[PATCH] D89969: [SLP] Consider alternatives for cost of select instructions.

[Florian Hahn via Phabricator via llvm-commits]

fhahn added a comment.

In D89969#2350457 <https://reviews.llvm.org/D89969#2350457>, @spatel wrote:

> I haven't looked at the details here, but I just want to chime in that I'm cautiously optimistic that we'll have the basic cost model for intrinsics straightened out soon ( example: c963bde0152a <https://reviews.llvm.org/rGc963bde0152a741d58b9e1e9ac485d8f2e6c6d58> ).
> That should then allow us to canonicalize to min/max/abs intrinsics in instcombine. I'm not sure if this patch makes that goal harder, easier, or neither. We will eventually want the vectorizers to recognize and produce those intrinsics directly.

I don't think this patch would make things harder. The change itself may become less important, if some of those scalar selects are turned into min/max before SLPVectorizer sees them. We can remove it, as @xbolva00 suggested, if it turns out it is not needed any more.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D89969/new/

https://reviews.llvm.org/D89969