WIP: Use more system libs

[hmaarrfk]

I removed the split package so I can actually debug things locally. Kinda annoying that conda debug doesn't work with split packages. Even if I get the output id correct, I can't see to active the environment correctly.

Checklist

• Used a personal fork of the feedstock to propose changes
• Bumped the build number (if the version is unchanged)
• Reset the build number to 0 (if the version changed)
• Re-rendered with the latest conda-smithy (Use the phrase @conda-forge-admin, please rerender in a comment in this PR for automated rerendering)
• Ensured the license file is being packaged.

[conda-forge-linter]

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 (recipe) and found some lint.

Here's what I've got...

For recipe:

• When defining a source/url please add a sha256, sha1 or md5 checksum (sha256 preferably).
• requirements: build: cmake != 3.23.0 should not contain a space between relational operator and the version, i.e. cmake !=3.23.0
• requirements: run_constrained: pytorch-cpu = 1.11.0 should not contain a space between relational operator and the version, i.e. pytorch-cpu =1.11.0
• requirements: run_constrained: pytorch-gpu = 99999999 should not contain a space between relational operator and the version, i.e. pytorch-gpu =99999999
• requirements: run_constrained: pytorch-gpu = 1.11.0 should not contain a space between relational operator and the version, i.e. pytorch-gpu =1.11.0
• requirements: run_constrained: pytorch-cpu = 99999999 should not contain a space between relational operator and the version, i.e. pytorch-cpu =99999999

[conda-forge-linter]

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 (recipe) and found some lint.

Here's what I've got...

For recipe:

• When defining a source/url please add a sha256, sha1 or md5 checksum (sha256 preferably).
• requirements: run_constrained: pytorch-cpu = 1.11.0 should not contain a space between relational operator and the version, i.e. pytorch-cpu =1.11.0
• requirements: run_constrained: pytorch-gpu = 99999999 should not contain a space between relational operator and the version, i.e. pytorch-gpu =99999999
• requirements: run_constrained: pytorch-gpu = 1.11.0 should not contain a space between relational operator and the version, i.e. pytorch-gpu =1.11.0
• requirements: run_constrained: pytorch-cpu = 99999999 should not contain a space between relational operator and the version, i.e. pytorch-cpu =99999999

[hmaarrfk]

Torch is failing eventually with:

/home/conda/feedstock_root/build_artifacts/pytorch_1653938249907/_build_env/bin/../lib/gcc/x86_64-conda-linux-gnu/7.5.0/../../../../x86_64-conda-linux-gnu/bin/ld: cannot find -lgloo_cuda

[ngam]

Do we pass this? https://github.com/pytorch/builder/blob/c14b882cca0b4c425771d87feb6ebea2fc7ab242/conda/pytorch-nightly/meta.yaml#L73

[ngam]

I am not saying it will solve it, I just noticed they pass that env var for some reason.

[hmaarrfk]

I don't think we have a gloo package yet at conda forge.

Unless we go through with a major unbundling step. I don't think it is worthwhile to do it too peicemeal.

I haven't built gloo with GPU support.

[ngam]

agree, we will want to do some unbundling eventually...