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

[REQUEST] Update example_walker.py to use tile API #417

Open
shi-eric opened this issue Jan 7, 2025 · 4 comments
Open

[REQUEST] Update example_walker.py to use tile API #417

shi-eric opened this issue Jan 7, 2025 · 4 comments
Assignees
Labels
enhancement New feature or request

Comments

@shi-eric
Copy link
Contributor

shi-eric commented Jan 7, 2025

Description

wp.matmul() will eventually be removed, so we want to make sure this example uses the tile programming model.

@shi-eric shi-eric added the enhancement New feature or request label Jan 7, 2025
@shi-eric
Copy link
Contributor Author

shi-eric commented Jan 7, 2025

BTW, one complication is that the tile API doesn't work with CUDA 11 builds due to the use of nvJitLink...

@daedalus5
Copy link
Contributor

Yeah that is a complication. Possibly will be an issue with automated testing? Is it possible to filter example tests by required CUDA version? Although we have other examples now that use the Tile API, so I guess it's no different.

This just means we don't directly support matrix multiplication unless you're running CUDA 12+ after the deprecation.

@shi-eric
Copy link
Contributor Author

shi-eric commented Jan 8, 2025

There's also a usage of wp.matmul() in the linear solvers module.

@gdaviet
Copy link
Contributor

gdaviet commented Jan 8, 2025

Yes, wp.matmul is currently used as the default dense matrix-vector multiplication algorithm for the iterative linear solvers. A tile-based alternative should be possible, and I don't believe the linear solvers are part of the officially supported API, so limiting support to cuda 12 might be an option.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
enhancement New feature or request
Projects
None yet
Development

No branches or pull requests

3 participants