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.

Sign up for GitHub

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

Jump to bottom

[FEA] numba-cuda compatibility #117

Open
gmarkall opened this issue Nov 18, 2024 · 1 comment · May be fixed by #129
Open

[FEA] numba-cuda compatibility #117

gmarkall opened this issue Nov 18, 2024 · 1 comment · May be fixed by #129
Labels
feature request New feature or request

Comments

@gmarkall
Copy link
Contributor

Presently pynvjitlink patches and tests against Numba with its builtin CUDA target. The current release of numba-cuda incorporates the pynvjitlink patch so that patching is no longer required.

pynvjitlink should be updated so that:

  • If numba-cuda is in use, then it does not patch its linker.
  • If numba-cuda is in use, then it is at least version 0.0.18 (earlier versions don't include the patch, and may not be compatible with being patched by pynvjitlink).

This will also allow the Numba-related tests in the pynvjitlink CI to pass with later CUDA toolkits (e.g. 12.5) as it has fixed the expected PTX patterns generated by newer toolkits in the tests of atomics (See https://github.com/NVIDIA/numba-cuda/pull/40/files#diff-d48759455e529fd3994f3380640329c652dae73f5360cb977359c8c43461925b).
@gmarkall gmarkall added the feature request New feature or request label Nov 18, 2024
@gmarkall gmarkall mentioned this issue Nov 18, 2024
Open
@isVoid
Copy link
Contributor

isVoid commented Feb 18, 2025 •
edited
Loading

Can you clarify this issue? It seems like there's some conflict between point 1 and point 2.

I think I understand this more now. You meant that if numba-cuda is in use, pynvjitlink should not apply the patch to numba linker. Because after 0.0.18, the patch is already included. And before 0.0.18, pynvjitlink is no longer compatible.

@isVoid isVoid linked a pull request Feb 18, 2025 that will close this issue
Open
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
feature request New feature or request
Projects
None yet
Milestone
No milestone
Development

Successfully merging a pull request may close this issue.

Raise when pynvjitlink is in use with Numba-CUDA isVoid/pynvjitlink
2 participants
@gmarkall @isVoid