[PATCH] D54095: [X86] X86DAGToDAGISel::matchBitExtract(): extract 'lshr' from `X`

[Roman Lebedev via Phabricator via llvm-commits]

lebedev.ri added a comment.

In https://reviews.llvm.org/D54095#1296929, @andreadb wrote:

> In https://reviews.llvm.org/D54095#1296822, @andreadb wrote:
>
> >
>
>
> ...
>  An alternative approach that uses MOVZ+OR instead of a MOV8rr is not bad. In particular, if the compiler knows how to optimize out the MOVZ by using the "known-zero-bits", then we may want to produce an OR instead of a partial register move.


Yeah, i think i will go with MOVZ+OR.


Repository:
  rL LLVM

https://reviews.llvm.org/D54095