SolidStateDetectors.jl icon indicating copy to clipboard operation
SolidStateDetectors.jl copied to clipboard

Run CI with two Julia threads active

Open oschulz opened this issue 2 years ago • 13 comments

oschulz avatar May 21 '23 15:05 oschulz

What happens if we set this to 8 threads? ^^ And any idea why we are getting an OutOfMemoryError on x86?

fhagemann avatar May 21 '23 17:05 fhagemann

What happens if we set this to 8 threads?

Well, the default GitHub Actions runner only has two virtual CPUs, so that should work but provide no speed-up.

And any idea why we are getting an OutOfMemoryError on x86

Nope ...

oschulz avatar May 21 '23 17:05 oschulz

I'll re-run all CI jobs to time them again.

oschulz avatar May 21 '23 20:05 oschulz

Am I seeing it right that the ubuntu x64 tests on Julia 1 take almost 4 hours (?)

fhagemann avatar May 25 '23 03:05 fhagemann

Am I seeing it right that the ubuntu x64 tests on Julia 1 take almost 4 hours (?)

Yes, and it seems to be reproducible, and I don't understand why. Before (this PR) it was about 1.5h. But the tests on the other platforms, that also used to take around 1.5 h, do run in about 50 minutes now, which makes sense.

oschulz avatar May 25 '23 08:05 oschulz

I've added test timers to this PR (should be useful in general), let's see where all the time is spent.

oschulz avatar May 28 '23 18:05 oschulz

Tests on Julia v1.9 keep failing with out-of-memory, not sure why (and why only when using multithreading).

Docs failures should be fixed by #341.

oschulz avatar May 29 '23 08:05 oschulz

ubuntu x64 is still significantly slower than the rest..

fhagemann avatar Jan 26 '24 20:01 fhagemann

ubuntu x64 is still significantly slower than the rest..

Yes, massively - but I haven't the faintest intuition on why, it looks really strange.

oschulz avatar Jan 27 '24 09:01 oschulz

How long does it run locally, on two core, x86-64, in comparison? Maybe the GitHub x86-64 runners are just in higher demand and more heavily oversubscribed?

oschulz avatar Jan 27 '24 09:01 oschulz

Let's see whether something has changed here ever since..

fhagemann avatar Sep 18 '24 14:09 fhagemann

@oschulz looks like the tests all speed up except for the one also testing the code coverage. Any idea why?

fhagemann avatar Oct 07 '24 07:10 fhagemann

I rebased this PR onto the current main

fhagemann avatar Oct 27 '24 13:10 fhagemann

The test also testing the code coverage seemed to take significantly longer than the rest. Let's see whether using #434 solves this.

fhagemann avatar Nov 18 '24 16:11 fhagemann

Looks like the number of processors for public repositories was updated to four: https://docs.github.com/en/actions/using-github-hosted-runners/using-github-hosted-runners/about-github-hosted-runners#standard-github-hosted-runners-for-public-repositories

fhagemann avatar Dec 22 '24 18:12 fhagemann

The last force-push was a rebase onto the current main.

fhagemann avatar Dec 22 '24 23:12 fhagemann