open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I'm running the latest telemac version, and I'm having trouble running the 3D 'source' example case in parallel (without any modifications to the .cas file). While other example cases work fine in parallel (e.g., 'stratif_wind'), 'source' seems to only work in serial mode.

is there some special treatment needed for parallelization in source cases ?
I attached my log file for the parallel run.
thanks,
Shai

let me refine that a little,

there's a sensitivity to the ratio between the number of vertical layers and the number of parallel processors.

Hello Shai,

Indeed there is a sensitivity to the number of subdomains when you run parallel computations. In your case, you seem to ask for 80 cores, so that in the subdomain where the output file is written, you only have 32 3D elements which is very low!
Having 2,000 elements or less per core may lead to strange results. Less than 100 for sure.
The geometry file of the TELEMAC-3D source example is not too heavy to be run easily and quickly. It is not built to be run for dozens of core.

If you run this example with 2 or 4 cores, it works.

Chi-Tuan

hi again,

So I took your advice, and generated a finer mesh using BK.
now the run crashes for some other reason. I believe it has something to do with the the vertical location of the point source. The mesh seems to be generated correctly, with the dimensions I set. The source point is supposed to be somewhere within the domain boundaries. However, the report (out.txt) says its above the water surface.


what am I missing here?

Hello Shai,

When you refine the 2D mesh, the size of triangles is often smaller and that often leads to a need to use a smaller time step... If using a time step like 0.1 s, your computation runs.

If you look a the output files PE*LOG in the temporary directory, in particular the ones with a size different from the others (the suspicious ones), you can read that you have crossed planes: your first ideas to solve is to reduce the time step.

Hope this helps,

Chi-Tuan

Hi Chi-Tuan,
your solution worked fine.
However, the core of the problem remains. I am unable to place the source point between the bottom plane and the water surface. because of that, the mesh is being stretched at the the first timestep to reach the source, which causes the layer crossings.

Hello Shai,

As written in the TELEMAC-3D user manual:
"TELEMAC-3D places the sources on the nearest mesh level. In this case, it is recommended to use fixed levels at sources elevations in order to avoid unwanted vertical movement of the sources during the simulation."
You should try to use MESH TRANSFORMATION = 3 and define a fixed level in the user_mesh_transf subroutine.

Hope this helps,

Chi-Tuan

It worked great CT, thank you very much.