Re: tlug: Re: multi-processor linux configuration ?

["Manuel M. T. Chakravarty" <chak@example.com>]

Chris Sekiya <chris@example.com> wrote,

> On Wed, 14 Oct 1998, Darren Cook wrote:
> 
> > I'd have thought the problem was with applications that parallelize well
> > and those that don't, and the language is of secondary importance.
> 
> It helps when the language has extensions for parallel execution.  Various
> flavors of fortran have parallel provisions (PARDO, for instance), but I'm
> not aware of any for C[1].

These days, parallel applications are actually often written
in C.  One reason is that one of the most popular (and most
tedious) ways of writing a parallel program is the use of
MPI (a message passing library).  It interfaces without any
problems to C.  Parallelizing Fortran is more important for
vector computers and maybe also for virtual-shared memory
machines.  For relatively high-latenzy distributed-memory
machines (like clusters of Linux machines) I don't think
there is any chance with a parallelizing Fortran compiler.
maybe HFP (High Performance Fortran)[1] is an option -- it
requires a many annotations by the programmer that guide
data distribution etc.

Re parallel versions of C: There are plenty:

  CC++ : http://globus.isi.edu/ccpp/

  Cid  : http://www.crl.research.digital.com/projects/scalable/default.html

  Cilk : http://theory.lcs.mit.edu/~cilk/

  MPC++: http://www.rwcp.or.jp/people/mpslab/mpc++/mpc++.html
  
  mpC  : http://www.ispras.ru/~mpc/index.html

  Split-C: http://www.cs.berkeley.edu/projects/parallel/castle/split-c/

To name a few...

Cheers,

Manuel

[1] http://www.crpc.rice.edu/HPFF/home.html
---------------------------------------------------------------
Next Nomikai: 20 November, 19:30 Tengu TokyoEkiMae 03-3275-3691
Next Meeting: 12 December, 12:30 Tokyo Station Yaesu central gate
---------------------------------------------------------------
Sponsor: PHT, makers of TurboLinux http://www.pht.co.jp