[PATCH] D57059: [SLP] Initial support for the vectorization of the non-power-of-2 vectors.

[Alexey Bataev via Phabricator via llvm-commits]

ABataev marked an inline comment as done.
ABataev added inline comments.


================
Comment at: llvm/test/Transforms/SLPVectorizer/X86/zext.ll:263
+; SLM-NEXT:    [[TMP20:%.*]] = extractelement <2 x i16> [[TMP12]], i32 1
+; SLM-NEXT:    [[V7:%.*]] = insertelement <8 x i16> [[V6]], i16 [[TMP20]], i32 7
 ; SLM-NEXT:    ret <8 x i16> [[V7]]
----------------
RKSimon wrote:
> Its really odd that SLM fails to use zext <8 x i8> to <8 x i16> like SSE2, I think the custom SLM extract/insert costs are affecting something unexpected?
Yes, the cost of ExtractElement instructions affects vectorization, must be fixed in another patch for the vectorization of the building of aggregates.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D57059/new/

https://reviews.llvm.org/D57059