[PATCH] D102002: [PassManager] unify vector passes between regular and LTO pipelines

[Sanjay Patel via Phabricator via llvm-commits]

spatel added a comment.

In D102002#2742500 <https://reviews.llvm.org/D102002#2742500>, @lebedev.ri wrote:

> One would guess that Unroll might introduce extra opportunities for SLP,
> but then proper LoopVectorize //should// have caught them already.

Yes, I think we have gone back-and-forth on that, and that might explain why LTO is different than regular.

> I'm also unsure about LoopLoadElimination - how exactly does it help?

I don't know. I'm leaning toward reduction + rebuild-with-justification as you can see, but I missed that diff initially! I can remove it too and see if that causes any tests to wiggle.
OTOH, this could end up being a multi-month slog as we get regression complaints. So it's probably better to surgically remove one pass at a time until we reach something close to what this patch is currently showing.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D102002/new/

https://reviews.llvm.org/D102002