projects / platform / upstream / llvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: c68b469) | patch
[AMDGPU] Fix hazard handling of v_cmpx to permlane
authorStanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
Wed, 8 Jun 2022 20:42:28 +0000 (13:42 -0700)
committerStanislav Mekhanoshin <Stanislav.Mekhanoshin@amd.com>
Thu, 9 Jun 2022 17:33:54 +0000 (10:33 -0700)
commit5c974d086c22d519e3e649f1d5fe213b03d6ba25
tree043d20e4bb1468278bd38f57af36b5c69083b2e3tree | snapshot
parentc68b469e07884bc1dd69fb946af61ae476ac139bcommit | diff
[AMDGPU] Fix hazard handling of v_cmpx to permlane

- VOP3 and SDWA forms of V_CMPX were not handled
- Hazard only exists if the compare defines EXEC (i.e. V_CMPX)
  forwarded to the permlane.

Differential Revision: https://reviews.llvm.org/D127344
llvm/lib/Target/AMDGPU/GCNHazardRecognizer.cpp diff | blob | history
llvm/test/CodeGen/AMDGPU/vcmpx-permlane-hazard-sdwa.mir [new file with mode: 0644] blob
llvm/test/CodeGen/AMDGPU/vcmpx-permlane-hazard.mir diff | blob | history
Domain: System / Toolchain;