TELEMAC-MASCARET Forum: Parallel version problem (1/2)

Welcome, Guest

Page:
1
2

TOPIC: Parallel version problem

Parallel version problem 6 years 5 months ago #31655

mcRove	Goodmorning at all, I did a simulation for the propagation of the waves with a very big domain. The results with scalar version are good but the same simulation with the parallel version gave very casual and wrong results. Very non sense results. I can't understand why.
	The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31656

c.coulet OFFLINE Moderator Posts: 3772 Thank you received: 1044	Hi did you use a user fortran? regards
	Christophe The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31662

mcRove

hi,
thanks for the reply.
I compiled openTelemac in linux using gfortran.
i'm running tomawac in parallel. unfortunately, the results of the parallel computation do not match the reuslts of the scalar computation( single processor).

The results of the scalar computation are correct whereas the results of the parallel computation are completely wrong.

If you can provied any help i would be grateful to you.

I attached my systel.cfg file

Attachments:

systel.cfg (2KB)

The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31665

c.coulet OFFLINE Moderator Posts: 3772 Thank you received: 1044	Hi You didn't answer to my question ! In your computation, did you use a "user fortran"? Did you check the Tomawac example runs in scalar and parallel correctly? Regards
	Christophe The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31668

mcRove	I'm sorry, I misunderstood your question. I used a 'user fortran' file called 'limwac.f'. I took this fortran subrutine from the example Manche. I attached the file that I used. R Attachments: limwac.f (15KB)
	The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31669

c.coulet OFFLINE Moderator Posts: 3772 Thank you received: 1044	Did you check the example Manche works in scalar and in parallel with correct results? I had a brief look to this example and it seems there is no parallel run of this case during the validation process...
	Christophe The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31670

mcRove	the example of Manche run in parallel give results that are good. I have to do somethings of particular for work in parallel? Thanks R
	The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31675

c.coulet OFFLINE Moderator Posts: 3772 Thank you received: 1044	Sometimes it could be necessary take care about parallelism because node numbering change but if it works in the Manche case, this means this should be ok on this point... I don't have any other idea regards
	Christophe The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31681

tfouquet OFFLINE Moderator Posts: 307 Thank you received: 120	Hello If you had good results with case manche in parallel that means you were very lucky. Normally it does not work correctly in parallel because of nodes numbering. You fill find an example of using function GLOBAL_TO_LOCAL_POINT to deal with paralllelism in examples/sisyphe/littoral/Tom_user_fortran/bord_wac.f hope it helps.
	The administrator has disabled public write access.

Parallel version problem 6 years 5 months ago #31701

mcRove	hi, i tryed to follow the example and i wrote this subroutine for my case. The results are bad. what i wrong? Attachments: bord_wac.f (2KB) limwac_2018-10-10.f (15KB)
	The administrator has disabled public write access.