[PATCH] D105673: [SelectionDAG] Fix the representation of ISD::STEP_VECTOR.

Craig Topper via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Wed Jul 21 09:45:46 PDT 2021


craig.topper added a comment.

In D105673#2892667 <https://reviews.llvm.org/D105673#2892667>, @frasercrmck wrote:

> Without having dug in I suspect the RV32 regressions can be fixed with some improvements to the scalar splatting. In many of these cases we should be able to know that the scalar -- while not a constant -- is a sign-extended XLEN-sized value.
>
> I'm okay with doing that separately after this merges because it sounds enough to me like a new feature. Plus 64-bit vectors on RV32 aren't our strongest suit to begin with. Perhaps someone else may disagree though? @craig.topper?

I'm also ok doing this after this merges. 64-bit vectors on RV32 aren't a priority for me right now.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D105673/new/

https://reviews.llvm.org/D105673



More information about the llvm-commits mailing list