[PATCH] D159354: [SROA] Limit the number of allowed slices when trying to split allocas

[Alexander via Phabricator via llvm-commits]

alex-t added a comment.

This change breaks the AMDGPU backend performance. It causes the 85% performance drop in certain rocBLAS benchmarks.
The approach itself seems weird to me. You'd better add a target hook to let the target decide about the thresholds instead of deciding for others.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D159354/new/

https://reviews.llvm.org/D159354