TELEMAC-MASCARET Forum: Parallel computation NAN (1/1)

Welcome, Guest

TOPIC: Parallel computation NAN

Parallel computation NAN 3 years 4 months ago #38791

Deng OFFLINE Fresh Boarder Posts: 4	Hi, I face a problem with the parallel computation using v8p1r1. When I used the parallel processors=48,it runs normally without NaN.But When i used the parallel processors=96, the NaN appears. What may cause such problem？ Best JD
	The administrator has disabled public write access.

Parallel computation NAN 3 years 4 months ago #38792

c.coulet OFFLINE Moderator Posts: 3722 Thank you received: 1031	Hi Hard to say with only this information. Try to increase the precision of solver to 1e-9 or 1e-10 and check You could also upgrade to v8p2 ... Hope this helps
	Christophe The administrator has disabled public write access.

Parallel computation NAN 3 years 4 months ago #38796

pham OFFLINE Administrator Posts: 1559 Thank you received: 602	Hello JD, What is the number of triangular elements? If it is not big enough, 96 processors may lead to problems in your computation. But as Christophe wrote, hard to say with only this information. You should read the rules of this forum (in particular #7). Hope this helps, Chi-Tuan
	The administrator has disabled public write access.

Parallel computation NAN 3 years 4 months ago #38803

Deng
OFFLINE
Fresh Boarder
Posts: 4

Hi, Chi-Tuan

Sorry for the minmum information given. Here is the error information. The total element of the mesh is 551443. I need it to run for a long duration,so I need more processors.

END OF TPXO INITIALISATION
USING STREAMLINE VERSION 7.3 FOR CHARACTERISTICS
PROSOU : IN CARTESIAN COORDINATES, THE CORIOLIS
PARAMETER IS READ IN THE STEERING FILE
IT IS THE KEY WORD 'CORIOLIS
COEFFICIENT', IT IS UNIFORM IN SPACE
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 100 RELATIVE PRECISION: NaN
INITIAL QUANTITY IN SUSPENSION FOR CLASS 1 : 0.000000 M3
INITIAL QUANTITY IN SUSPENSION FOR CLASS 2 : 0.000000 M3
EQUNOR (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: NaN
EQUNOR (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 50 RELATIVE PRECISION: NaN
GRACJG (BIEF) : EXCEEDING MAXIMUM ITERATIONS: 100 RELATIVE PRECISION: NaN

Also, TPXO INITIALISATION takes for up to 5-7 hours at the cluster, which make it difficult for debugging.

Thanks,
Deng

The administrator has disabled public write access.

Parallel computation NAN 3 years 4 months ago #38808

pham
OFFLINE
Administrator
Posts: 1559
Thank you received: 602

Hello,

Do you have the same behaviour without modelling sediment transport?
Do you use TPXO tidal solutions for both initial and boundary conditions?

You can try to reduce tidal data base size be trying what I wrote in topic:
www.opentelemac.org/index.php/kunena/16-...alling-tpxo?start=10

It is not totally equivalent but can reduce the time of TPXO initialisation step.

You should also use debug options in the systel.cfg files.
You can have a look at the $HOMETEL/configs/systel.edf.cfg file and the flags for debug compilations.

Hope this helps,

Chi-Tuan

The administrator has disabled public write access.