This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU: Minor cleanups to help with f16 support
ClosedPublic

Authored by arsenm on Jul 7 2015, 6:47 PM.

Details

Reviewers
tstellarAMD
Summary

The main change is inverting the condition for the
operand class classes so that VT.Size == 16 uses VGPR_32
instead of 64.

Diff Detail

Event Timeline

arsenm updated this revision to Diff 29231.Jul 7 2015, 6:47 PM
arsenm retitled this revision from to AMDGPU: Minor cleanups to help with f16 support .
arsenm updated this object.
arsenm added a subscriber: llvm-commits.
tstellarAMD accepted this revision.Aug 21 2015, 4:12 PM
tstellarAMD edited edge metadata.

LGTM.

This revision is now accepted and ready to land.Aug 21 2015, 4:12 PM
arsenm closed this revision.Aug 21 2015, 4:51 PM

r245764