[MITgcm-support] nPx vs. nPy

Gus Correa gus at ldeo.columbia.edu
Fri Jul 27 11:26:53 EDT 2012


Hi Taimaz

Another detail is that you need to recompile the code
whenever you change the number of processors and subgrid sizes
in SIZE.h, as most memory is allocated statically.
Well, at least it used to be this way.
Did you recompile?

Gus Correa

On 07/26/2012 05:38 PM, taimaz.bahadory wrote:
> My total grid is 130x100, which matches a Nx=13x1x10 and Ny=100x1x1.
> As I said, when I play with y-direction (i.e. changing nPy to from 1 to
> 10 and sNy from 100 to 10, while nPx is 1 and sNx is 130), everything
> works fine. The problem arises when I try to increase the nPx to
> something else than 1, also I'm careful of the total product to match
> the total grid size mentioned in data file.
>
>
>
> On Thu, Jul 26, 2012 at 6:21 PM, Gus Correa <gus at ldeo.columbia.edu
> <mailto:gus at ldeo.columbia.edu>> wrote:
>
>     Hi Taimaz
>
>     Nx, Ny, Nr are the number of points in your [full] domain,
>     which should match the domain size in your initial condtion files,
>     etc, so they should be fixed for each experiment you run.
>     If there is no match the run fails.
>
>     Given Nx, you need to find out the the values
>     nPx, and sNx such that their product is Nx
>     [assuming nSx is 1, i.e. you're not using threads/OpenMP].
>     Likewise for nPy, sNy, etc.
>
>     Note that in SIZE.h is defined as
>     Nx  = sNx*nSx*nPx,
>     and Ny  = sNy*nSy*nPy.
>     Hence, it is you need to choose the
>     right factors to get the expected products Nx and Ny.
>
>     I hope this helps,
>     Gus Correa
>
>
>     On 07/26/2012 03:56 PM, taimaz.bahadory wrote:
>
>         Attached
>
>
>         On Thu, Jul 26, 2012 at 5:17 PM, Chris Horvat
>         <horvat at fas.harvard.edu <mailto:horvat at fas.harvard.edu>
>         <mailto:horvat at fas.harvard.edu
>         <mailto:horvat at fas.harvard.edu>__>> wrote:
>
>              Taimaz,
>
>              Could you include your SIZE.h file so we can see what the
>         other set
>              parameters are?
>
>              Chris Horvat
>
>
>              On Thu, Jul 26, 2012 at 3:45 PM, taimaz.bahadory
>         <taimaz.bahadory at mun.ca <mailto:taimaz.bahadory at mun.ca>
>         <mailto:taimaz.bahadory at mun.ca
>         <mailto:taimaz.bahadory at mun.ca>__>> wrote:
>
>                  Hi;
>
>                  I have configures a code to run my topography file using
>                  parallel system (mpi). Everything works fine when I set
>         nPy=10
>                  (or anything), but the errors come out when I try to
>         break the
>                  x-direction to different sections, i.e. increasing nPx to
>                  something more than 1.
>                  summary:
>
>                    * nPx = 1; nPy = 10 => works fine
>                    * nPx = 10; nPy = 1 => fails
>
>
>                  Is there any limitation on the x-axis?
>
>                  _________________________________________________
>                  MITgcm-support mailing list
>         MITgcm-support at mitgcm.org <mailto:MITgcm-support at mitgcm.org>
>         <mailto:MITgcm-support at mitgcm.__org
>         <mailto:MITgcm-support at mitgcm.org>>
>
>         http://mitgcm.org/mailman/__listinfo/mitgcm-support
>         <http://mitgcm.org/mailman/listinfo/mitgcm-support>
>
>
>
>              _________________________________________________
>              MITgcm-support mailing list
>         MITgcm-support at mitgcm.org <mailto:MITgcm-support at mitgcm.org>
>         <mailto:MITgcm-support at mitgcm.__org
>         <mailto:MITgcm-support at mitgcm.org>>
>
>         http://mitgcm.org/mailman/__listinfo/mitgcm-support
>         <http://mitgcm.org/mailman/listinfo/mitgcm-support>
>
>
>
>
>         _________________________________________________
>         MITgcm-support mailing list
>         MITgcm-support at mitgcm.org <mailto:MITgcm-support at mitgcm.org>
>         http://mitgcm.org/mailman/__listinfo/mitgcm-support
>         <http://mitgcm.org/mailman/listinfo/mitgcm-support>
>
>
>
>     _________________________________________________
>     MITgcm-support mailing list
>     MITgcm-support at mitgcm.org <mailto:MITgcm-support at mitgcm.org>
>     http://mitgcm.org/mailman/__listinfo/mitgcm-support
>     <http://mitgcm.org/mailman/listinfo/mitgcm-support>
>
>
>
>
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support




More information about the MITgcm-support mailing list