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

[Tim Corringham via Phabricator via llvm-commits]

timcorringham added a comment.

There is some more info about the motivation for this in https://github.com/GPUOpen-Drivers/llpc/pull/863

Basically it gives the front end more flexibility in how it treats unroll hints. It has been shown that a lot of user supplied hints are not optimal for amdgpu if obeyed strictly. Hopefully this is generic enough to be used in other as yet unforeseen ways.


Repository:
  rG LLVM Github Monorepo

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

https://reviews.llvm.org/D84779