projects / oota-llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d819fac) | patch
DAGCombiner: Assume invariant load cannot alias a store
authorMatt Arsenault <Matthew.Arsenault@amd.com>
Fri, 10 Jul 2015 22:17:40 +0000 (22:17 +0000)
committerMatt Arsenault <Matthew.Arsenault@amd.com>
Fri, 10 Jul 2015 22:17:40 +0000 (22:17 +0000)
commitee6d5d1c9e67ac32b9904b18c5f55b98fdd7357b
treee741ea254e0d1683cb562c5fd0bd25f5e307273etree | snapshot
parentd819fac2945672f211b0ef8fdedffa2e124b46e2commit | diff
DAGCombiner: Assume invariant load cannot alias a store

The motivation is to allow GatherAllAliases / FindBetterChain
to not give up on dependent loads of a pointer from constant memory.

This is important for AMDGPU, because most loads are pointers
derived from a load of a kernel argument from constant memory.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@241948 91177308-0d34-0410-b5e6-96231b3b80d8
lib/CodeGen/SelectionDAG/DAGCombiner.cpp diff | blob | history
test/CodeGen/AMDGPU/invariant-load-no-alias-store.ll [new file with mode: 0644] blob
C/C++ LLVM-based compilers that forbids OOTA behaviors