[PATCH] D28810: [X86][SSE] Use MOVMSK for all_of/any_of reduction patterns

[Andrea Di Biagio via Phabricator via llvm-commits]

andreadb added inline comments.


================
Comment at: test/CodeGen/X86/vector-compare-all_of.ll:583-585
+; AVX2-NEXT:    movw $-1, %ax
+; AVX2-NEXT:    cmovnew %cx, %ax
 ; AVX2-NEXT:    cwtl
----------------
Rather than `movw+cmovnew+cwtl`, can we select instead `movl+cmovnel`?
The advantage is that we remove the partial register writes on AX, and we can get rid of the sign extend at the end (cwtl).


Repository:
  rL LLVM

https://reviews.llvm.org/D28810