diff options
author | Alex Voicu <alexandru.voicu@amd.com> | 2025-09-29 22:50:15 +0100 |
---|---|---|
committer | GitHub <noreply@github.com> | 2025-09-29 22:50:15 +0100 |
commit | d481e5f9b7f4bde74fc4909a8a67bbd758991b33 (patch) | |
tree | 11a80b436ba2e8fb02a0b4b719ef24b657304a42 /llvm/lib/Support/VirtualFileSystem.cpp | |
parent | af5c1a696cffb48712526cc862dc1cafc9ce6e3c (diff) | |
download | llvm-d481e5f9b7f4bde74fc4909a8a67bbd758991b33.zip llvm-d481e5f9b7f4bde74fc4909a8a67bbd758991b33.tar.gz llvm-d481e5f9b7f4bde74fc4909a8a67bbd758991b33.tar.bz2 |
[AMDGPU][SPIRV] Use SPIR-V syncscopes for some AMDGCN BIs (#154867)
AMDGCN flavoured SPIR-V allows AMDGCN specific builtins, including those
for scoped fences and some specific RMWs. However, at present we don't
map syncscopes to their SPIR-V equivalents, but rather use the AMDGCN
ones. This ends up pessimising the resulting code as system scope is
used instead of device (agent) or subgroup (wavefront), so we correct
the behaviour, to ensure that we do the right thing during reverse
translation.
Diffstat (limited to 'llvm/lib/Support/VirtualFileSystem.cpp')
0 files changed, 0 insertions, 0 deletions