Provide full CUDA errors for device scan messages. - Githubissues

glotzerlab / hoomd-blue

Molecular dynamics and Monte Carlo soft matter simulation on GPUs.

http://glotzerlab.engin.umich.edu/hoomd-blue

BSD 3-Clause "New" or "Revised" License

335 stars 131 forks source link

Provide full CUDA errors for device scan messages. #1803

Closed joaander closed 3 months ago

joaander commented 4 months ago

Description

Provide full CUDA errors for device scan messages.

Motivation and context

HIP errors state "Unknown error" for almost every CUDA error. This does not help users.

How has this been tested?

CI checks.

Change log

Changed:

* Provide the full CUDA error message when scanning devices
  (`#1803 <https://github.com/glotzerlab/hoomd-blue/pull/1803>`__).

Checklist:

[x] I have reviewed the Contributor Guidelines.
[x] I agree with the terms of the HOOMD-blue Contributor Agreement.
[x] My name is on the list of contributors (sphinx-doc/credits.rst) in the pull request source branch.

joaander commented 3 months ago

While this does not provide a better message in #1802 as I hoped, the code does work and may provide a better message for someone in the future. I tested it by triggering an error manually before the message is formatted.