Page MenuHomePhabricator

[AMDGPU] Preliminary patch for divergence driven instruction selection. Load offset inlining pattern changed.
ClosedPublic

Authored by alex-t on Sep 12 2018, 4:19 AM.

Details

Reviewers
rampitec
Summary

With new selection both I32 and U32 may be selected.
V_ADD_U32 added to the pattern.

Diff Detail

Event Timeline

alex-t created this revision.Sep 12 2018, 4:19 AM
This revision is now accepted and ready to land.Sep 12 2018, 12:06 PM
arsenm added inline comments.Sep 12 2018, 8:16 PM
lib/Target/AMDGPU/SIInstrInfo.cpp
4086–4087

I thought we always just selected one of these, and one was just the alias to deal with the name change problems

alex-t set the repository for this revision to rL LLVM.Sep 13 2018, 5:07 AM
alex-t closed this revision.Sep 13 2018, 5:51 AM