This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Precommit test for i8 vector CopyToReg handling patch
ClosedPublic

Authored by jrbyrnes on Aug 31 2023, 11:31 AM.

Details

Summary

Adds test to show impact on cross block CopyToReg & CopyFromReg handling for n x i8, and shows NFC on CC

Diff Detail

Event Timeline

jrbyrnes created this revision.Aug 31 2023, 11:31 AM
Herald added a project: Restricted Project. · View Herald TranscriptAug 31 2023, 11:31 AM
jrbyrnes requested review of this revision.Aug 31 2023, 11:31 AM
Herald added a project: Restricted Project. · View Herald TranscriptAug 31 2023, 11:31 AM

function-args.ll and call-argument-types.ll seem to be missing all the i8 vectors

llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
3

don't need -verify-machineinstrs

5

add a v3 and v5 version

46–47

update_llc_test_checks is too dumb to deal with split argument lists across lines, can you move these

148

v16/v32?

arsenm requested changes to this revision.Sep 5 2023, 1:35 PM
This revision now requires changes to proceed.Sep 5 2023, 1:35 PM
jrbyrnes updated this revision to Diff 555945.Sep 5 2023, 3:03 PM
jrbyrnes marked 4 inline comments as done.

Address comments + Add tests & change test structure of vni8-across-blocks

arsenm added inline comments.Sep 12 2023, 3:19 AM
llvm/test/CodeGen/AMDGPU/gfx-callable-argument-types.ll
3181

use load null instead of load undef

3182

not sure if call returns are covered

arsenm accepted this revision.Sep 12 2023, 3:19 AM
This revision is now accepted and ready to land.Sep 12 2023, 3:19 AM
This revision was landed with ongoing or failed builds.Sep 13 2023, 11:27 AM
This revision was automatically updated to reflect the committed changes.
jrbyrnes marked 2 inline comments as done.