Jeffrey B. Layton wrote:
I know this will set off a language war, but I've found that anything
technical or scientific I need to code is always easier in Fortran than
anything else (well except Matlab, but I usually prototype in Matlab
and move to Fortran for extra speed). For many other tasks, I use
Python or bash or perl or whatever I have handy.
So my general opinion about coding for clusters is to use whatever you
want. You can usually find thread libraries or MPI libraries for that
language. Of course, it's always best if you code in Fortran :)
Jeff
_______________________________________________
Beowulf mailing list, [email protected]
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
I say this a different way: the fastest programs are FORTRAN programs,
no matter what language they are written in.
I write my fortran programs in C, but
I've also written fortran programs in Common LISP, using the
amazingly capable LOOP macro, and they run just fine as well.
(And finally, my favorite story. I taught a programming languages class
at Stanford once, which covered Pascal, LISP, APL, and SNOBOL, with
a programming assignment in each. The SNOBOL assignment was
to solve a crossword puzzle. I was looking for correctness rather
than speed, but one fellow turned in an amazingly fast SNOBOL
solution. When I held up the listing and looked
at it from arms-length, it even looked like FORTRAN.)
I am not sure what is going on here, but I suspect there is something
about our machine designs that makes them do a really good job
on load/store/arithmetic/if/loop "fortran" programs.
-L
_______________________________________________
Beowulf mailing list, [email protected]
To change your subscription (digest mode or unsubscribe) visit
http://www.beowulf.org/mailman/listinfo/beowulf
