This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Enable divergence driven ISel for ADD/SUB i64
ClosedPublic

Authored by alex-t on Mar 18 2020, 10:09 AM.

Details

Summary

Currently we custom select add/sub with carry out to scalar form relying on later replacing them to vector form if necessary.
This change enables custom selection code to take the divergence of adde/addc SDNodes into account and select the appropriate form in one step.

Diff Detail

Event Timeline

alex-t created this revision.Mar 18 2020, 10:09 AM
arsenm added inline comments.Mar 18 2020, 10:22 AM
llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
1020–1027

static const

alex-t updated this revision to Diff 251427.Mar 19 2020, 11:37 AM

static const

alex-t marked an inline comment as done.Mar 19 2020, 11:38 AM
arsenm accepted this revision.Mar 19 2020, 12:07 PM
This revision is now accepted and ready to land.Mar 19 2020, 12:07 PM
vpykhtin accepted this revision.Mar 20 2020, 1:02 AM

LGTM.

This revision was automatically updated to reflect the committed changes.
Herald added a project: Restricted Project. · View Herald TranscriptMar 20 2020, 7:33 AM