llvm icon indicating copy to clipboard operation
llvm copied to clipboard

[CI] Run ESIMD tests on Arc in precommit when ESIMD changed

Open sarnex opened this issue 2 years ago • 10 comments

Let's see if this works

sarnex avatar Feb 07 '24 16:02 sarnex

@aelovikov-intel Looks like ESIMD + Matrix tests on DG2 take ~14 min. Is that too long?

Let's ignore the failures for now, I don't know what's going on but if we want to proceed with this change I'll investigate.

sarnex avatar Feb 07 '24 20:02 sarnex

We've had about 75 pre-commit jobs in the last 24 hours. Just pre-commit would take about 18 hours at that pace, and we also use that single runner for post-commit/nightly.

I'd say yes, too long.

aelovikov-intel avatar Feb 07 '24 20:02 aelovikov-intel

@aelovikov-intel If we only ran it when there were ESIMD changes is that still too long? I'm not sure how many PRs we have per day but probably 1-3. Maybe I'll actually have to use REST.

sarnex avatar Feb 07 '24 20:02 sarnex

@aelovikov-intel If we only ran it when there were ESIMD changes is that still too long? I'm not sure how many PRs we have per day but probably 1-3. Maybe I'll actually have to use REST.

In that case, it would be the right thing to do, even if that is too long. :)

I'm also wondering how much the two timed-out tests affected total time. Maybe if we can fix that, the overall time would drop as well.

aelovikov-intel avatar Feb 07 '24 20:02 aelovikov-intel

Good idea, let me try disabling those just for testing and see if it's still long. If so, I'll try to implement the code owner idea.

sarnex avatar Feb 07 '24 20:02 sarnex

6m 46s with timed out tests disabled, looks like I'm learning REST ;(

sarnex avatar Feb 08 '24 14:02 sarnex

:white_check_mark: With the latest revision this PR passed the C/C++ code formatter.

github-actions[bot] avatar Feb 08 '24 18:02 github-actions[bot]

@aelovikov-intel Mind reviewing? I think it's ready. In the current run the new runner didn't run because there were no ESIMD changes. There was a previous run here and the only fail is unrelated and failing in HEAD.

sarnex avatar Feb 08 '24 20:02 sarnex

Will address feedback when I'm back from vacation!

sarnex avatar Feb 08 '24 21:02 sarnex

@aelovikov-intel Do you mind taking another look at this? I attempted to address your feedback, but I'm not sure if I did it in the cleanest way.

sarnex avatar Feb 21 '24 18:02 sarnex

@intel/dpcpp-esimd-reviewers Please review this one ASAP, we can have DG2 testing in precommit after this is merged

sarnex avatar Feb 26 '24 16:02 sarnex

Failures not related, see https://github.com/intel/llvm/pull/12816 and https://github.com/intel/llvm/actions/runs/8022089783/job/21917005858

sarnex avatar Feb 26 '24 18:02 sarnex