[PATCH] D149281: Not disable loop unroll for vectorized loops on AMDGPU target

[Nikita Popov via Phabricator via llvm-commits]

nikic added a comment.

@alex-t The premise of the change was that if a loop is vectorized, the vectorizer is also responsible for unrolling it via interleaving. The question is why the vectorizer failed to unroll the loop in your workload. We should understand that before adding additional options.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D149281/new/

https://reviews.llvm.org/D149281