[PATCH] D105390: [X86] Lower insertions into upper half of an 256-bit vector as broadcast+blend (PR50971)

[Craig Topper via Phabricator via llvm-commits]

craig.topper added inline comments.


================
Comment at: llvm/test/CodeGen/X86/avx512-insert-extract.ll:653
 define <16 x i16> @insert_v16i16(<16 x i16> %x, i16 %y, i16* %ptr) {
-; CHECK-LABEL: insert_v16i16:
-; CHECK:       ## %bb.0:
-; CHECK-NEXT:    vpinsrw $1, (%rsi), %xmm0, %xmm1
-; CHECK-NEXT:    vextracti128 $1, %ymm0, %xmm0
-; CHECK-NEXT:    vpinsrw $1, %edi, %xmm0, %xmm0
-; CHECK-NEXT:    vinserti128 $1, %xmm0, %ymm1, %ymm0
-; CHECK-NEXT:    retq
+; KNL-LABEL: insert_v16i16:
+; KNL:       ## %bb.0:
----------------
Is this really better? I assume this what we get for an AVX2 target too. Not just KNL?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D105390/new/

https://reviews.llvm.org/D105390