Beyond the numerous MPI tutorials, I am aware of several very different
examples of MPI codes.
1) the NAS parallel benchmarks, which incidentally provide a comparison
between MPI and Open-MP implementations of the same algorithms
NAS Parallel Benchmarks Team
NASA Ames Research Center
Mail Stop: T27A-1
Moffett Field, CA 94035-1000
E-mail: npb_at_[hidden]
Fax: (650) 604-3957
http://www.nas.nasa.gov/Software/NPB/
2) the scalapack project
http://www.netlib.org/scalapack/
which points to *lots* of open source MPI software
3) on the national french IDRIS supercomputer, *lots* of examples
supporting the tutorials, e.g.
http://www.idris.fr/docs/docu/support_cours/parallele.html
(arghhh in french, sorry !).
Hoping this helps.
Best.
Pierre.
PS. In my own MPI code (e.g.
http://www-laog.obs.ujf-grenoble.fr/~valiron/ccr12/index.html) I have
developed several simple fortran "auto-adaptative" MPI routines that may
be of general interest. If this rouses some interest, I'll publish them.
- discover which tasks share the same filesystem and which don't, and
create a communicator to duplicate data files from the master task. this
routine is especially useful on a cluster if you want to use the local
scratch filesystem on each node.
- automatic loop distribution and load balancing with minimal
communication overhead (assuming the loop presents no dependances). This
is especially useful for irregular loops and/or inhomogeneous MPI
environments.
- and also numerous small utility routines, for instance routines to
automate the creation of datatypes for MPI_Bcast
Hugh Merz wrote:
>>> Well, a google search for: mpi code examples produces very good
>>> results...
>>
>> Yes, but they're mostly tutorials: you can't just download a CFD code
>> that will run, as you could for say (ug) windows.
>
>
> In that case you would have to search google for "mpi cfd" and the
> third hit would give you:
> http://www-berkeley.ansys.com/cfd/CFD_codes_p.html
>
> "CFD codes list - free software"
>
> The reason why you only find tutorials is that most people who write
> mpi codes are writing them for private research, and these are
> organized on the www by the problem they solve, not by the fact that
> they use mpi.
>
> If you want to find robust mpi applications you have to narrow the
> search.
>
>>> If somebody wants me to set up a wiki for mpi I would be happy to do
>>> it.
>>> Since I don't expect the bandwidth for such a thing to be huge I could
>>> even set it up on my server.
>>>
>>> Of course, it would be up to the community to put stuff there to
>>> make it
>>> useful.
>>
>> I'd be happy to help, but as you asked, is the community there?
>> Anyone else like the idea?
>
>
> I think providing a centralized mpi code repository would be a good
> thing. The MPI applications indicated on the LAM site are good places
> to start for finding common mathematical solvers, but there is no "one
> place to go" if you are looking for more complex programs (as you have
> indicated above).
>
> As well as this many research codes are still closed, in that they are
> owned by commercial entities, government agencies and academics who are
> unwilling to share source due to the amount of time (money) invested
> in them. Having an open source repository would help to reduce
> overlap in development (I bet many graduate students have effectively
> written the same program!) and provide incentive for groups who are
> fearful of losing their edge to take the lead and put their code out
> there. Then again maybe I'm living in a naive fantasy world.
>
> Are there any plans to have something of this nature published on the
> open-mpi.org site?
>
> Just my 2 cents.
>
> Hugh
>
>
>> Thanks,
>>
>> -Jack C
>> http://www.crepinc.com/
>>
>> _______________________________________________
>> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
--
Soutenez le mouvement SAUVONS LA RECHERCHE :
http://recherche-en-danger.apinc.org/
_/_/_/_/ _/ _/ Dr. Pierre VALIRON
_/ _/ _/ _/ Laboratoire d'Astrophysique
_/ _/ _/ _/ Observatoire de Grenoble / UJF
_/_/_/_/ _/ _/ BP 53 F-38041 Grenoble Cedex 9 (France)
_/ _/ _/ http://www-laog.obs.ujf-grenoble.fr
_/ _/ _/ mail: Pierre.Valiron_at_[hidden]
_/ _/ _/ Phone: +33 4 7651 4787 Fax: +33 4 7644 8821
_/ _/_/
|
