Hello,
Not speaking for Van (who can perfectly well
speak for himself :-), when I teach parallelism
to Fortran programmers, the following reactions
usually result:
OpenMP- OK for SMP, but what if 100's of processors are needed?
HPF- Great for some problems, OK for some, but it's impossible
to manage the communications overhead
MPI- "My subroutines are now four times as big,
and I haven't added any physics!", which I call
"message passing shock"
Co-array Fortran does well on SMP, so it covers OpenMP;
it does well on DMP, so it covers MPI; it's understandable;
and it's portable between SMP and DMP, and therefore
to mixtures of the two (as in when your Beowulf has dual
processor nodes).
Peter Shenkin wrote:
>
> Hi, Van,
>
> On Thu, 20 Jul 2000, Van Snyder wrote:
>
> > We're interested to use Fortran 95 with a Beowulf cluster. We're not
> > excited about OpenMP or HPF or explicit MPI.
>
> Just curious: can you say why you're not into each of the above?
>
> -P.
>
> --
> ** Whether the playing field is level depends on the coordinate system. ***
> ********* Peter S. Shenkin; Schrodinger, Inc.; (201)433-2014 x111 *********
> *********** [log in to unmask]; http://www.schrodinger.com ***********
--
Cheers!
Dan Nagle
Purple Sage Computing Solutions, Inc.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|