[PATCH] D23570: Fix SystemZ compilation abort caused by negative AND mask

[Ulrich Weigand via llvm-commits]

uweigand added a comment.

I agree it makes sense to do this transformation in PerformDAGCombine.  Patch looks good to me.

However, once we do this here, aren't the peepholes in SystemZInstrInfo.td now superfluous?  If so, they should be removed again ...

Also, please update the comment before combineSHIFTROT so it matches the new behaviour of the routine.


https://reviews.llvm.org/D23570