[PATCH] D104911: [OpenMP] Match initial thread pattern on AMDGPU

[Joseph Huber via Phabricator via llvm-commits]

jhuber6 created this revision.
jhuber6 added reviewers: jdoerfert, JonChesterfield.
Herald added subscribers: ormris, guansong, hiraditya, t-tye, tpr, dstuttard, yaxunl, kzhuravl.
jhuber6 requested review of this revision.
Herald added subscribers: llvm-commits, sstefan1, wdng.
Herald added a project: LLVM.

The AAExecutionDomain pass used to push globalized memory calls to
global shared memory doesn't match the pattern AMDGPU generates. This
means the optimizations won't work on anything other than an NVPTX
target. This patch adds AMDGPU's pattern to the check.

Depends on D102423 <https://reviews.llvm.org/D102423>


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D104911

Files:
  llvm/lib/Transforms/IPO/OpenMPOpt.cpp
  llvm/test/Transforms/OpenMP/replace_globalization.ll

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D104911.354484.patch
Type: text/x-patch
Size: 4217 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20210625/1ec14cac/attachment.bin>