This is an archive of the discontinued LLVM Phabricator instance.

[AMDGPU] Add v_mov_b64 gfx940 opcode
ClosedPublic

Authored by rampitec on Mar 4 2022, 1:45 PM.

Diff Detail

Event Timeline

rampitec created this revision.Mar 4 2022, 1:45 PM
Herald added a project: Restricted Project. · View Herald TranscriptMar 4 2022, 1:45 PM
rampitec requested review of this revision.Mar 4 2022, 1:45 PM
Herald added a project: Restricted Project. · View Herald TranscriptMar 4 2022, 1:45 PM
Herald added a subscriber: wdng. · View Herald Transcript
foad accepted this revision.Mar 7 2022, 3:02 AM

Looks OK to me. I have never really understood why we have duplicate logic for expanding 64-bit vgpr copies in both copyToPhysReg and expandPostRAPseudo.

llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
1846

I think the setDesc should logically be inside this "if", but I guess it doesn't make any difference in practice.

This revision is now accepted and ready to land.Mar 7 2022, 3:02 AM

Looks OK to me. I have never really understood why we have duplicate logic for expanding 64-bit vgpr copies in both copyToPhysReg and expandPostRAPseudo.

One is needed for phys regs, the other was mostly created to handle and fold immediate moves (even though works with registers). So the logic isn't really duplicated, copyToPhysReg does not have immediate handling logic.

This revision was landed with ongoing or failed builds.Mar 7 2022, 12:07 PM
This revision was automatically updated to reflect the committed changes.