[rocketLauncher]

Hello foamers,

I am running the same case across a number of pre-processing tools and always using simplefoam. Up to now, I imported my mesh made on Salome into openfoam, decomposed the mesh, and ran a parallel calculation without issue.

Today I am using cfMesh for the first time. I used this command to speed up the meshing process before I launched my Allrun file.

export OMP_NUM_THREADS = 8,8

Then I did my usual

decomposePar

and

mpirun -np 10 simpleFoam -parallel | tee simpleFoam.log

.

The simulation is unusually slow. Any ideas as to why? The size of the mesh is the same, no issues with convergence, and no changes on the set up on my side. It's quite odd.

Thank you in advance for your insights.

[dlahaye]

For small cell count, overhead in parallel communication is likely to dominate computation (independent of how the mesh is generated).

How many cells in the mesh do you have?

[rocketLauncher]

At the moment, I have about 1 million cells. Previously, I was using around 700k cells with no issues on the parallel computation, so I don't believe the size is the issue.

[dlahaye]

OK.

Does it help here do share statistics for both mesh generation methods and for various number of processors?

[rocketLauncher]

Thank you for the quick reply. I am afraid I didn't fully understand what you meant in your last message. Could you rephrase?

[dlahaye]

Thank you for your reply.

I am looking to the issue you raise as outsider. The input I provide is limited by my outsider role.

Having said this, you raise the issue on the parallel performance of simpleFoam given a mesh generated by cfMesh.

I wonder:

1/ how does performance compare with the meshes generated by Salome?

2/ how is performance affected by the number of processors used?

[rocketLauncher]

Hello,

I always use the same number of processors (10). The performance difference has to do with the computational time. The clock time would be 2 or 3 times slower that with the Salome Mesh.

I have somewhat resolved the issue since yesterday. In the initial simulation I was using an stl file rather than an fms one. Since diving further into the user manual and other posts here on the forum I have managed to use the associated python scripts with Salome 9.12 to generate the fms file.

Thanks for you help.