[PATCH] D134228: [RISCV] Massage debug loc to improve scheduling of addressing for scatter/gather

[Philip Reames via Phabricator via llvm-commits]

reames abandoned this revision.
reames added a comment.

D134382 <https://reviews.llvm.org/D134382> handled the motivating address scaling problem.  I went back and took a look at the subset of this patch for the extension, and the test diffs were a real mix.  The improvements seen were fairly minor, but we seemed to sometimes trigger register spilling (particularly at high lmul) by extending live ranges.  Given the SEW of the extend isn't tied to the addressing SEW or the data SEW, I don't think there's a solid argument to prefer one over the other here.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D134228/new/

https://reviews.llvm.org/D134228