open TELEMAC-MASCARET The mathematically superior suite of solvers

Hello, I am having some problems with a telemac3d simulation using parallel processors. The model runs without a problem in 2d using a liquid boundaries file. However I get the following error in 3d

NUMBER OF LIQUID BOUNDARIES: 2

================================================================================
ITERATION 0 TIME 0 D 0 H 0 MN 0.0000 S ( 0.0000 S)
================================================================================

---

MASS BALANCE
INITIAL MASS OF WATER IN THE DOMAIN : 7257946119.16674
SOURCE POINT 1 OUTSIDE DOMAIN
SOURCE POINT 2 OUTSIDE DOMAIN
SOURCES : Q(1) = 51.5000000000000
SOURCES : TR(1,1) = 0.000000000000000E+000
SOURCES : TR(1,2) = 20.0000000000000
USING STREAMLINE VERSION 6.2 FOR CHARACTERISTICS
THE LIQUID BOUNDARIES FILE CONTAINS
4527 LINES WITH:
SL(1) VTU(2) VTV(2)
forrtl: severe (174): SIGSEGV, segmentation fault occurred
Image PC Routine Line Source
out17222_intel64. 000000000040550B Unknown Unknown Unknown

Stack trace terminated abnormally.
rank 0 in job 67 deepblue.corp.cefas.co.uk_50191 caused collective abort of all ranks
exit status of rank 0: return code 174
## Erreur : Fin anormale : cd /home/ow02/projects/Sizewell/t3d/size_t3d_z2.cas17222_tmp; /export/apps/s/mpich2/bin/mpirun -n 2 ./out17222_intel64.exe :44544
## Error : System command failed for /home/ow02/projects/Sizewell/t3d/size_t3d_z2.cas17222_tmp/telemac3d.bat :44544
________________________________________________________
Execution finished: telemac3d.bat
________________________________________________________
No compilation/linking/file errors detected.
Execution errors detected.
Please see messages in stdout above or study stderr output.

Working directory: /home/ow02/projects/Sizewell/t3d/size_t3d_z2.cas17222_tmp
can be manually deleted with: /home/ow02/projects/Sizewell/t3d/delete_size_t3d_z2.cas17222

Returning exit status 1

=========================================================
Telemac System 5.6 to 6.2 - Perl scripts version 6.2
=========================================================
...stopping.

The 3d simulation runs fine in scalar with no parallel processors but the simulation time is far too long. Is this a problem with the way the liquid boundaries file is split up by telemac for the parallel processors? Does anyone have any experience with this?

Thanks

Olly

Hello,

This should raise no problem in parallel, but VTU and VTV are not standard, so you probably developed something special to recover VTU and VTV, that would not work in parallel ? So your fortran file would be helpful.

Regards,

JMH

Hello, thank you for the response. I have attached my fortran file to this message which shows the addition of VTV and VTU. However, I have also tried using the standard VIT3 subroutine with velocity magnitude in the liquid boundaries file rather than U and V and I get the same error. I have also tried forcing the liquid boundaries file on one boundary but I also get a similar error. Any help is very much appreciated.

Thank you

Olly

Hello,

I looked at your Fortran file. It seems that the array FRICOEF is hardcoded with boundary point numbers of the original MESH (moreover it is an automatic array of size NPTFR2). This is probably a (or the) cause of error. Think that all numbers are changed in // and that a sub-domain may even have 0 boundary points. To give the values to the local boundary points, use the array BOUNDARY_COLOUR%I(K) which is the original number of boundary point K. So to set the value of point 247, do the loop:

DO K=1,NPTFR2
IF(BOUNDARY_COLOUR%I(K).EQ.247) FRICOEF(K)= value of original point 247
... idem for other points
ENDDO

I hope that helps,

JMH

Ok thanks for your advice I will add that to the code.

Olly