[PATCH] D115261: [LV] Disable runtime unrolling for vectorized loops.

[Philip Reames via Phabricator via llvm-commits]

reames added a comment.

In theory, I'm not a fan of this patch for many of the reasons @lebedev.ri spelled out.  Our cost model really should be good enough to avoid unrolling something which is not profitable without needing to have one pass overrule the heuristic in another.  In practice, I think this is a reasonable workaround for a very deep and hard to fix problem (poor unroll heuristics).  Assuming perf results are neutral, I'd be fine with this going in.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D115261/new/

https://reviews.llvm.org/D115261