projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: e574833) | patch
[AMDGPU] Handle memset users in PromoteAlloca
authorpvanhout <pierre.vanhoutryve@amd.com>
Thu, 16 Mar 2023 13:35:27 +0000 (14:35 +0100)
committerpvanhout <pierre.vanhoutryve@amd.com>
Tue, 28 Mar 2023 13:01:55 +0000 (15:01 +0200)
commitd7b4b76956e2da8b251c4e6121624477e1491211
tree6edb46b3975efc73ffdd4cc404efe5331f3b093ftree | snapshot
parente574833c2bc46a39150156eb973d0efb142bc618commit | diff
[AMDGPU] Handle memset users in PromoteAlloca

Allows allocas with memset users to be promoted.

This is intended to prevent patterns such as `memset(&alloca, 0, sizeof(alloca))` (which I think can be emitted by frontends) from preventing a vectorization of allocas.

Fixes SWDEV-388784

Reviewed By: arsenm

Differential Revision: https://reviews.llvm.org/D146225
llvm/lib/Target/AMDGPU/AMDGPUPromoteAlloca.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/promote-alloca-memset.ll [new file with mode: 0644] blob
Domain: System / Toolchain;