[rescue] Small servers (was Re: WTT: 1.5G of PC2700 for 1G of PC100)

[Joshua Boyd]

On Mon, May 05, 2008 at 02:00:03PM -0600, Robert Darlington wrote:
> And absolutely none of that applies to high performance computing.
> Try running object oriented code where time counts!  I guess you could
> always throw more computing power at a problem, but the problems I've
> worked on take months to solve with computers that cost half a billion
> dollars.  Going OO would increase the time to solution many times
> over.  Fortran 77 for life, pointers be damned.

Depending on the type of performance needed and the sort of problem you
are working on, OO code may be a very good idea.

I say that it is best to start with the most programmer efficient method
that could possibly work, then optimize from there.  If the fastest way
to get up and running and to prove your idea is to use a HLL running
across a cluster, then do so first, and optimize from there.  And
depending on the HLL you choose, you may not have much more work to do.

I don't know how well it normally works, nor do I know how to use it at
all, but I've seen some very impressive parallel performance from Jocaml
in demonstrations.  And for other situations there is Erlang.  And even
Python has potential to be an acceptable starting point.