Hello everybody,
I have some problems when running ROMS with MPI.
I am using the latest version of ROMS for the MEditerranean sea. I compared the results obtained with one and four processors and the results start to diverge quickly.
This divergence starts along the tile boundaries.
The velocities are the first variables where this difference appears first.
I have also checked the UPWELLING case and while there is not difference in the history files, the v tendencies (v_prsgrd, v_vadv, .. etc) also are different in the cyclic boundary points.