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

[Matt Arsenault via Phabricator via llvm-commits]

arsenm added a comment.

In D84779#2206378 <https://reviews.llvm.org/D84779#2206378>, @timcorringham wrote:

> As I said in my reply to Matt, I made it target specific to avoid cluttering the llvm.loop metadata with something that is only used by one target. I am happy to make it llvm.loop.unroll.threshold if that is considered better.
>
> However, I'm not really sure how generic it is as no other target has felt the need to do this (at least publicly). The function attribute is also amdgpu specific. Neither are used in any generic code, only in the target specific getUnrollingPreferences().
>
> Is the consensus to change it to llvm.llop.unroll.threshold?

Yes, but first I think we need to back up and establish why directly putting arbitrary thresholds in the IR is useful in the first place


Repository:
  rG LLVM Github Monorepo

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

https://reviews.llvm.org/D84779