Simplify matrix configuration for CI workflows

[sjain-stanford]

Addresses #1207.

Provisioned jobs:

# ubuntu - x86_64 - llvm in-tree     - pytorch binary - build+test    # most used dev flow and fastest signal
# ubuntu - x86_64 - llvm out-of-tree - pytorch source - build+test    # most elaborate build
# macos  - arm64  - llvm in-tree     - pytorch source - build only    # cross compile, can't test arm64

Main changes

• Spawn macos builds from a separate matrix (in the same workflow). It made sense to do this as they are fairly different from ubuntu (cross compile, use a different cmake configuration). This simplifies the matrix configuration and exclusions quite a bit, and makes the workflow a bit more tractable and maintenance friendly.
• Remove the submodule md5sum step for ccache config. This was broken for a while now.
• Removes unused matrix options - os, targetarch, python-version, llvmtype.
• Address ZSTD comment on @powderluv's cross compile PR.

Further improvements (to be addressed in follow-on):

• ubuntu-x86_64 out-of-tree integration tests fail (error); only run unit tests for now (tests are excluded in current CI too)

Passing workflow:

https://github.com/sjain-stanford/torch-mlir/actions/runs/2840676309

[powderluv]

This is looking great. The arm64 Pytorch src build should work if you remove this one line

torch-mlir/build_tools/build_libtorch.sh

Line 134 in 51bfe25

pip uninstall torch

It is trying to uninstall the systemwide torch.

[sjain-stanford]

The arm64 Pytorch src build should work if you remove this one line

Thanks, patched. Let's wait to see if the arm64 pytorch source workflow goes through with the fix. If there are more errors, I can revert to pytorch binary and land it for now (to avoid cache evictions the longer this is open). I'll wait for an "all green" CI before landing, but if this looks good otherwise, please feel free to ✅ this.

[powderluv]

LGTM. Feel free to switch arm64 pytorch source in a follow on.

[powderluv]

nicely done. The silly cache gets generated again when it merges so we got to wait for it again

[sjain-stanford]

The silly cache gets generated again when it merges so we got to wait for it again

Ah I was wondering why it didn't restore from cache after landing because the keys didn't change. Good to know this is normal. Maybe it treats GHA runs on PRs differently than runs on push to main. Oh well...

[sjain-stanford]

... and thank you for the help in reviewing it!