This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU/GlobalISel: Select G_SHL
ClosedPublic

Authored by arsenm on Jul 15 2019, 5:54 AM.

Details

Summary

I think this manages to not break the DAG handling with the divergent
predicates because the stadalone divergent patterns end up with a
higher priority than the pattern on the instruction definition.

The 16-bit versions don't work yet.

Diff Detail

Event Timeline

arsenm created this revision.Jul 15 2019, 5:54 AM

This looks OK to me, but I'm not as familiar with how this will impact SDAG. Might want to have someone else take a look too.

nhaehnle accepted this revision.Jul 16 2019, 2:42 AM

Yeah, let's take this.

This revision is now accepted and ready to land.Jul 16 2019, 2:42 AM
arsenm closed this revision.Jul 16 2019, 1:15 PM

r366254, though irritatingly I had to split the tests to not fail in a release build