This is an archive of the discontinued LLVM Phabricator instance.

AMDGPU : Add intrinsic for instruction v_cvt_pk_u8_f32
ClosedPublic

Authored by wdng on Aug 9 2016, 3:27 PM.

Diff Detail

Repository
rL LLVM

Event Timeline

wdng updated this revision to Diff 67423.Aug 9 2016, 3:27 PM
wdng retitled this revision from to AMDGPU : Add intrinsic for instruction v_cvt_pk_u8_f32.
wdng updated this object.
wdng added reviewers: arsenm, tstellarAMD.
wdng set the repository for this revision to rL LLVM.
arsenm edited edge metadata.Aug 10 2016, 12:13 PM

There should probably be constant folding for this put into InstCombineCalls

include/llvm/IR/IntrinsicsAMDGPU.td
410

Remove the name comment

test/CodeGen/AMDGPU/v_cvt_pk_u8_f32.ll
9–13

You should have a test for 0..3

wdng updated this revision to Diff 67613.Aug 10 2016, 2:55 PM
wdng edited edge metadata.
wdng marked 2 inline comments as done.
  1. Remove comments
  2. Add 0...3 LIT tests and one test for all.

Just a few minor comments, with those changes LGTM.

include/llvm/IR/IntrinsicsAMDGPU.td
415–419

This seems like an unrelated whitespace change.

lib/Target/AMDGPU/SIInstructions.td
1568

Unnecessary whitespace change

test/CodeGen/AMDGPU/v_cvt_pk_u8_f32.ll
61

We don't need the #1 attribute on these test functions.

This revision was automatically updated to reflect the committed changes.