[PATCH] D102748: [LoopUnroll] Don't unroll before vectorisation

[Dávid Bolvanský via Phabricator via llvm-commits]

xbolva00 added a comment.

>> This, I think, also relies on the loop vectoriser which seems more powerful than SLP vectorisation currently.

Right.
Also,  SLP does not have the ability to add runtime checks*. So we blindly fully unroll loops, which are otherwise vectorizable with Loop Vectorizer with RT checks, and then we miss many vectorization opportunities.

I really would like to see loop vectorizer before SLP.

- See PR40976


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102748/new/

https://reviews.llvm.org/D102748