open TELEMAC-MASCARET The mathematically superior suite of solvers

it worked!!!!!! I cannot believe it. So my problem was ubugfopenmpi in config.
Ok so it means that I can extract my cross sections now. That is great news!

Ok, I still have problem with parallel running, but I will try to fix it later somehow. I will leave you in piece now. Thank you so much for your help. I will keep my fingers crossed that my postel will work for bigger cases as one i used now is very short. I will let you know. Thanks again. If you do not hear from me that everything is fine and have a nice week-end!

Violeta

Hi Christophe,

Postel worked on scalar even for larger case.

Just a quick question about running in scalar on cluster. I try to run very big case, well in terms of dimensions it is ok: 500x500x150 all in metres, but in terms of mesh it is very fine: it is horizontally 1m mesh for 120m and then 10m, but verticaly I have 500 layers so it is 1m evenly in vertical, so it is very dense. I could not run such case on my desktop but I hoped that I could do it on cluster in scalar, since I still have problems with parallel settings. I have just tried but it fails to run without any error message. Do you think it is possible to run such a big case in scalar or it has to be in parallel only? Just would like to know you expert opinion.

Thanks again for all your help for last two days.

Kind Regards!

Violeta

Violeta

Greats for postel3d. This definitely means postel is not parallelized and should absolutely be run in scalar mode.

About your case, it's a little bit more complex as it's a very huge mesh.
Running such model in scalar will probably require to much memory ( but this could be checked).
If so this means you have to run it in parallel.

For this particular problem you met with parallelism, I think you should create a specific topic.
Before that, check if it runs in scalar (this allow you to see if the steering file is ok).
Then try to run it in parallel for a small duration (10 time step for example). You could also run different tests, with different number of layers to see if there is a limit. Probably the number of processors will have an influence.
This will allow you to give the maximum informations to the community in order to help you to solve this problem.

Good luck with this new challenge

Thank you so much Christophe for your help and very useful advice. I will do some testing first and then will put a new post on forum. Thanks again.

Have a fantastic week-end!

Violeta