open TELEMAC-MASCARET The mathematically superior suite of solvers

Dear all,

I'm trying to setup a model to propagate wave conditions from offshore to the coast. The model is attached in the zip-file.

The model seems to run fine in serial mode, and runs parallel on up to 7 cpu. When I try to run the model on 8 (or more) cpu it crashes (while reading the frequency of the first time step?).
I tested several versions (v6p3r0, v7p2r3, v7p3r1), each time editing the limwac.f routine.

For each version I got a different error message, but all run on up to 7 cpu and no more.
v6p3r0: |Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
v7p2r3: |STOP 1
v7p3r1: |application called MPI_Abort(MPI_COMM_WORLD, 2) - process 5 (log-file attached)

For what it's worth, I can run the same mesh in telemac2d with TPXO boundary conditions on up to 32 cpu (maximum available to me at the moment).

Any help on pinpointing why this parallelisation limit in TOMAWAC is reached would be much appreciated.
I noticed there is a slight difference in maximum number of elements and points around a point when I run in serial or parallel mode (9 and 9 versus 9 and 10,see log-files attached)?

Hello,

I think your problem comes from the opening and the reading of your formatted ascii files on hard-coded channel (800,..). In parallel mode, each processor try to open the files on the same reading unit and may read the same file reaching its end prematurely.
Even with 7 procs, your results may not be correct..

I think you should try to gather your boundary conditions on the same file and then read them thanks the "FORMATTED FILE 1" in your steering file, and with "READ(NFO1,*)" command on limwac subroutine.

Regards,
Laurent