[PATCH] D91515: [X86] Use GF2P8AFFINEQB to implement vector bitreverse.

Tue Nov 17 04:04:34 PST 2020

RKSimon accepted this revision.
RKSimon added a comment.
This revision is now accepted and ready to land.

LGTM - I'm not sure whether its worth supporting this for pre-SSSE3 but technically its possible (and we already do this a lot for recent AVX512 features on nonVLX fictional hardware).

================
Comment at: llvm/test/Analysis/CostModel/X86/bitreverse.ll:21
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+gfni,+avx512f | FileCheck %s -check-prefixes=GFNIAVX512F
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -cost-model -analyze -mattr=+gfni,+avx512bw | FileCheck %s -check-prefixes=GFNIAVX512BW

----------------
These can all probably have a GFNI 'common' prefix to reduce some duplication

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D91515/new/

https://reviews.llvm.org/D91515