Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fix HIP assert in cooperative groups #12

Conversation

matyas-streamhpc
Copy link

The header was broken because asm("trap;") is a CUDA function and HIP does not support it.

The issue appeared only with the -O0 flag because of the NDEBUG switch.

The recent HIP version supports device assert, therefore we are recommending its use in the recent bug fix.

If the mentioned solution is not preferred, we can replace asm("trap;") with asm("s_trap 1;") as a working alternative.


The MR solves ROCm/HIP#233 and relates to ROCm/HIP#3317

Perhaps it relates to ROCm/ROCm#1729

@iassiour
Copy link
Contributor

Hi @matyas-streamhpc this has already been changed in internal clr:

hip_abort is now defined like:
#define __hip_abort()
{ abort(); }

Could you please confirm that it solves the issue you are seeing. The change should become available in the next release (6.0) and should also become available in the external clr repo here with the next sync.

@matyas-streamhpc
Copy link
Author

Thanks! I can confirm it: I tried it and it works well.

@matyas-streamhpc matyas-streamhpc deleted the fix_hip_amd_cooperative_group_assert branch November 30, 2023 16:06
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants