[libc-commits] [PATCH] D139839: [libc] Add a loader utility for AMDHSA architectures for testing

[Joseph Huber via Phabricator via libc-commits]

jhuber6 added a comment.

In D139839#3999316 <https://reviews.llvm.org/D139839#3999316>, @sivachandra wrote:

> In D139839#3991165 <https://reviews.llvm.org/D139839#3991165>, @sivachandra wrote:
>
>> 1. I will rename the current `loader` directory to `startup` and you can move the GPU startup implementation there. As it stands in this patch, it is a trivial implementation so you can land it separately after I do the renaming.
>
> This is now done.
>
>> ii) Develop the ability to run any unit test as an integration test. We really want to be explicit in listing targets in the libc project so one approach could be to list integration test targets which point to the unit test sources. And, this can be done for a Linux host and the GPU.
>
> Once you land the GPU startup code, we can discuss this more.

Sounds good, I'll make a separate patch for it.

I propose we rewrite the loader implementation to use OpenMP runtime calls to perform the loading, but keep the general scheme the same. I think this is a good compromise so we can treat the GPU as a standalone system, but forego re-implementing kernel launch code that exists in-tree. If we wanted to prevent the dependency then we can go back and write a dedicated loader for that architecture.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D139839/new/

https://reviews.llvm.org/D139839