This is an archive of the discontinued LLVM Phabricator instance.

[X86][AVX512BW] support in byte shift and SAD
ClosedPublic

Authored by AsafBadouh on Aug 30 2015, 4:26 AM.

Diff Detail

Repository
rL LLVM

Event Timeline

AsafBadouh updated this revision to Diff 33533.Aug 30 2015, 4:26 AM
AsafBadouh retitled this revision from to [X86][AVX512BW] support in byte shift and SAD.
AsafBadouh updated this object.
AsafBadouh added reviewers: delena, igorb.
AsafBadouh added a subscriber: llvm-commits.
delena added inline comments.Aug 30 2015, 4:58 AM
../llvmTmp/lib/Target/X86/X86InstrAVX512.td
6788 ↗(On Diff #33533)

While DAG lowering, we build SHIFT nodes with Imm8. I suppose that Imm32 will not work in this case.
Could you, please, add IR tests for these nodes?

AsafBadouh updated this revision to Diff 33580.Aug 31 2015, 7:40 AM

apply Elena's comments

delena edited edge metadata.Sep 1 2015, 1:58 AM

Please apply the comments and you can commit the patch.

../llvmTmp/lib/Target/X86/X86InstrSSE.td
4187 ↗(On Diff #33580)

The line is too long.

../llvmTmp/test/CodeGen/X86/avx512bwvl-intrinsics.ll
4197 ↗(On Diff #33580)

This file is for intrinsics. Put in avx-isa-check.ll.

This revision was automatically updated to reflect the committed changes.