Skip to content

Add serial and 64-process Channel test to CI#31

Closed
loganoz wants to merge 3 commits intodevelopfrom
feature/WallFuncionParallelTest
Closed

Add serial and 64-process Channel test to CI#31
loganoz wants to merge 3 commits intodevelopfrom
feature/WallFuncionParallelTest

Conversation

@loganoz
Copy link
Owner

@loganoz loganoz commented Nov 21, 2025

Expanded the CI workflow to run the Channel test in serial (1 process) and with 64 processes, in addition to the existing 8-process test. This improves coverage of different parallel configurations for the Navier-Stokes Channel solver.

Expanded the CI workflow to run the Channel test in serial (1 process) and with 64 processes, in addition to the existing 8-process test. This improves coverage of different parallel configurations for the Navier-Stokes Channel solver.
@loganoz loganoz requested a review from psole-m November 21, 2025 12:56
@loganoz
Copy link
Owner Author

loganoz commented Nov 21, 2025

@psole-m, this pull request just runs develop branch (so old wall model) with 1, 8 and 64 MPI processes. Let's check if the old implementation gave the same results where it run (at least with 1 and 8 proc).

@loganoz
Copy link
Owner Author

loganoz commented Nov 21, 2025

@oscarmarino, is this behaviour expected?

@oscarmarino
Copy link
Collaborator

I think it should be the same, it is strange that we have different results. In principle the wall functions here (old one) is MPI independent, so it could be something in the channel module (I tested it in a different mesh and only a few partitions...)

Introduces a new Channel-no-source test for Navier-Stokes, including control and setup files. Updates CI workflow to run this test in serial and parallel modes, and adds it to the Solver test configuration.
@loganoz
Copy link
Owner Author

loganoz commented Nov 24, 2025

@oscarmarino, @psole-m I created a new test case Channel-no-source. Here I set "use channel = .false.". The runs are with 1, 8 and 64 processes. I didn't update the residuals, so I don't expect it to pass. However, if the residuals do not change, the problem is with the parallelisation of the channel source.

@loganoz loganoz closed this Feb 16, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

Comments