There were many reasons to introduce dynamic processes in MPI-2
(although it was a contentious issue). Migrating from PVM was one.
Other users genuinely wanted real "accept" and "connect" style dynamic
communication.
There really is no "correct" way to use it -- I've seen a lot of apps
use the PVM-style stuff (e.g., run ./a.out and have it launch the rest
of the workers), but that stuff seems kinda useless (IHMO -- mpirun /
MPI_INIT can do all that with the user writing any additional code to
support it). A few recent codes have started to do more interesting
things, like doing distributed computing, seek out other MPI processes
to join the computation, etc. Having a real multi-threaded-capable MPI
really enables the power of this kind of paradigm (again, IMHO). LAM
can do all the dynamic MPI-2 stuff, but isn't multi-threaded capable.
Open MPI will be; it will probably be interesting to explore such
paradigms.
On Apr 11, 2005, at 12:42 PM, Heiko Bauke wrote:
> Hi,
>
> I apologise for my off-topic question. But I'm sure to find people here
> who can answer my question.
>
> So far I have used in my programms MPI-1.1 features only. Now I would
> like to explore MPI-2.0. At the moment I am wondering, what was the
> reason to introduce dynamic process creation and what is the right way
> to use this feature. Is it just to mimic a PVM programming style?
>
>
> with regards
>
> Heiko
>
> --
> -- Es gehört viel Erfahrung dazu, wie eine Anfängerin zu küssen.
> -- (Zsa Zsa Gabor)
> -- Supercomputing in Magdeburg @ http://tina.nat.uni-magdeburg.de
> -- Heiko Bauke @ http://www.uni-magdeburg.de/bauke
>
> _______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
>
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
