projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: cbd93ce) | patch
[AMDGPU] Skip invariant loads when avoiding WAR conflicts
authorPiotr Sobczak <Piotr.Sobczak@amd.com>
Wed, 12 May 2021 07:23:59 +0000 (09:23 +0200)
committerPiotr Sobczak <Piotr.Sobczak@amd.com>
Wed, 12 May 2021 08:57:05 +0000 (10:57 +0200)
commit68137ef5682f936a5db14202a69548eee6294256
treeb2dd474e417eddbc807082edeebdc8c4a9cbc182tree | snapshot
parentcbd93cee9bf014402a7405479ba21f6f3340a126commit | diff
[AMDGPU] Skip invariant loads when avoiding WAR conflicts

No need to handle invariant loads when avoiding WAR conflicts, as
there cannot be a vector store to the same memory location.

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D101177
llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/mul24-pass-ordering.ll diff | blob | history
llvm/test/CodeGen/AMDGPU/stack-realign.ll diff | blob | history
llvm/test/CodeGen/AMDGPU/waitcnt-no-redundant.mir diff | blob | history
Domain: System / Toolchain;