[PATCH] D152019: [RISCV][CostModel] Model vrgather.vv as being quadradic in LMUL

[Philip Reames via Phabricator via llvm-commits]

reames added a comment.

In D152019#4392239 <https://reviews.llvm.org/D152019#4392239>, @craig.topper wrote:

> I think the vrgather.vv for reverse needs at most 2 sources for each VLEN piece. Is it likely hardware would optimize for that?

It's certainly possible.  Do you have existence proof in either direction?  I'm guessing at the moment.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D152019/new/

https://reviews.llvm.org/D152019