[MITgcm-devel] FW: [MITgcm-support] size.h

[Chris Hill]

Ed,

 When you have a chance could you check whats happening with the mitgcm
online doc search field (the test field under Quick Links)? I just tried
"parallel", "parallelism", "hydrostatic", "mpirun". None of these matched
anything.

Chris

> -----Original Message-----
> From: mitgcm-support-bounces at mitgcm.org 
> [mailto:mitgcm-support-bounces at mitgcm.org] On Behalf Of Ed Hill
> Sent: Wednesday, April 07, 2004 6:11 PM
> To: MITgcm-support
> Subject: Re: [MITgcm-support] size.h
> 
> On Wed, 2004-04-07 at 18:55, Michael Schaferkotter wrote:
> > greetings:
> > i/ve successfully run the plume on the shelf problem and i
> want to modify the example.
> > 
> > looking in SIZE.h i see that:
> > 
> >      &           sNx = 160,
> >      &           sNy =   1,
> >      &           nSx =   2,
> >      &           nSy =   1,
> >      &           nPx =   1,
> >      &           nPy =   1,
> >      &           Nx  = sNx*nSx*nPx,
> >      &           Ny  = sNy*nSy*nPy,
> >      &           Nr  =  60)
> > 
> > the domain is 320x1x60.
> > 
> > why is nSx set to 2, i/m using a single processor (i
> think). there was
> > nothing in the make that i could see indicated multi-processor.
> 
> Hi Michael,
> 
> Almost all of the verification examples currently have multiple tiles 
> so that they will give the same (or almost the
> same) results whether they are run on single or multi-processor 
> systems.  If you look in:
> 
>   http://dev.mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/verification/
> 
> you will also see that most tests come with both "code/SIZE.h" and 
> "code/SIZE.h_mpi" files.  They are used by the verification/testreport 
> script to do daily MPI verification runs:
> 
>   http://mitgcm.org/testing.html
> 
> 
> > also, i couldn/t find the info on the web site about
> parallel setup. i
> > have access to a bunch of machines and would like to try an MPI run.
> 
> 
> I just looked through the docs and you're right--we should add more 
> detail about how to use MPI.  In the mean time, heres the "quick"
> directions:
> 
>   1) Determine the locations of your MPI compiler and/or libraries 
>      and put them into an options file ("optfile").  It will be 
>      easiest to start with one of the examples in:
> 
> http://dev.mitgcm.org/cgi-bin/viewcvs.cgi/MITgcm/tools/build_options/
> 
>      such as "linux_ia32_g77+mpi_cg01" or "linux_ia64_efc+mpi" and 
>      then edit it to suite your machine.  You may need help from 
>      your local systems admin to determine the location of the MPI 
>      libraries or, if necessary, to have them installed.
> 
> 
>   2) Build the code with the genmake2 "-mpi" command such as:
> 
>      ../../../tools/genmake2 -mods=../code -mpi -of=YOUR_OPTFILE
>      make depend
>      make
> 
> 
>   3) Run the code with the MPI "loader" or "run" program provided
>      with your particular implementation of MPI.  Typical MPI 
>      packages such as mpich will use something like:
> 
>      mpirun -np 4 -machinefile mf ./mitgcmuv
> 
> 
> And thats it.  If you have more specific problems please send them to 
> this list and we'll try to help.
> 
> Ed
> 
> ps - Attached are the commands used recently to install MPICH
>   and build and run MITgcm on a fairly standard (if there is
>   such a thing!) Linux cluster.
> 
> 
> --
> Edward H. Hill III, PhD
> office:  MIT Dept. of EAPS;  Room 54-1424;  77 Massachusetts Ave.
>             Cambridge, MA 02139-4307
> email:   eh3 at mit.edu,  ed at eh3.com
> URL:     http://web.mit.edu/eh3/
> phone:   617-253-0098
> fax:     617-253-4464
> 
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: mpi_example.txt
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20040407/0f0b6395/attachment.txt>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20040407/0f0b6395/attachment.sig>
-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: ATT00203.txt
URL: <http://mitgcm.org/pipermail/mitgcm-devel/attachments/20040407/0f0b6395/attachment-0001.txt>