Page MenuHomePhabricator

[DA][TTI][AMDGPU] Add option to select GPUDA with TTI
ClosedPublic

Authored by kerbowa on Jan 20 2020, 8:06 AM.

Details

Summary

Enable the new diveregence analysis by default for AMDGPU.

Diff Detail

Event Timeline

kerbowa created this revision.Jan 20 2020, 8:06 AM
Herald added a project: Restricted Project. · View Herald TranscriptJan 20 2020, 8:06 AM
rampitec added inline comments.Jan 20 2020, 11:29 AM
llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
72

Do you need this new option if there is -use-gpu-divergence-analysis?

kerbowa marked an inline comment as done.Jan 20 2020, 11:40 AM
kerbowa added inline comments.
llvm/lib/Target/AMDGPU/AMDGPUTargetTransformInfo.cpp
72

The new cl option disables the new DA for AMDGPU. You can't do that with -use-gpu-divergence-analysis.

This revision is now accepted and ready to land.Jan 20 2020, 11:41 AM
This revision was automatically updated to reflect the committed changes.

Tests on Windows have been broken for over 12h now: http://lab.llvm.org:8011/builders/clang-x64-windows-msvc/builds/13808 I've reverted this for now in cd470717d15854fb778f2c96667ba36672af6cf5 . Sorry about the toil!