[PATCH] D130956: [X86][MC] Always emit `rep` prefix for `bsf`

[Craig Topper via Phabricator via llvm-commits]

craig.topper added inline comments.


================
Comment at: llvm/test/CodeGen/X86/clz.ll:49
 ; X86:       # %bb.0:
-; X86-NEXT:    bsfw {{[0-9]+}}(%esp), %ax
+; X86-NEXT:    rep bsfw {{[0-9]+}}(%esp), %ax
 ; X86-NEXT:    retl
----------------
pengfei wrote:
> craig.topper wrote:
> > We need to promote bsrw to bsrl. tzcntl is faster than tzcntw. tzcntw has a false dependency to preserve the upper 48 bits of the result register in Intel CPUs. But separate patch please.
> Can we allow to always do that? The result is not equal for tzcnt if we care the src is 0: https://godbolt.org/z/enahabbM7
I meant in SelectionDAG.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D130956/new/

https://reviews.llvm.org/D130956