open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear all,

Is it possible to impose about 100 different tracers in the parallell version of Telemac-3D ?
I tried to do so by changing the MAXTRA parameter in declarations_telemac3d.f and then, due to an error, I had also to change the MAXBLOCK parameter.

I re-compiled all the system but at the end I can only impose about 60 different tracers. When I try to impose a larger number of tracers, the calculation crashes at thomps.f. When I try to traceback the error, it is somewhere in the scaract.f subroutine.

Do you have any idea why I get the error? Or how can I impose the 100 tracers?

Kind regards,
Maria

Hello Maria,

Whooof... 100 tracers! Actually we never tried. What you did is correct, but maybe not sufficient, it could well be that one of the arrays used for chacteristics and the buffers in MPI are not sufficiently dimensioned. For example there is a "basket" in streamline.f which is dimensioned for 10 advected variables, and this value 10 is also in a structure for communicating with MPI, so I'm afraid that it will be a large development, and I am surprised that it works for 60 tracers ! However these tracers need not be advected with characteristics and that could save you. So my questions are:

* how many tracers are advected with characteristics ?
* does it work in scalar mode ?

Regards,

Jean-Michel

Hello Jean-Michel,

Thank you for your quick reply and explanation! I tried in scalar mode and it does not work either.
Regarding your first question, I'm using an explicit advection scheme for the passive tracers (scheme advection=14 (I'm working with tidal flats)) since I'm concerned about "mass" conservation. So, if I well understood your question the tracers do not need to be advected with the characteristics.


Kind regards,
Maria

Hello Maria,

OK, I understand that Thompson advects all the tracers, this is why we are limited by the size of the basket in streamline, even if you use another advection scheme for tracers. There is in streamline.fr (bief) a variable MAX_BASKET_SIZE set to 10 : the maximum number of advected variables, for you it should be at least 104, but the problem is that this 10 may be hardcoded elsewhere, e.g. it is in declarations_parallel for defining CHARAC_TYPE with a component BASKET(10). This does not explain that it works with 60 tracers, it should not!

Regards,

Jean-Michel

Dear Jean-Michel.

thank you again for all the suggestions and recommendations.
I'm going to re-check all this.

Kind regards
Maria

Dear Jean-Michel,

Sorry for the delayed answer. You were right: the 60 tracers were not working. I do not understand why the model kept running...Meanwhile I have replaced the value 10 by 104 in variables MAX_BASKET_SIZE (declarations_parallel.f and org_charac_type.f) and BASKET (streamline.f). The code can run with 100 tracers, but results(velocities) start not to be the same when I impose the 19th tracer (I'm imposing passive tracers). Do you any idea why this happens?Do you think that the dimensions defined in org_charac_type1.f for CH_DELTA, CH_BLENGTH and CH_TYPES are related to that?


Kind regards,
Maria