jmc at ocean.mit.edu
Tue Dec 7 15:56:25 EST 2010
I would recommand to use MPI with a number of processors equal
to the number of tiles (minimum number of tiles with cubed-sphere
grid is 6, so should try with at least 6 procs).
The reason is that there is a problem somewhere in the vegetation
part of fizhi code if using multiple tiles per procs (nSx*nSy > 1).
On Tue, Dec 07, 2010 at 03:30:46PM -0500, Bruno Deremble wrote:
> I am trying to run the fizhi tutorial examples. But when I compile mitgcm with ifort, the run crashes very quickly (during the call of the sorad subroutine): I obtain a 'segmentation fault'
> I do not observe this problem with gfortran. But for several reasons, I still would like to use ifort.
> Thank you for your help
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
More information about the MITgcm-support