[PATCH] D78937: [CostModel] Use isExtLoad in BasicTTI

[Dave Green via Phabricator via llvm-commits]

dmgreen added a comment.
Herald added a subscriber: wuzish.

This would conflict with D79162 <https://reviews.llvm.org/D79162>. That's not the prettiest patch in the world, but it solves some real problems we have in the vectorizer and moving further away from it seems like a mistake.

I get that given an (accurate) context instruction I, it's good to make use of it to produce better costs. But ideally the costmodel should always be costing hypothetical instructions, not relying on real ones.

On a concrete level, this might now be using the type of the actually load (say i32), and not the type the costmodel has provided (v4i32, for example)?


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D78937/new/

https://reviews.llvm.org/D78937