[MITgcm-support] parallel computing with mitgcm

Martin Losch Martin.Losch at awi.de
Tue Jan 25 04:48:20 EST 2011

Hi Asic,

I am not qualified to answer your question, but nobody else seems to be answering.

The MITgcm uses MPI, but multithreading is also possible. All subdomains (tiles), however, have to have the same size.

The MITgcm code is in F77 and all fields and infrastructure are allocated statically, that's why you have to have all parameters available in advance. Given that I do not think (but I am not qualified) that it is possible to work with the MITgcm in a cloud-like enviroment (as little as I understand what that is). 
We have used the MITgcm on a system that used MOSIX some 9 years ago (http://www.mosix.org/), but without MPI parrallelization.


On Jan 24, 2011, at 8:51 PM, Azbayar Demberel wrote:

> Hi all,
> I'm a total novice and I'm trying to understand the parallelism primitives supported in mitgcm. Can anyone explain to me why is it necessary to provide the number of parallel processors in advance? Or to quote from online documentation: "The numeric value of number of processors np must be equal to the product of the processor grid settings of nPx and nPy in the file SIZE.h"
> Is there a version of mitgcm that supports runs on a grid/cloud-like environment, where experiments are divided into multiple jobs, which can be queued and executed once resources become available?
> Thanks,
> Asic
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support

More information about the MITgcm-support mailing list