[MITgcm-support] Artificial Zonal Symmetries

[Martin Losch]

Hi Alon,

how do you compile and run the model? I think you need to do something like this:

$rootdir/tools/genmake2 -of $youroptfile -mods $yourcode -mpi

Don't forget the -mpi, that tells genmake2 to generate a makefile that will include the mpi-related code, then

make CLEAN && make depend && make etc.

mpirun -np 36 mitgcmuv_ad

unrelated comments:
- why is Oly .ne. Olx? I would always have the two overlaps have the same value.
- why do you use nSx/y=2? Are you trying to do multithreading? I would start with one tile per cpu (npx), so nsx/y=1 (and adjust snx/y accordingly) and once that works you can try 4 tiles per cpu (although it has been explained to my about a million times why this can be useful, I still have not seen an application where this actually increases the overall performance, ie. time to solution).

Martin

On Jun 4, 2012, at 6:10 PM, Alon Stern wrote:

> Hello
> 
> My name is Alon Stern, and I am a Phd student at the Courant Institute
> at NYU. I have been using MITGCM for the last few months to try
> simulate the Antarctic Circumpolar Current.
> 
> I have set up the model to be a channel: Solid (no slip) boundaries in
> the north and south, and periodic boundaries on the east and west of
> the domain. I am forcing the fluid with a wind stress which is
> constant in the x-direction.
> 
> I am using the following setup in my SIZE.h file
> 
> sNx=35
> sNy=50
> OLx=4
> OLy=2
> nSx=2
> nSy=2
> nPx=6
> nPy=6
> Nx=sNx*nSx*nPY
> Ny=sNy*nSy*nPy
> Nr=25
> 
> 
> The problem that I am facing is that the results of the simulation
> come out to be zonally periodic in the x-direction, with 6 copies of
> the same pattern in the domain. Ie: I see the same eddy pattern repeat
> 6 times in the zonal direction. It seems that this is occurring
> because nPx=6. I have done tests with other amounts of processors and
> I get the same issue with the same pattern repeating a number of times
> equal to the number of processors in the x-direction.
> 
> I have tried to change the number of processors in the middle of a
> simulation, and in this way I can lower the number of symmetries down
> to 2, but I can not get rid of them completely.
> 
> Do you have any idea what is causing this effect, and how I can get
> rid of it? Please let me know if I have not been sufficiently clear
> with describing the problem.
> 
> Thanks
> Alon
> alon at cims.nyu.edu
> 
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support