[PATCH] D108738: [SelectionDAG] Optimize bitreverse expansion to minimize the number of mask constants.

[Craig Topper via Phabricator via llvm-commits]

craig.topper added inline comments.


================
Comment at: llvm/test/CodeGen/X86/vector-bitreverse.ll:690
 ; SSE2-NEXT:    por %xmm1, %xmm0
-; SSE2-NEXT:    movdqa {{.*#+}} xmm1 = [51,51,51,51,51,51,51,51,51,51,51,51,51,51,51,51]
-; SSE2-NEXT:    pand %xmm0, %xmm1
-; SSE2-NEXT:    psllw $2, %xmm1
-; SSE2-NEXT:    pand {{\.?LCPI[0-9]+_[0-9]+}}(%rip), %xmm0
-; SSE2-NEXT:    psrlw $2, %xmm0
+; SSE2-NEXT:    movdqa %xmm0, %xmm1
+; SSE2-NEXT:    psrlw $2, %xmm1
----------------
pengfei wrote:
> pengfei wrote:
> > Seems we have 2 extra move here. Is this a regression?
> Seems we reduce the constant load :)
Right we increased the number of instructions, but decreased the number of loads.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D108738/new/

https://reviews.llvm.org/D108738