open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello,

I have a configuration of TELEMAC3D on v6p2, in MPI mode.
I updated the configuration files for version v7p1r1.

On v6p2, I could calculate on several numbers of processors.

For the first test of the upgraded version, I could not calculate with more than 4 processors.
If I used more than 4 processors, I have the following error during the execution:
runcode::main:
:
|runCode: Fail to run
|/usr/bin/mpiexec -wdir /home/faure/simu/v7/test1/simu_berre_2007_v7_notra.txt_2016-12-15-12h38min30s -n 8 /home/faure/simu/v7/test1/simu_berre_2007_v7_notra.txt_2016-12-15-12h38min30s/out_princi_berre_v7
|~~~~~~~~~~~~~~~~~~
|At line 213 of file /home/faure/opentelemac/v7p1r1/sources/telemac2d/thomps.f
|Fortran runtime error: Index '590' of dimension 1 of array 'it1' outside of expected range (558:1)

In PJ, you can find the .cas file, the log of PARTEL, and the execution log file.
On my open boundary, I have imposed only the water height (with tide data)

Thanks a lot for your help ! :)
Vincent

Hello,

Now we would need the full case to test it.

Your message shows that you are using strict checking options that slow down your runs (though they are useful for debugging). Maybe it would work without (even if there is probably a problem behind this message).

With best regards,

Jean-Michel Hervouet

hello,

Thanks for you answer.
For information, I am a TELEMAC beginner, and moreover, this configuration has been developped by someone else (on v6p2).

What are these "strict checking options" ?
How can I send your the files of my case (I would prefer in a private way)

Thanks !
Vincent

Hello,

The "strict checking options" are the compiler directives, which depend on your compiler. You may ask to trace back the errors, to check array dimensions, etc. For example in my systel.cfg file I have:

cmd_obj: mpif90 -c -O0 -g -fbounds-check -ffpe-trap=invalid,zero,overflow -fbacktrace -Wall -fconvert=big-endian -frecord-marker=4 -DHAVE_MPI <mods> <incs> <f95name>

because I mostly do developments and want to see all the errors. For a study it would be faster with only:

cmd_obj: mpif90 -c -O3 -g -fconvert=big-endian -frecord-marker=4 -DHAVE_MPI <mods> <incs> <f95name>

where moreover I have set the optimisation level at 3 instead of 0.


To contact me you can ask to be a contact if you go to my profile, then I can send you personal messages and give you my E-mail.

With best regards,

Jean-Michel Hervouet

Hello,

I think the problem comes from:

I suppose that in your case 2*NPTFR > NPOIN.

Did you send the full case to Jean-Michel?

Best regards,
JF

Hello,

I have send my email to Jean-Michel thanks to a private message, and i am waiting for his answer.

If you are right, is there any solution ?

thanks !
Vincet

Hello,

You can try to replace point_telemac3d.f in your_telmemac_dir_v7p1/source/telemac3d by the one in attachment.

And recompile TELEMAC(run compileTELEMAC.py).

Regards,

Hello !

I have upgrade my configuration to V7P2R0.

I have tested the proposition from JacquesFontaine by modifying a line in point_telemac3d.f :
SIZ=MAX(2*NPTFR2,BIEF_NBMPTS(IELM3,MESH3D))

However, I still have the same error as previously.

Thanks !
Vincent

Hello,

Could you send me your setup? I'll try to reproduce your bug...

Best regards