open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello everyone,

i have problems using a higher number of parallel processors for suspended sediment calculation. For example my model (telemac2d&sisyphe) works fine with my max. available number of 64 processors only for bed load transport (with 8 classes of sediment). If i add suspension(=YES) the simulation does not start if i use more than 8 processors. For 2,4 and 8 processors it works.
If i use 9 classes of sediment i can run the simulation with max. 16 processors.

Does anyone observed the same error? Or is this an internal programm problem that the number of sediment classes refers somehow to the number of processors...

Best regards,

Markus

Hi Markus

Normally, parallel computation are not linked to the caracteristics of computation.
Nevertheless, we know there is actually a limitation in the computation of characteristics for more than 10 variables...
It's probably your situation with the suspended sediment...
This should be confirm by others developers but if this is the reason, the developement of a solution is a huge job...

Hope this helps

Hi Christophe,

first thanks for your reply.
What are this variables? For example: Velocity (u,v), water depth, bed-evolution, Transport-rate (QS),... and also the % of each sediment class.

But then the question is why leads an increase of the classes (from 8 to 9) to an increase of running parallel processors (8 to 16)..

Best regards,

Markus

Hi
I'm not sure and I didn't look into the subroutines to confirm my opinion but I suppose this is only in the management of the suspension so the variables concerned are only the sediment class. There is probably 1 or 2 other varaibles used on the same call to the characteristics algorithm.
For the increase of parallel run, I've no idea but I would be suspicious about the results due to memory problem without crash...

Regards