Re: Need some light on using Ada or not

[jonathan <johnscpg@googlemail.com>]

On Feb 20, 9:15 pm, Pascal Obry <pas...@obry.net> wrote:
> Jonathan,
>
> > My preferred plan B would link to MPI, which distributes mpi-tasks
> > over the available cores.  I don't know if that will work either!
>
> That won't be better. MPI is for non shared memory used in clusters.
> Tasking on a single machine is the best option AFAIK.
>
> --
>
> --|------------------------------------------------------
> --| Pascal Obry                           Team-Ada Member
> --| 45, rue Gabriel Peri - 78114 Magny Les Hameaux FRANCE
> --|------------------------------------------------------
> --|    http://www.obry.net -  http://v2p.fr.eu.org
> --| "The best way to travel is by means of imagination"
> --|
> --| gpg --keyserver keys.gnupg.net --recv-key F949BD3B

Hi Pascal,

MPI is more versatile than you think.  When I run 8 MPI
tasks on a single machine with 8 cores, then it works beautifully.
When I run 48 MPI tasks on six 8-core machines, then it works fine,
but
you notice the degraded bandwidth from the network. All of this is
transparent to user .. all he sees is the 8 cores or the 48 cores. The
basic MPI communications model is a remote rendezvous, which I use
extensively. I have nothing but praise for it. Asynchronous
communication
is available but has never been much use in my applications.

But actually, I'ld be surprised if MPI worked well in the present
application (benchmarking "new" and unchecked_deallocation). I am
*still* not planning to do any work on this benchmark.

J.