Adding cuda-nvtx recipe

[adibbley]

Checklist

• Title of this PR is meaningful: e.g. "Adding my_nifty_package", not "updated meta.yaml".
• License file is packaged (see here for an example).
• Source is from official source.
• Package does not vendor other packages. (If a package uses the source of another package, they should be separate packages or the licenses of all packages need to be packaged).
• If static libraries are linked in, the license of the static library is packaged.
• Package does not ship static libraries. If static libraries are needed, follow CFEP-18.
• Build number is 0.
• A tarball (url) rather than a repo (e.g. git_url) is used in your recipe (see here for more details).
• GitHub users listed in the maintainer section have posted a comment confirming they are willing to be listed there.
• When in trouble, please check our knowledge base documentation before pinging a team.

[conda-forge-webservices]

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 (recipes/cuda-nvtx) and found it was in an excellent condition.

[bdice]

Recipe looks fine aside from minor suggestions on doc_url. I haven't checked the package contents yet.

[jakirkham]

Currently we have nvtx-c, which packages as similar library based on the NVTX repo's C source code.

This in turn is used by nvtx, Python bindings to the C library, which is based on this Python code.

With this PR, we would now have cuda-nvtx, which packages NVTX based on the CTK copy.

This situation is somewhat reminiscent of the CCCL situation we just worked through

Wonder if we should do something to...

1. resolve conflicts
2. disambiguate packages
3. sunset unused components
4. ?

Would be interested in hearing thoughts from folks here 🙂

[jakirkham]

cc @shwina (who may also have thoughts on this)

[shwina]

We could change the nvtx Python package to depend on cuda-nvtx? (and do away with nvtx-c)

[bdice]

I support @shwina’s plan.

[jakirkham]

Generally that seems reasonable. One suggestion would be to introduce a conflict in nvtx-c analogous to what we did for cub ( conda-forge/cub-feedstock#23 )

[vyasr]

One change request.

Also, do we need to make changes to nvtx-c before we merge this? CC @jakirkham

[vyasr]

Approving assuming the nvtx-c conflict is addressed at some point.

[kkraus14]

I think we need a repodata patch on nvtx-c to ensure these don't clobber each other and then we can maybe consider making it an alias?

Additionally, what is the relationship and versioning in relation to the GitHub repo https://github.com/NVIDIA/NVTX?

[jakirkham]

To avoid clobbering nvtx-c, suggesting that we place the contents here in targets on Windows as we did with cuda-cccl (for a similar reason)

[jakirkham]

Noticed a clobbering issue regarding headers on Windows CI

In looking into this, it appears cuda-nvtx may not actually be needed on Windows since there are no .dlls in the redist or the package here and the headers are already in cuda-nvtx-dev

Suggesting we disable cuda-nvtx builds on Windows

[jakirkham]

If we drop cuda-nvtx on Windows, there would be some follow up changes needed. Added those below

[jakirkham]

Maybe this is a better approach for the clobbering issue on Windows?

Suggested change

	files:
	- lib/libnv*.so.* # [linux]
	- targets/{{ target_name }}/lib/*.so.* # [linux]
	files: [] # [win]
	files: # [linux]
	- lib/libnv*.so.* # [linux]
	- targets/{{ target_name }}/lib/*.so.* # [linux]

[conda-forge-webservices]

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

I wanted to let you know that I linted all conda-recipes in your PR (recipes/cuda-nvtx) and found some lint.

Here's what I've got...

For recipes/cuda-nvtx:

• Selectors are suggested to take a <two spaces>#<one space>[<expression>] form. See lines [39]

[conda-forge-webservices]

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 (recipes/cuda-nvtx) and found it was in an excellent condition.

[jakirkham]

Resolved the Windows clobbering issues by dropping cuda-nvtx. That package was not needed on Windows anyways (no runtime libraries & headers in cuda-nvtx-dev).

[jakirkham]

Thanks all! 🙏

Let's follow up on anything else in the feedstock 🙂