Matt Arsenault c74d601f15 AMDGPU: Don't report 2-byte alignment as fast ... This is apparently worse than 1-byte alignment. This does not attempt to decompose 2-byte aligned wide stores, but will stop trying to produce them. Also fix bug in LoadStoreVectorizer which was decreasing the alignment and vectorizing stack accesses. It was assuming a stack object was an alloca that could have its base alignment changed, which is not true if the pointer is derived from a function argument.		2020-02-11 18:35:00 -05:00
..
CMakeLists.txt	[VectorCombine] new IR transform pass for partial vector ops	2020-02-09 10:04:41 -05:00
LLVMBuild.txt
LoadStoreVectorizer.cpp	AMDGPU: Don't report 2-byte alignment as fast	2020-02-11 18:35:00 -05:00
LoopVectorizationLegality.cpp
LoopVectorizationPlanner.h	[LV] Do not try to sink dead instructions.	2020-01-28 08:28:03 -08:00
LoopVectorize.cpp	[LV] Hoist code to mark conditional assumes as dead to caller (NFC).	2020-01-28 08:50:44 -08:00
SLPVectorizer.cpp	[Alignement][NFC] Deprecate untyped CreateAlignedLoad	2020-01-23 13:34:32 +01:00
VectorCombine.cpp	[VectorCombine] remove unused debug counter; NFC	2020-02-11 08:24:07 -05:00
Vectorize.cpp	[VectorCombine] new IR transform pass for partial vector ops	2020-02-09 10:04:41 -05:00
VPlan.cpp
VPlan.h
VPlanDominatorTree.h
VPlanHCFGBuilder.cpp
VPlanHCFGBuilder.h
VPlanLoopInfo.h
VPlanPredicator.cpp
VPlanPredicator.h
VPlanSLP.cpp
VPlanTransforms.cpp
VPlanTransforms.h
VPlanValue.h
VPlanVerifier.cpp
VPlanVerifier.h
VPRecipeBuilder.h