[PATCH] D30176: [X86][SSE] Attempt to extract vector elements through target shuffles

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon marked 4 inline comments as done.
RKSimon added inline comments.


================
Comment at: test/CodeGen/X86/promote-vec3.ll:19
+; SSE3-NEXT:    pextrw $1, %xmm0, %edx
+; SSE3-NEXT:    pextrw $2, %xmm0, %ecx
 ; SSE3-NEXT:    punpcklwd {{.*#+}} xmm0 = xmm0[0],xmm1[0],xmm0[1],xmm1[1],xmm0[2],xmm1[2],xmm0[3],xmm1[3]
----------------
filcab wrote:
> Interesting.
In a future patch we can hopefully fold extract(insert(v,x,c),c) -> x


Repository:
  rL LLVM

https://reviews.llvm.org/D30176