open TELEMAC-MASCARET The mathematically superior suite of solvers

Hi

I am using version 6.3 of Telemac2D , and I am calculating the flow through cross sections which are defined in the file :
SECTIONS INPUT FILE
It works fine in scalar with both options:
• node numbers are given, and the second line of the file is: 4 -1
• coordinates are given, and the second line of the file is: 4 0

and the file SECTIONS OUTPUT FILE is properly generated.

But when I try to run in parallel I get the error message below.

I also noted that in the temp directory all files T2DSEC00003-00000 are there, but they are empty.

Can you help please?


job aborted:
rank: node: exit code[: error message]
0: thierry1.45woodburn.faure.ca: 2: process 0 exited without calling finalize
1: thierry1.45woodburn.faure.ca: 2: process 1 exited without calling finalize
2: thierry1.45woodburn.faure.ca: 2: process 2 exited without calling finalize
3: thierry1.45woodburn.faure.ca: 2: process 3 exited without calling finalize
_____________
runcode::main:
C:\Detroit:
|runCode: Fail to run
|C:\opentelemac-mascaret\mpich2\bin\mpiexec.exe -wdir C:\Detroit\cas_section.
dat_2013-12-16-21h28min41s -n 4 C:\Detroit\cas_section.dat_2013-12-16-21h28min41
s\out_tele_Detroit_C4_v6p3.exe
|~~~~~~~~~~~~~~~~~~
|At line 62 of file C:\opentelemac-mascaret\v6p3\sources\telemac2d\read_secti
ons_telemac2d.f (unit = 45, file = 'C:\Detroit\cas_section.dat_2013-12-16-21h28m
in41s\T2DSEC00003-00001')
|At line 62 of file C:\opentelemac-mascaret\v6p3\sources\telemac2d\read_secti
ons_telemac2d.f (unit = 45, file = 'C:\Detroit\cas_section.dat_2013-12-16-21h28m
in41s\T2DSEC00003-00000')
|Fortran runtime error: End of file
|Fortran runtime error: End of file
|At line 62 of file C:\opentelemac-mascaret\v6p3\sources\telemac2d\read_secti
ons_telemac2d.f (unit = 45, file = 'C:\Detroit\cas_section.dat_2013-12-16-21h28m
in41s\T2DSEC00003-00002')
|Fortran runtime error: End of file
|At line 62 of file C:\opentelemac-mascaret\v6p3\sources\telemac2d\read_secti
ons_telemac2d.f (unit = 45, file = 'C:\Detroit\cas_section.dat_2013-12-16-21h28m
in41s\T2DSEC00003-00003')
|Fortran runtime error: End of file
|~~~~~~~~~~~~~~~~~~

C:\Detroit>

Hi Thierry

All I could say is your problem is in the domain decomposition step (partel).
Maybe you could try to run it manually to check is there is an execution error message

Good luck

Thank you Christophe

All I can say is that in the output screen, prior to the compilation, the file is properly “copied” into T2DSEC in the temp directory, at the same time as all the other input files.

There is no indication on the screen , after the compilation, that the file is “duplicated”, whereas there is proper “duplication” of T2DFO2.

Thierry

Thierry

I'm not sure the T2DSEC should be duplicate.

You could try to run manually the partel program in the temporary file for example for the geometry file T2DGEO
There is a question about sections...
Maybe a problem of partel parameter?

Good luck

PS: please update your profile!

Hi Christophe

I do not understand what you mean by “run manually the partel program in the temporary file”. The file in SECTIONS INPUT FILE should be duplicated and not partitioned since it is just a text file similar to the FORMATTED DATA FILE 2.


Has anybody had this problem before: not being able to get the flow through CONTROL SECTIONS in parallel?

Thank you
Thierry

Hello,

I've just recently completed simulations using v6p3r1 and am encountering the same problem. Under scalar all works fine. Under parallel, simulations stops at the beginning. So, I gave up on the control sections feature and will try to compute flows through sections as a post processing step.

If someone has a solution, I would be interested in hearing it.

Pat

Hi Thierry

I'm note sure the section file should be just duplicated...
The fact it was a text file is not sufficient...

In the dictionary, this file is not declared as SCAL like some other files which are duplicated for the parallel computation. It is declared as SECTION which will probably induce another specific treatment.

About partel, when you try to run your case, the files are copied into a temporary directory in which the partitioning operations are made. My proposal, after the crash is to go into this temporary directory and try to run partel on the command line (you should probably enter the full path to partel).
Then partel will run and ask you some informations (which are passed automatically in a normal execution of telemac).
For the geometry file, you just have to enter T2DGEO.
T2DCLI for boundary conditions.
Then the number of subdomains, the method (choose metis).
Then there is a question about sections. Enter the right answer and eventually if required the name of section file T2DSEC and check the result.
If execution fail, this will give us some informations about the origin of the problem, if it's OK, this will probably mean there is a problem with the script.

Hope this helps

Hello

Mybe this could help to resolve the problem.
So far, to give coordinates of control section you need to write 4 0 in the second line of the file. However, this is wrong, because you have to write 4 1 instead (or a strictly positive number).
This was corrected in the new user manual available on the manual section www.opentelemac.org/downloads/MANUALS/TE...r_manual_en_v6p2.pdf

With my best regards

Riadh ATA

Thanks Pat and Riadh for your input. Christophe I will try your method.

In the mean time I have tried several things.

With the existing declaration of the section file, the file is duplicated in the temp directory but they are all empty.

If I modify the dictionary and declare the file PARAL instead of SECTION then it is properly duplicated. It is read properly because the output file shows all the good names of all sections, but all flows are zero.

If I declare the file SCAL, it is not duplicated. It is still read properly because, again, the output shows all the good names. But all flows are zero.


The problem does not seem to be 4 0 or 4 1 in the second line. Because with 4 1 it is looking for integers, and the coordinates are of course real, and it give an error message.

Any more thought on the problem?

Thierry

Following my last post, I have a specific question for the developers of Telemac.

Section 5.2.2 of the manual does not show the example of the case where the coordinates are given.

With the coordinates option, do we give the X,Y coordinates of a point close to the section end node (This is what I did , since it is what we used to do for sources, Telemac looking for the node closest to the point), or do we give the exact coordinate of the node ? In this case how many decimal do we need?

Thanks
Thierry