This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Promote alloca to vector in opt
ClosedPublic

Authored by rampitec on May 21 2020, 10:10 AM.

Details

Summary

Promote alloca to vector before SROA and loop unroll. If we manage
to eliminate allocas before unroll we may choose to unroll less.

Diff Detail

Event Timeline

rampitec created this revision.May 21 2020, 10:10 AM
Herald added a project: Restricted Project. · View Herald TranscriptMay 21 2020, 10:10 AM
arsenm added inline comments.May 21 2020, 10:45 AM
llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
486

I think we should start adding the pass pipeline tests like other targets have. Without looking I'm not 100% sure how many times this runs, or before which SROA runs

rampitec updated this revision to Diff 265559.May 21 2020, 11:37 AM
rampitec marked an inline comment as done.

Added opt pipeline test.

If you want I can pre-commit it.

If you want I can pre-commit it.

Yes, that would make most sense

rampitec updated this revision to Diff 265571.May 21 2020, 12:10 PM

Pre-commited opt pipeline test and rebased.

arsenm accepted this revision.May 21 2020, 1:18 PM
This revision is now accepted and ready to land.May 21 2020, 1:18 PM
This revision was automatically updated to reflect the committed changes.