This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Fix shouldClusterMemOps to process flat loads
ClosedPublic

Authored by rampitec on Sep 5 2017, 9:41 PM.

Diff Detail

Repository
rL LLVM

Event Timeline

rampitec created this revision.Sep 5 2017, 9:41 PM
vpykhtin accepted this revision.Sep 6 2017, 7:22 AM

LGTM. However it looks like we should fix TD to follow single naming convention.

This revision is now accepted and ready to land.Sep 6 2017, 7:22 AM

LGTM. However it looks like we should fix TD to follow single naming convention.

vdata is for store, vdst is for load, so convention seems fine.

arsenm added inline comments.Sep 6 2017, 8:27 AM
lib/Target/AMDGPU/SIInstrInfo.cpp
368–369 ↗(On Diff #113954)

I think falling back like this is more hazardous than checking if it's flat first

This revision was automatically updated to reflect the committed changes.