optimizing for size. An assert introduced in r344743 triggered the above PR's showing it does happen. This patch fixes this behavior by preventing vectorization in such cases. Differential Revision: https://reviews.llvm.org/D53612 llvm-svn: 345959 |
||
---|---|---|
.. | ||
CMakeLists.txt | ||
LLVMBuild.txt | ||
LoadStoreVectorizer.cpp | ||
LoopVectorizationLegality.cpp | ||
LoopVectorizationPlanner.h | ||
LoopVectorize.cpp | ||
SLPVectorizer.cpp | ||
Vectorize.cpp | ||
VPlan.cpp | ||
VPlan.h | ||
VPlanDominatorTree.h | ||
VPlanHCFGBuilder.cpp | ||
VPlanHCFGBuilder.h | ||
VPlanHCFGTransforms.cpp | ||
VPlanHCFGTransforms.h | ||
VPlanLoopInfo.h | ||
VPlanValue.h | ||
VPlanVerifier.cpp | ||
VPlanVerifier.h | ||
VPRecipeBuilder.h |