[PATCH] D129537: [X86][DAGISel] Don't widen shuffle element with AVX512

[LuoYuanke via Phabricator via llvm-commits]

LuoYuanke added inline comments.


================
Comment at: llvm/test/CodeGen/X86/avx512-shuffles/shuffle-blend.ll:169
+; AVX512BW-NEXT:    vpsubb %xmm1, %xmm0, %xmm2 {%k1}
+; AVX512BW-NEXT:    vmovdqa %xmm2, %xmm0
+; AVX512BW-NEXT:    ret{{[l|q]}}
----------------
RKSimon wrote:
> regression? you might need to improve the 128-bit limit logic to account for vXi16 specifically
There is PBLENDW, but there is no PBLENDB instruction, so it is better to widen to v2Xi16 from vXi8. Besides there is more instruction for 16-bit element (e.g., movsh). I'll investigate more on this issue.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D129537/new/

https://reviews.llvm.org/D129537