This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU/SI: Handle wait states required for DPP instructions
ClosedPublic

Authored by tstellarAMD on Feb 23 2016, 8:23 AM.

Diff Detail

Repository
rL LLVM

Event Timeline

tstellarAMD retitled this revision from to AMDGPU/SI: Handle wait states required for DPP instructions.
tstellarAMD updated this object.
tstellarAMD added a reviewer: arsenm.
tstellarAMD added a subscriber: llvm-commits.

Fix case where dpp instruction was first in block.

arsenm added inline comments.Feb 23 2016, 6:53 PM
lib/Target/AMDGPU/SIInsertWaits.cpp
507 ↗(On Diff #48876)

Typo strt

509–510 ↗(On Diff #48876)

You should be able to use range loop with predecessors()

516 ↗(On Diff #48876)

There could be a break here since there will only ever be one fall through block, although it doesn't really matter because with structured control flow there is at most one predecessor

tstellarAMD marked 3 inline comments as done.

Fix typo. Use range for loop. Simplify loop.

arsenm accepted this revision.Mar 4 2016, 10:34 PM
arsenm edited edge metadata.

LGTM

This revision is now accepted and ready to land.Mar 4 2016, 10:34 PM
This revision was automatically updated to reflect the committed changes.