[PATCH] D99949: [AMDGPU][OpenMP] Add amdgpu-arch tool to list AMD GPUs installed

Jon Chesterfield via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Wed Apr 14 07:14:36 PDT 2021


JonChesterfield added a comment.

In D99949#2688869 <https://reviews.llvm.org/D99949#2688869>, @gregrodgers wrote:

> 1. It does not provide the infrastructure to identify runtime capabilities to satisfy requirements of a compiled image.

I believe we only require a value for '-march=' to unblock running tests on CI machines. I'd guess you're referring to target id stuff where clang fills in reasonable defaults already.

In D99949#2688869 <https://reviews.llvm.org/D99949#2688869>, @gregrodgers wrote:

> 2. It is not architecturally neutral or extensible to other architectures.

Yes. By design. Clang calling into some unrelated tool for nvptx seems fine, nvidia-smi may already suffice for that purpose.

If we want to do something clever about guessing whether the machine has an nvidia card or an amdgpu one, clang can call both this tool and the nvidia one, and check both return codes.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D99949/new/

https://reviews.llvm.org/D99949



More information about the cfe-commits mailing list