This is an archive of the discontinued LLVM Phabricator instance.

[X86] LowerFunnelShift: prefer unpack-based algorithm
Needs ReviewPublic

Authored by Nekotekina on Apr 28 2023, 9:54 AM.

Details

Summary

Continuation of the https://reviews.llvm.org/D149071

Using unpack for splitting and using double-bitwidth shifts may improve performance according to https://uica.uops.info tests.

  1. No cross-lane shuffles
  2. No dirtying double-width registers

Diff Detail

Event Timeline

Nekotekina created this revision.Apr 28 2023, 9:54 AM
Herald added a project: Restricted Project. · View Herald TranscriptApr 28 2023, 9:54 AM
Herald added a subscriber: hiraditya. · View Herald Transcript
Nekotekina requested review of this revision.Apr 28 2023, 9:54 AM