(j3.2006) Liaison to IFIP WG 2.5

[Aleksandar Donev]

On Wednesday 22 August 2007 11:54, Lawrie Schonfelder wrote:

> You appear to be angry about parameterised data types
I am not angry. They are useful and provide consistency. But they are not 
nearly as useful to *real* scientific programmers as is parallelism. 
Pragmatism is more important than purism in cases like this...

> I would claim that there is no, absolutely null, nil,
> applications that can be programmed with co-arrays that cannot be
> programmed more succinctly without them. 
We should distinguish here between applications and algorithms. An application 
is a problem to be solved by a computer. Sure, any problem that can be solved 
in parallel can be solved in serial as well. But parallel *algorithms* cannot 
be programmed with existing Fortran. And people design and use them. So 
explicit parallelism is necessary.

> That makes co-arrays very 
> different conceptually
Sure, we can argue philosophy all day.

> and practically. 
Yes, explicit parallelism is much more important practically than any 
type-system feature added since derived types in F90.

Best,
Aleks

-- 
Aleksandar Donev, Ph.D.
Lawrence Postdoctoral Fellow @ LLNL
High Performance Computational Materials Science and Chemistry
E-mail: donev1 at llnl.gov
Phone: (925) 424-6816  Fax: (925) 423-0785
Address: P.O.Box 808, L-367, Livermore, CA 94551-9900
Web: http://cherrypit.princeton.edu/donev