[PATCH] D109989: [X86] Improve i8 all-ones element insertion in pre-SSE4.1

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added inline comments.


================
Comment at: llvm/test/CodeGen/X86/insertelement-ones.ll:314-315
 ; SSE2:       # %bb.0:
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [0,255,255,255,255,255,255,255,255,255,255,255,255,255,255,255]
-; SSE2-NEXT:    movl $255, %eax
-; SSE2-NEXT:    movd %eax, %xmm2
-; SSE2-NEXT:    pandn %xmm2, %xmm1
-; SSE2-NEXT:    pslldq {{.*#+}} xmm2 = zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,zero,xmm2[0]
-; SSE2-NEXT:    por %xmm1, %xmm2
-; SSE2-NEXT:    por %xmm2, %xmm0
+; SSE2-NEXT:    orps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
+; SSE2-NEXT:    orps {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
 ; SSE2-NEXT:    retq
----------------
lebedev.ri wrote:
> Are we intentionally not merging constant vectors in hope of promoting their reuse?
Have these been lowered to constant pool loads before reassociation sees it?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109989/new/

https://reviews.llvm.org/D109989