Enable shm_comm support for arm

[phalani-paladugu]

This PR enables shared memory communication in single node for arm hosts - #7625

[delock]

Overall looks good to me. I'm also glad that the refractory from @heyujiao99 in #7519 really helps alot! Around 100 lines of code to support a new HW architecture, wow! Kudos to @phalani-paladugu and @heyujiao99 !

[phalani-paladugu]

[phalani-paladugu]

Hi, is there a known issue with the cpu-torch-latest CI job?
In the last two runs, the unit tests appear to complete successfully, but the job never finishes and eventually fails.
I don’t see any test failures in the logs—only that the workflow seems to hang after the tests have passed.

[sfc-gh-truwase]

@phalani-paladugu, yes, something seems to be wrong with cpu-torch-latest. Right now I see that it is running for over 2hrs, which is unexpected.

@loadams do you know what could be wrong?

[sfc-gh-truwase]

@phalani-paladugu FYI, @tohtana is investigating in the linked PR.

[tohtana]

Hi @phalani-paladugu,
Based on the investigation in #7851, I think this change will fix the hang issue in the cpu test. It's not a fully clean fix, but at least we can unblock merging your PR.
As I cannot push a commit to this PR (not editable by maintainers), can you merge the commit?

[phalani-paladugu]

Hi @tohtana ,
Thanks for the investigation and for proposing the fix. I have added the change you suggested. It looks like a few checks are currently awaiting approval. Should be good to go once they have cleared.

[tohtana]

All tests passed and the PR is merged now. This is a significant update. Thank you @phalani-paladugu for your contribution!