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

[Sam Parker via Phabricator via llvm-commits]

samparker added a comment.

> ideally the costmodel should always be costing hypothetical instructions, not relying on real ones.

This probably isn't true for most optimisations though, but of course the vectorizer is an important (but small) user of this. I would suspect that vectorizer should only really be passing the instruction when calculating the scalar cost of the loop.  But I'll revisit this once the context stuff is in.


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

https://reviews.llvm.org/D78937