[PATCH] D42935: [X86] Use min/max for vector ult/ugt compares if avoids a sign flip.

[Simon Pilgrim via Phabricator via llvm-commits]

RKSimon added a comment.

I've been hitting some of the same issues while finishing the TRUNC(CLAMP()) truncation -> PACK support. Would it be a good idea to enable custom lowering of all (U,S)(MAX,MIN) on SSE2+ targets? It cleans out a lot of CMPGT/CMPGE mismatches and makes it easier to match a lot of this stuff using the SMIN etc. opcodes directly.


https://reviews.llvm.org/D42935