This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Switch AnnotateUniformValues to MemorySSA
ClosedPublic

Authored by rampitec on May 5 2021, 5:05 PM.

Details

Summary

This shall speedup compilation and also remove threshold
limitations used by memory dependency analysis.

It also seem to fix the bug in the coalescer_remat.ll
where an SMRD load was used in presence of a potentially
clobbering store.

Fixes: SWDEV-272132

Diff Detail

Event Timeline

rampitec created this revision.May 5 2021, 5:05 PM
rampitec requested review of this revision.May 5 2021, 5:05 PM
Herald added a project: Restricted Project. · View Herald TranscriptMay 5 2021, 5:05 PM
Herald added a subscriber: wdng. · View Herald Transcript
arsenm accepted this revision.May 5 2021, 5:08 PM
This revision is now accepted and ready to land.May 5 2021, 5:08 PM
rampitec updated this revision to Diff 343254.May 5 2021, 6:03 PM
rampitec edited the summary of this revision. (Show Details)
  • Drop now unused dfs include.
  • Forgot to add llc-pipeline test.
  • Make clang-tidy happy.
arsenm accepted this revision.May 5 2021, 6:26 PM
rampitec updated this revision to Diff 343263.May 5 2021, 6:37 PM
  • SetVector include also not needed anymore.
  • MSSAWalker is not always needed and cached by the MSSA anyway, so request it only when needed.

PSDB passed, I am about to submit it after check-llvm.

This revision was landed with ongoing or failed builds.May 5 2021, 6:47 PM
This revision was automatically updated to reflect the committed changes.