This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU/GlobalISel: Mark 32-bit G_FADD as legal
ClosedPublic

Authored by tstellar on Sep 29 2017, 8:17 PM.

Diff Detail

Repository
rL LLVM

Event Timeline

tstellar created this revision.Sep 29 2017, 8:17 PM
kristof.beyls added inline comments.Oct 2 2017, 2:13 AM
test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir
1 ↗(On Diff #117252)

I just find "-mcpu=fiji" a bit strange for this test - I'm a bit surprised an -mcpu would be needed here.
Otherwise, without knowing hardly anything about the AMDGPU target, the rest of this patch looks fine to me.

arsenm accepted this revision.Oct 3 2017, 3:58 PM

LGTM

This revision is now accepted and ready to land.Oct 3 2017, 3:58 PM
tstellar added inline comments.Oct 24 2017, 11:46 AM
test/CodeGen/AMDGPU/GlobalISel/legalize-fadd.mir
1 ↗(On Diff #117252)

It's not necessary, I will drop it before I commit.

This revision was automatically updated to reflect the committed changes.