[Beowulf] cluster softwares supporting parallel CFD computing

[Jim Lux]

At 09:12 AM 9/17/2006, Geoff Jacobs wrote:
>Jim Lux wrote:
><snip>
> >
> > Yes, the software structure was badly designed for the interconnect.
> > HOWEVER, the whole point of computing resources is that I really
> > shouldn't have to design the whole software system around the
> > peculiarities of the hardware platform.
>
>It sure would be nice if software could be partitioned in an efficient
>manner, including considerations of interconnect topology, with a
>compiler flag. Alas, this we must still do ourselves.


I suspect that the number of people writing software for which there 
IS a big dependence on interconnect properties AND for which the 
compiler could actually do something useful, is fairly small 
(probably more than I can count on my fingers and toes, but I doubt 
it runs into the thousands).

This sort of falls in the compiler/optimizer category, and it's a 
pretty non-trivial problem.  Those few folks who actually write 
compilers  (as opposed to loading yet another syntax description into 
yacc/lex) have got their hands full dealing with the easier 
optimizations that are used by more people.


>This is a strange sentiment to be displaying on this list, if I may say
>so. Beowulfers sit further out on the frontier so they can derive more
>performance from their hardware budgets. The consequence of this is
>additional development time, application complexity, and sometimes
>hardware specificity.

Well, yes... but one can dream, can't one?  It's that "DWIM" (do what 
i mean) compiler.



>--
>Geoffrey D. Jacobs
>
>Go to the Chinese Restaurant,
>Order the Special

James Lux, P.E.
Spacecraft Radio Frequency Subsystems Group
Flight Communications Systems Section
Jet Propulsion Laboratory, Mail Stop 161-213
4800 Oak Grove Drive
Pasadena CA 91109
tel: (818)354-2075
fax: (818)393-6875