Mehdi Amini 2b72a53db9 Fix LoopLoadElimination to keep original alignment on the inital hoisted store ...

This is fixing a bug where Loop Vectorization is widening a load but
with a lower alignment. Hoisting the load without propagating the alignment
will allow inst-combine to later deduce a higher alignment that what the pointer
actually is.

Differential Revision: https://reviews.llvm.org/D28408

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@291281 91177308-0d34-0410-b5e6-96231b3b80d8

2017-01-06 21:06:51 +00:00

..

backward.ll

…

cond-load.ll

[LLE] Don't hoist conditionally executed loads

2016-06-28 04:02:47 +00:00

def-store-before-load.ll

…

forward.ll

Fix LoopLoadElimination to keep original alignment on the inital hoisted store

2017-01-06 21:06:51 +00:00

loop-simplify-dep.ll

[LLE] Add missed LoopSimplify dependence

2016-03-10 23:54:39 +00:00

memcheck.ll

Allow LLE/LD and the loop versioning infrastructure to use SCEV predicates

2015-11-09 13:26:09 +00:00

multiple-stores-same-block.ll

…

non-consecutive.ll

[LLE] Add missing check for unit stride

2016-03-09 20:47:55 +00:00

opt-size.ll

[LoopLoadElim] Don't allow versioning when optForSize

2016-02-05 01:14:05 +00:00

symbolic-stride.ll

[LAA] Enable symbolic stride speculation for all LAA clients

2016-06-17 22:35:41 +00:00

type-mismatch.ll

[LLE] Check for mismatching types between the store and the load earlier

2016-03-24 17:59:26 +00:00

unknown-dep.ll

…