Are CUDA targets still supported for DPCTL?

[niteya-shah]

I am trying to install the Codeplay plugins for Nvidia GPUs and am getting the error

warning  libmamba Failed to load subdir: Transfer finalized, status: 404 [https://developer.codeplay.com/conda/linux-64/repodata.json.zst] 0 bytes
warning  libmamba Failed to load subdir: Transfer finalized, status: 404 [https://developer.codeplay.com/conda/noarch/repodata.json.zst] 0 bytes
[+] 0.0s
warning  libmamba Unable to retrieve repodata (response: 404) for 'https://developer.codeplay.com/conda/linux-64/repodata.json'
warning  libmamba Unable to retrieve repodata (response: 404) for 'https://developer.codeplay.com/conda/linux-64/repodata.json'
critical libmamba Multiple errors occurred:
    Transfer finalized, status: 404 [https://developer.codeplay.com/conda/noarch/repodata.json] 5254 bytes
    Subdir conda-forge/noarch not loaded!
    Subdir nodefaults/noarch not loaded!
    Subdir https://developer.codeplay.com/conda/noarch not loaded!
    If you run into this error repeatedly, your package cache may be corrupted.
    Please try running `mamba clean -a` to remove this cache before retrying the operation.
    
    If you still are having issues, please report the error on `mamba-org/mamba`'s issue tracker:
    https://github.com/mamba-org/mamba/issues/new?assignees=&labels=&projects=&template=bug.yml

None of codeplay's conda repos seem to accessible. Does this mean CUDA is deprecated?

[ndgrigorian]

@niteya-shah
CUDA isn't deprecated from dpctl side, but I believe codeplay is not releasing the plugin anymore. There isn't one since 2025.1.

To use with CUDA, you will have to install the OS compiler from Intel LLVM repo, which still builds/releases nightly compiler with CUDA adapters.

The OS compiler workflow in .github can give you a rough idea of how to do it. Be warned that the OS compiler often breaks, so you may have to step back several days to over a week to find one that works.

It would be a good idea for us to track a "sort of known good" OS compiler release to recommend. In addition, there is work starting to support builds with AdaptiveCpp, which supports CUDA out of the box.

I will try to verify that CUDA builds are possible w/ OS compiler as well independently.

[clawdney]

Hi! I'm interested in helping with this. Could you share more context about the use case?