[PATCH] D127447: [RISCV] Cost model for scalable reductions

[Philip Reames via Phabricator via llvm-commits]

reames added a comment.

@craig.topper You seem to be assuming a much different implementation that I was.  I was thinking of these as likely being constant time operations.

For your model, I think we'd have to be conservative for scalable vectors and use the maximum possible VL.  That would fit naturally in the structure, but would strongly bias against using them.  Should I switch to that?

I suspect we're also going to need a process feature flag here at some point, but we can defer that.  :)


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D127447/new/

https://reviews.llvm.org/D127447