diff options
| author | Nicolai Hähnle <nicolai.haehnle@amd.com> | 2025-08-23 00:07:59 -0700 |
|---|---|---|
| committer | GitHub <noreply@github.com> | 2025-08-23 00:07:59 -0700 |
| commit | 46762421c30a361c439ad5930f1fd026601db7f5 (patch) | |
| tree | a413b64f076ffd0c5255aeed500c3902a41f449b /llvm/lib/Transforms/Utils/SimplifyCFG.cpp | |
| parent | 304ef65f7bbc575b110ba3c7f3ae01c3d8f027e6 (diff) | |
| download | llvm-46762421c30a361c439ad5930f1fd026601db7f5.zip llvm-46762421c30a361c439ad5930f1fd026601db7f5.tar.gz llvm-46762421c30a361c439ad5930f1fd026601db7f5.tar.bz2 | |
AMDGPU/GFX12: Do not wait unnecessarily before barriers (#154970)
The barrier intrinsic itself should not have memory semantics. Frontends
should use appropriate fence instructions for memory effects, and some
frontends want to rely on that for performance (e.g. wait only for LDS
before a barrier).
See the code comment for more detail.
Diffstat (limited to 'llvm/lib/Transforms/Utils/SimplifyCFG.cpp')
0 files changed, 0 insertions, 0 deletions