[MITgcm-support] Running a test in parallel

Jody Klymak jklymak at uvic.ca
Fri Feb 3 17:21:51 EST 2017


Hi Luis,

Nx  = sNx*nSx*nPx = 10*9*2 = 180.

If you want two processors in x, set sNx=45,nSx=1, and nPx=2

(ahem, though be aware that some compilers w/ optimizations don’t like odd sNx)

Cheers,   Jody



> On 3 Feb 2017, at  14:16 PM, Luis Cebamanos <l.cebamanos at epcc.ed.ac.uk> wrote:
> 
> Hi Jody,
> 
> This sounds sensible, where can I find nx times ny value? My SIZE.h
> under the global_ocean.90x40x15 directory looks like this:
> 
>      PARAMETER (
>     &           sNx =  10,
>     &           sNy =  10,
>     &           OLx =   3,
>     &           OLy =   3,
>     &           nSx =   9,
>     &           nSy =   4,
>     &           nPx =   2,
>     &           nPy =   2,
>     &           Nx  = sNx*nSx*nPx,
>     &           Ny  = sNy*nSy*nPy,
>     &           Nr  =  15)
> 
> 
> Cheers,
> Luis
> 
> On 03/02/2017 20:40, Klymak Jody wrote:
>> Sorry my fault.  If you can't read a record on a file it usually means that it is the wrong size.  Does nx times ny in the file equal what's in Size.h?   
>> 
>> Cheers.   Jody
>> 
>> Sent from my iPhone
>> 
>>> On Feb 3, 2017, at 11:09, Luis Cebamanos <l.cebamanos at epcc.ed.ac.uk> wrote:
>>> 
>>> Hi Jody,
>>> 
>>> I am sorry for the confusion. It is still on the same system, just
>>> different partitions. The computing nodes can only see some partitions
>>> of the system.
>>> I guess my question is, why is MITgcm failing to read  bathymetry.bin
>>> although the file is created?
>>> 
>>> Thanks,
>>> Luis
>>>> On 03/02/2017 18:58, Jody Klymak wrote:
>>>> Hi Luis,
>>>> 
>>>> You must recompile `mitgcmuv` on the new computer.  Executable programs aren’t usually portable between architectures.  Hopefully there is a helpful file in `tools/build_options` to help w/ your genmake2 step.
>>>> 
>>>> Cheers,   Jody
>>>> 
>>>> 
>>>>> On 3 Feb 2017, at  10:53 AM, Luis Cebamanos <l.cebamanos at epcc.ed.ac.uk> wrote:
>>>>> 
>>>>> Hi all,
>>>>> 
>>>>> I am totally new to MITgcm, so apologies if I say any  nonsense. I m
>>>>> trying to run a test in parallel,so I started with
>>>>> global_ocean.90x40x15. I changed the SIZE.h to run on 4 mpi processes
>>>>> and built it successfully. Now, I am working on a Cray system, meaning
>>>>> that it has to be run on a different location. So I copied over to the
>>>>> right place /work/MITgcm/ the following directories from
>>>>> global_ocean.90x40x15:
>>>>> 
>>>>> global_ocean.90x40x15/input
>>>>> global_ocean.90x40x15/build/mitgcmuv
>>>>> tutorial_global_oce_latlon/input
>>>>> 
>>>>> I then created the directory /work/MITgcm/run and run the following:
>>>>> ln -s ../input/* .
>>>>> ../input/prepare_run
>>>>> ln -s ../mitgcmuv .
>>>>> 
>>>>> My script calls MITgcm:
>>>>> 
>>>>> aprun -n 4 -N 4 -d 1 ./mitgcmuv
>>>>> 
>>>>> It appears to start running properly but soon fails with the following
>>>>> errors:
>>>>> 
>>>>> lib-4016 : UNRECOVERABLE library error
>>>>> A READ operation tried to read a nonexistent record (721).
>>>>> 
>>>>> Encountered during a direct access unformatted READ from unit 9
>>>>> Fortran unit 9 is connected to a direct unformatted unblocked file:
>>>>> "bathymetry.bin"
>>>>> 
>>>>> lib-4016 : UNRECOVERABLE library error
>>>>> A READ operation tried to read a nonexistent record (730).
>>>>> 
>>>>> lib-4016 : UNRECOVERABLE library error
>>>>> A READ operation tried to read a nonexistent record (370).
>>>>> 
>>>>> 
>>>>> Could someone please help me to run a simple test case ?
>>>>> 
>>>>> Regards,
>>>>> Luis
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> 
>>>>> The University of Edinburgh is a charitable body, registered in
>>>>> Scotland, with registration number SC005336.
>>>>> _______________________________________________
>>>>> MITgcm-support mailing list
>>>>> MITgcm-support at mitgcm.org
>>>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>>> --
>>>> Jody Klymak    
>>>> http://web.uvic.ca/~jklymak/
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> 
>>>> _______________________________________________
>>>> MITgcm-support mailing list
>>>> MITgcm-support at mitgcm.org
>>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>>> 
>>> The University of Edinburgh is a charitable body, registered in
>>> Scotland, with registration number SC005336.
>>> _______________________________________________
>>> MITgcm-support mailing list
>>> MITgcm-support at mitgcm.org
>>> http://mitgcm.org/mailman/listinfo/mitgcm-support
>> 
>> _______________________________________________
>> MITgcm-support mailing list
>> MITgcm-support at mitgcm.org
>> http://mitgcm.org/mailman/listinfo/mitgcm-support
> 
> 
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
> _______________________________________________
> MITgcm-support mailing list
> MITgcm-support at mitgcm.org
> http://mitgcm.org/mailman/listinfo/mitgcm-support

--
Jody Klymak    
http://web.uvic.ca/~jklymak/








More information about the MITgcm-support mailing list