Welcome, Guest
Username: Password: Remember me
  • Page:
  • 1
  • 2

TOPIC: v6p3 problems with parallel mode

v6p3 problems with parallel mode 9 years 11 months ago #15262

Hello,

I've a v6p2 version of the Telemac suite running perfect in parallel mode (somebody else did the set up). Now I'm trying to install v6p3 with perl, because the v6p2 was installed with perl as well. However the system runs on a single processor but not in parallel mode. Therefor I've some basic questions:
As the manual says that the INCMPI should point to the include directory of MPI-2 implementation I've the question: where do I finde the MPI-2 Implementation. I just copied the "mpi" folder from the v6p2 version where I found the "gfortran_linux" subfolder to the v6p3 folder and changed the path of the INCMPI to the new folder. Is this the right way?
Should LKLIB just refer to the libmetis.a or are to other places as well?
The compilation process with makeall90 and the justmake command works without errors.

PS: Maybe I should wait for the v7 release, but I think then I'll have the same problem

Thank you
Engelbert
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 11 months ago #15271

  • yugi
  • yugi's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 851
  • Thank you received: 244
There are some modifications to do in the systel.ini file between the v6p2 and the v6p3.

Could you post your systel.ini here so i can give you more informations.

Hope it helps.
There are 10 types of people in the world: those who understand binary, and those who don't.
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 11 months ago #15285

Hello,

after I did some changes in my systel.ini from the v6p2 version parallel mode isn't running anymore with v6p2 even if I recompile with the older systel.ini which used to work well.

I've attached the error logs from version v6p2 and v6p3 and my systel.ini from version v6p3.

Hope someone can help me
In single mode both versions work

Thank you
Engelbert
Attachments:
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 11 months ago #15286

sorry here the systel.ini from v6p3
Attachments:
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 11 months ago #15317

Finally I managed to compile my v6p2 setup to work well in parallel mode but the v6p3 still does not.
FC_MPI, LK_MPI, LIBS_MPI, RUN_MPI in the systel.ini from v6p2 and v6p3 refer to the same locations.
The computation starts, but not in parallel, as you can see in the attached file.
Maybe someone can tell me where to search to fix the problem.

Thank you
Engelbert
Attachments:
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 11 months ago #15321

  • jmhervouet
  • jmhervouet's Avatar
Hello,

There is no error message, so this is strange, the program should go on. You can add:

DEBUGGER = 1

in your steering file to see more precisely in which subroutine you stop.

Another problem is that all the processors write their listings on the same page, this is very strange, normally you should see only processor 0, the others are redirected to files...

Regards,

JMH
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 11 months ago #15327

Hello JHM,

thank you for your answer even it is the weekend before christmas. Sorry for the inaccurate post (#15317). The program does go on, i just put the relevant output in the post. But the simulation is running very slow and not as fast as with v6p2 on 4 processors.

thank you
Engelbert
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 10 months ago #15363

  • yugi
  • yugi's Avatar
  • OFFLINE
  • openTELEMAC Guru
  • Posts: 851
  • Thank you received: 244
Hi the error seems to come from your cas file.

You have the following message in your run output:
 DIFFERENT NUMBER OF PARALLEL PROCESSORS:
 DECLARED BEFORE (CASE OF COUPLING ?):           0
 TELEMAC-2D :           4
 VALUE            0  IS KEPT
 DIFFERENT NUMBER OF PARALLEL PROCESSORS:
 DECLARED BEFORE (CASE OF COUPLING ?):           0
 TELEMAC-2D :           4
 VALUE            0  IS KEPT

So the code is run 4 times in sequential instead of once in parallel.

Could you post your cas files.
Hope it helps
There are 10 types of people in the world: those who understand binary, and those who don't.
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 10 months ago #15365

  • jmhervouet
  • jmhervouet's Avatar
Hello,

No, in case of coupling, only the telemac-2D value is looked at, the keyword given in the coupled program is discarded. I would rather think that it is a problem of configuration. After launching all the processors, their number is retrieved through mpi. When there is a problem, for example here if no processor is running, we find that there is a difference between 4 and the real number of running processors and the process is stopped. The real problem is here that the 4 runs are launched on the same processor.

With best regards,

Jean-Michel
The administrator has disabled public write access.

v6p3 problems with parallel mode 9 years 10 months ago #15387

Hello all,

I attached my files. Maybe you could check if there is a mistake in my cas file, but it is running well with v6p2.

Thank you
Engelbert
Attachments:
The administrator has disabled public write access.
  • Page:
  • 1
  • 2
Moderators: borisb

The open TELEMAC-MASCARET template for Joomla!2.5, the HTML 4 version.