open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I'm trying to run a test case where I impose some sources and sinks in my domain through the key words ABSCISSES DES SOURCES; ORDONNEES DES SOURCES; COTES DES SOURCES.

When running in serial mode it works fine. When I switch to the parallel mode, telemac3d returns me that the points that I'm assigning are outside of the domain and therefore at the end it does not assume the source terms that I'm imposing.

Do you have any suggestions to solve this problem?

Kind regards,
Maria

Hello Mary,

It is normal in parallel that the sources points appear to be outside a sub-domain, but the program does not stop for this. What is imposed in parallel is that the coordinates of those points coincide exactly with a point in the mesh (because with some tolerance we could end up in choosing 2 different mesh points in 2 different sub-domains). Anyway in that case the program stops and tells you what exact coordinates you should put in your steering file (we did that for all test cases in the examples).

This is general explanations that do not fit quite to the description of your problem, but it may shade some light on it, to go further...

With kind regards,

Jean-Michel Hervouet

Dear Jean-Michel,

Thanks for the quick reply.

I well understood that I have to assign the exact coordinates. And that is done. The model does not stop, but it says at the beginning (from the proxim.f subroutine) that the points I'm assigning are outside the domain (WRITE(LU,*) 'POINT SOURCE OU SPECTRE ',K,' HORS DOMAINE'). Therefore the model continues running but it does not take into account the sources I'm imposing.

Could it be a problem inside proxim.f subroutine? The coordinates given in the steering file are treated for the global or local sub-domains?

Kind regards,
Maria

Hello Maria,

No, the message is normal if the program does not stop, a sub-domain at least has found that the point is inside it. You should see in the listings that the points are recognised by at least one sub-domain (but the listings of processors others than 0 are erased if the run is OK, so not easy...). A solution could be to put subroutine proxim (from bief) in your FORTRAN FILE, and change the WRITE(LU,*) into PRINT*, so that the messages are not redirected.
Then the problem is to understand whether these sources are taken into account or not, at least you should see the discharges in the mass balance, even in processor 0 listing (the mass balance sums all processors, so is the same on all processors).

I hope it helps...

Jean-Michel

Dear Jean-Michel,

You are right! When getting the first message saying that the sources were out of the domain I thought that the model would no take them into account, but actually it does!

Thanks for your help!

Best regards,
Maria