open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi,

I use sources in a 3D model in a new subroutine called culverts (presented at the user conference). It works fine, but I noticed something strange: when I run the model with the sources, I get nice results. When I don't change anything and I run again, the some sources don't seem to work anymore. I'm running in parallel mode and noticed that when the source are on the boundary of a subdomain, they don't work anymore.

Is there a way I can add some code to make it work anyway or can I code something to prevent the sources are on the boundary of a subdomain?

For now I just have to be lucky the sources are within one subdomain.

Thanks,

Sven

Hello Sven,

We would have to look precisely at the Fortran sources, and there is certainly some parallel communication lacking somewhere. As I am not currently at work, I will not be able to look at this before mid-February.

With best regards,

Jean-Michel Hervouet

Thank you Jean-Michel

For the moment is there a way to work around this problem? When starting a simulation we depent on luck that the partitioning of the mesh will be good. Can I control the partitioning? I looked into METIS but did not find anything directly. Another idea was to save the EPART array from the partitioning subroutine of a succesfull run and use that agian for future simulations...

Sven