[PATCH] D25685: [DAG] use isConstOrConstSplat in ComputeNumSignBits to optimize SRA

[Simon Pilgrim via llvm-commits]

RKSimon accepted this revision.
RKSimon added a comment.
This revision is now accepted and ready to land.

LGTM - I think isConstOrConstSplatFP should probably moved over at the same time as isConstOrConstSplat (separate preliminary commit) followed by the SRA fix - make sense?

We can then investigate adding a "AllowUndefs" args to both of them in a followup commit and start the fun job of consolidating all those predicates.....


https://reviews.llvm.org/D25685