projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: d242b52) | patch
[AMDGPU][Uniformity] V_MBCNT* is never uniform
authorCarl Ritson <carl.ritson@amd.com>
Thu, 18 May 2023 04:27:58 +0000 (13:27 +0900)
committerCarl Ritson <carl.ritson@amd.com>
Thu, 18 May 2023 04:50:12 +0000 (13:50 +0900)
commit9602c7a0817f34fb4d01194119fe4b75c00ea5ed
tree1b7ad7297e6f4f8445d46b02daec67da768b3be0tree | snapshot
parentd242b52d0e477171f3f6b957bafece88edb88a4fcommit | diff
[AMDGPU][Uniformity] V_MBCNT* is never uniform

Mark V_MBCNT instructions add thread/lane position so will never
be uniform.

Reviewed By: foad

Differential Revision: https://reviews.llvm.org/D150759
llvm/lib/Target/AMDGPU/VOP2Instructions.td diff | blob | history
llvm/test/Analysis/UniformityAnalysis/AMDGPU/MIR/never-uniform.mir diff | blob | history
Domain: System / Toolchain;