CI: Split Unstructured <-> DGMulti; Parabolic

[DanielDoehring]

The "unstructured_dgmulti" and "parabolic" tests run almost twice as long as the other tests (over an hour compared to more than 30 min.) This PR aims to reduce CI time by parallelizing these tests.

[github-actions]

Review checklist

This checklist is meant to assist creators of PRs (to let them know what reviewers will typically look for) and reviewers (to guide them in a structured review process). Items do not need to be checked explicitly for a PR to be eligible for merging.

Purpose and scope

• The PR has a single goal that is clear from the PR title and/or description.
• All code changes represent a single set of modifications that logically belong together.
• No more than 500 lines of code are changed or there is no obvious way to split the PR into multiple PRs.

Code quality

• The code can be understood easily.
• Newly introduced names for variables etc. are self-descriptive and consistent with existing naming conventions.
• There are no redundancies that can be removed by simple modularization/refactoring.
• There are no leftover debug statements or commented code sections.
• The code adheres to our conventions and style guide, and to the Julia guidelines.

Documentation

• New functions and types are documented with a docstring or top-level comment.
• Relevant publications are referenced in docstrings (see example for formatting).
• Inline comments are used to document longer or unusual code sections.
• Comments describe intent ("why?") and not just functionality ("what?").
• If the PR introduces a significant change or new feature, it is documented in NEWS.md with its PR number.

Testing

• The PR passes all tests.
• New or modified lines of code are covered by tests.
• New or modified tests run in less then 10 seconds.

Performance

• There are no type instabilities or memory allocations in performance-critical parts.
• If the PR intent is to improve performance, before/after time measurements are posted in the PR.

Verification

• The correctness of the code was verified using appropriate tests.
• If new equations/methods are added, a convergence test has been run and the results
  are posted in the PR.

Created with ❤️ by the Trixi.jl community.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.88%. Comparing base (f5c2c75) to head (f06cd42).
Report is 6 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #2234   +/-   ##
=======================================
  Coverage   96.88%   96.88%           
=======================================
  Files         490      490           
  Lines       39491    39491           
=======================================
  Hits        38257    38257           
  Misses       1234     1234           

Flag	Coverage Δ
unittests	96.88% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ranocha]

It looks like we just gained 15 minutes or so:

• unstructured - ubuntu-latest - x64 (pull_request) Successful in 34m
• dgmulti - ubuntu-latest - x64 (pull_request) Successful in 50m
• parabolic_part1 - ubuntu-latest - x64 (pull_request) Successful in 21m
• parabolic_part2 - ubuntu-latest - x64 (pull_request) Successful in 38m
• parabolic_part3 - ubuntu-latest - x64 (pull_request) Successful in 45m

from a bit more than an hour to 50 minutes for dgmulti and 45 minutes for parabolic stuff.

I'm fine trying to speed up CI. However, a lot of time is spent compiling stuff. What do we get when we merge the first two parabolic CI jobs?

[DanielDoehring]

We could also group 1D & 3D for the DGMulti and split this from the 2D tests.

[ranocha]

MPI on Ubuntu still takes 50 minutes, limiting the CI runtime. Would you like to split it or shall we group other jobs since 50 minutes is already fine?

Did you check whether all jobs started together or whether we need to wait for some jobs to finish before new CI resources become available?

[DanielDoehring]

We could also split the MPI tests into 2D/3D or p4est/t8code, yes.

Right, in general there are some other CI jobs running and CI times of individual Jobs differ significantly from run to run.

[DanielDoehring]

Xref: #2197

[vchuravy]

Maybe unecessary after #2254 ?

[DanielDoehring]

So the imbalance between runs is still there (Although everything is now quicker)

[JoshuaLampert]

The longest run with 26 minutes is parabolic_part3. In this run, there is one elixir (elixir_navierstokes_taylor_green_vortex_amr.jl), which needs more than 5 minutes alone (more than 4 without compiling). So in order to speed up CI, we could also reduce the final simulation time of this elixir.

[ranocha]

The longest run with 26 minutes is parabolic_part3. In this run, there is one elixir (elixir_navierstokes_taylor_green_vortex_amr.jl), which needs more than 5 minutes alone (more than 4 without compiling). So in order to speed up CI, we could also reduce the final simulation time of this elixir.

Maybe we could already do this in another PR?

Edit: Done in #2267

[JoshuaLampert]

This can probably be closed after merging #2267 and #2268 as now (without this PR) building the documentation takes as long as the longest CI run. So splitting these doesn't really reduce the longest test run, but increases the total run time due to additional setup and precompilation.

[DanielDoehring]

This can probably be closed after merging #2267 and #2268 as now (without this PR) building the documentation takes as long as the longest CI run. So splitting these doesn't really reduce the longest test run, but increases the total run time due to additional setup and precompilation.

Yeah right. Do we know why the documentation run takes do long now?

[JoshuaLampert]

Didn't it always take so long?

[DanielDoehring]

Didn't it always take so long?

If that's the case I didn't notice, I thought it is like 5 min max 🤔
But I can very well be mistaken!

[ranocha]

The last dozens of builds took between 20 and 40 minutes, depending on the CI system etc.

[JoshuaLampert]

Let's close this for now. If you think, it's still useful, feel free to reopen @DanielDoehring.