[PATCH] D96037: [DAGCombiner] Don't fold FCOPYSIGN vector sign operand casts

[Fraser Cormack via Phabricator via llvm-commits]

frasercrmck accepted this revision.
frasercrmck added a comment.

In D96037#2542151 <https://reviews.llvm.org/D96037#2542151>, @luismarques wrote:

> I did propose that in the past: https://reviews.llvm.org/D66725
> I may have misunderstood the review feedback I got back then (spread over a variety of patches and personal email), but that feedback led to me abandoning that patch.
> I think part of the problem, at the time, was that my patches stressed too much the issues around the legality of the operation/operands, and not enough about the combine being desirable.
> When I re-analyzed this issue, before submitting this patch, it did seem to me that a hook would still be beneficial, even if many of the original problems had since been addressed by other means.

I see. I think the problem with implementing a hook is that it appears that no in-tree target benefits from this fcopysign optimization on vector types. So while I think it's the right idea in principle it's essentially the same as what you've got here as everyone would return false. I think this can go in as-is and a hook can come along later if a target wants it.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D96037/new/

https://reviews.llvm.org/D96037