Page MenuHomePhabricator

[DAGCombiner] Call SimplifyDemandedBits to simplify EXTRACT_VECTOR_ELT
Needs ReviewPublic

Authored by foad on Sep 30 2020, 7:12 AM.

Diff Detail

Event Timeline

foad created this revision.Sep 30 2020, 7:12 AM
Herald added a project: Restricted Project. · View Herald TranscriptSep 30 2020, 7:12 AM
foad requested review of this revision.Sep 30 2020, 7:12 AM
foad added inline comments.Sep 30 2020, 7:19 AM
llvm/test/CodeGen/Mips/cconv/vector.ll
963–1024

There are some regressions in this file but also some improvements. I haven't worked out what's going on yet.

llvm/test/CodeGen/Thumb2/mve-vecreduce-mla.ll
219–229

Regression here and in other cases that are now using muls instead of umull/umlal.

llvm/test/CodeGen/X86/vector-fshl-128.ll
188–190

Regression. Quite a few tests are now using pxor+punpckhdq instead of pshufd. I wonder if some kind of combine could spot this case and turn it back into pshufd.

foad added inline comments.Sep 30 2020, 7:34 AM
llvm/test/CodeGen/ARM/dagcombine-anyexttozeroext.ll
46–58

Regression.

llvm/test/CodeGen/ARM/vdup.ll
59–69

Regression in lots of cases in this file.

foad updated this revision to Diff 295595.Oct 1 2020, 9:15 AM

Rebase on D88570.

foad updated this revision to Diff 296629.Oct 7 2020, 2:57 AM

Rebase.

RKSimon added inline comments.Oct 7 2020, 3:44 AM
llvm/test/CodeGen/AMDGPU/sdiv64.ll
502

Is this a regression? It looks like we're lost track that we only need 1 element

llvm/test/CodeGen/ARM/func-argpassing-endian.ll
106

regression? we're no longer doing a single multiple load

llvm/test/CodeGen/ARM/vdup.ll
59–69

poor duplicate/splat detection?

@foad Any update on this?

yubing added a subscriber: yubing.Fri, Dec 4, 4:32 AM