cupy v13.6.0

[regro-cf-autotick-bot]

It is very likely that the current package version for this feedstock is out of date.

Checklist before merging this PR:

• Dependencies have been updated if changed: see upstream
• Tests have passed
• Updated license if changed and license_file is packaged

Information about this PR:

1. Feel free to push to the bot's branch to update this PR if needed.
2. The bot will almost always only open one PR per version.
3. The bot will stop issuing PRs if more than 3 version bump PRs generated by the bot are open. If you don't want to package a particular version please close the PR.
4. If you want these PRs to be merged automatically, make an issue with @conda-forge-admin,please add bot automerge in the title and merge the resulting PR. This command will add our bot automerge feature to your feedstock.
5. If this PR was opened in error or needs to be updated please add the bot-rerun label to this PR. The bot will close this PR and schedule another one. If you do not have permissions to add this label, you can use the phrase @conda-forge-admin, please rerun bot in a PR comment to have the conda-forge-admin add it for you.

Pending Dependency Version Updates

Here is a list of all the pending dependency version updates for this repo. Please double check all dependencies before merging.

Name	Upstream Version	Current Version
cudatoolkit	12.0.0
cutensor	2.3.0.6

_{This PR was created by the regro-cf-autotick-bot. The regro-cf-autotick-bot is a service to automatically track the dependency graph, migrate packages, and propose package version updates for conda-forge. Feel free to drop us a line if there are any issues! This PR was generated by https://github.com/regro/cf-scripts/actions/runs/17037632147 - please use this URL for debugging.}

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-linting service.

I just wanted to let you know that I linted all conda-recipes in your PR (recipe/meta.yaml) and found it was in an excellent condition.

[leofang]

DO NOT MERGE yet. I need to update the recipe to also build for CUDA 13.

[leofang]

@conda-forge-admin, please rerender

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/17049974361. Examine the logs at this URL for more detail.}

[leofang]

@conda-forge-admin, please rerender

[leofang]

Note to self: On Windows, CONDA_OVERRIDE_CUDA does not seem to be set in the CI anymore.

[conda-forge-admin]

Hi! This is the friendly automated conda-forge-webservice.

I tried to rerender for you, but it looks like there was nothing to do.

_{This message was generated by GitHub Actions workflow run https://github.com/conda-forge/conda-forge-webservices/actions/runs/17058550176. Examine the logs at this URL for more detail.}

[leofang]

Thanks for help from @carterbox @bdice, I've applied the CUDA 11, 12, 13 migrators from conda-forge/conda-forge-pinning-feedstock#7653 with a few minor tweaks, and successfully built CuPy v13.6.0 for CUDA 11, 12, and 13 following the upstream 🎉

The critical key to enable this is to ensure the timestamp of the CUDA 11 migrator (migrator_ts) comes the last, i.e. the highest. The discussion is summarized in this thread: conda-forge/conda-forge-pinning-feedstock#7653 (comment). I've verified that any other minor touches that I did to the migrators do not affect the re-rendering outcome. We can refresh the migrators once they're merged upstream, but let me skip that part to avoid wasting CI resources 🙂 Also the CUDA 11 migrator will be removed in the next CuPy release anyway (#315).

Next is we need to skip the optional dependencies (NCCL, cuTENSOR) for the CUDA 13 builds because they aren't ready yet. We can re-enable them later (#316, #317).

Another note is we need to set CUDA_PATH in the Windows + CUDA 13 CI, because the new package layout pushes everything one level down (an extra x64 subdirectory is added), which deviates from normal search paths $PREFIX\Library\{bin,lib}. This can be avoided if we incorporate this layout expectation into CuPy's build system, similar to what we've done in the past for supporting Linux + CUDA 12 (cc @kmaehashi for vis). For now I keep the diff minimal because I cannot spin the Windows pipeline locally via build-locally.py, which makes it very hard to develop/debug and I don't have time.

Final note is this observation (#314 (comment)). We should avoid relying on having CONDA_OVERRIDE_CUDA set in the Windows pipelines.

[kmaehashi]

Thank you as always for the excellent work here, @leofang, @carterbox and @bdice!

This can be avoided if we incorporate this layout expectation into CuPy's build system, similar to what we've done in the past for cupy/cupy#7603 (cc @kmaehashi for vis).

Yes, let's add the fix for CUDA 13 in the upstream. 👍

[leofang]

Yes, let's add the fix for CUDA 13 in the upstream. 👍

cupy/cupy#9378