[PATCH] D69575: Improve module.pcm lock file performance on machines with high core counts

[Ladd Van Tol via Phabricator via llvm-commits]

Ladd added a comment.

Apologies -- I haven't had time to try to reconcile these patches. The edge cases that @PaulTaykalo identified seem worth considering.

I'd be open to simplifying my patch to remove the kqueue, and using just the random wait behavior. This produces basically the same performance win in my measurements, and is not subject to the edge case. We could then pursue the more clever kqueue-based waiting as a follow-on.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D69575/new/

https://reviews.llvm.org/D69575