Page MenuHomePhabricator

AMDGPU: Fix inconsistent lowering of select of vectors

Authored by arsenm on May 24 2016, 11:33 AM.



f32 vectors would use a sequence of BFI instructions instead
of unrolled cmp + select. This was better in the case of a VALU
select with SGPR inputs, but we don't have a way of dealing with that
in the DAG.

Diff Detail

Event Timeline

arsenm updated this revision to Diff 58278.May 24 2016, 11:33 AM
arsenm retitled this revision from to AMDGPU: Fix inconsistent lowering of select of vectors.
arsenm updated this object.
arsenm added a reviewer: tstellarAMD.
arsenm added a subscriber: llvm-commits.
tstellarAMD accepted this revision.May 25 2016, 9:03 AM
tstellarAMD edited edge metadata.


This revision is now accepted and ready to land.May 25 2016, 9:03 AM
arsenm closed this revision.May 25 2016, 10:41 AM