use mixin class for espresso

[smoors]

No description provided.

[casparvl]

Code looks good. I'm doing a full run on both the main and feature branch for this test now to see if there are any differences.

One thing I've already noticed is that on main, the CI tag was only set on test cases smaller than 2 nodes (i.e. 1 node, or partial node). In the feature branch, it's set on all scales. I don't think that's desirable, I think this test takes longer and longer at larger scales, and I think we originally set that CI tag only on the 1-node scale to limit execution to reasonable time (full node takes around 15 minutes on a 128-core Rome node). So, we should probably keep that behaviour.

[casparvl]

To come back to my own remark: the 8-node case has been running for 1 hour and 15 minutes already, and isn't finished. So, yes, I think we should keep setting the CI tag only on the single-node cases, as we did before, to ensure only the test variants with reasonably short runtimes are run.

In terms of test performance, I'm seeing similar performance compared to main branch, so that's good.

[smoors]

no problem, i'll add it back, and i'll add a clear comment why we're doing this.

[smoors]

no problem, i'll add it back, and i'll add a clear comment why we're doing this.

ok done, we're now also adding a tag on the bench_name in the mixin class, because why not, right?

ready for re-review :)

[smoors]

ready for re-review :)

now for real!

[casparvl]

Tested using

reframe -c eessi/testsuite/tests/apps/espresso/espresso.py -t CI --run --system=snellius:rome -n 4.2.2-foss-2023b

And all seems to be fine. Approved!

[casparvl]

Oh @smoors there's a conflict with one of your earlier PRs that I already merged. If you solve it, I'll double check again and merge :)