[PATCH] D84779: [AMDGPU] Add amdgpu specific loop threshold metadata

[Johannes Doerfert via Phabricator via llvm-commits]

jdoerfert added a comment.

My concern is that we have a way to annotate loops with unroll specific metadata, and this introduces an AMDGPU specific way to do something that we want generically (as evident by the function attribute).
So, if you need per-loop control, why not introduce `‘llvm.loop.unroll.threshold`? I mean, at the end of the day, what is AMDGPU specific about this?


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D84779/new/

https://reviews.llvm.org/D84779