OOP and prime sieves

[m...@iae.nl]

Albert van der Horst commented on my parallel FIB posting
and said that I should work on a program that actually does
something useful:

> Another 5 years have passed so I think the time is right to
> push my proposal for a recursion benchmark, again.
> It contrasts with Fibonacci, that it makes more sense.
>
> \---------------------------------------------
> \ FILE: benchpin.frt
> \ LANGUAGE : ISO Forth
> \ TITLE : The corrected PI(N) recursion benchmark.

Because it is mutually recursive, benchpin is not easy
to adapt to processors where communication between
nodes is vastly slower than computational speed. Therefore
I looked for prime counting programs that might be
easier to parallelize. I found Kim Walisch's primesieve
project on http://primesieve.googlecode.com .

The code is in C++ and uses OpenMP. Just the thing,
OOP makes it real easy to port.

Reading the source has been a revelation for me. The
direction that object-oriented programming for parallel
architectures has taken is really mind-boggling. I will
never be able to duplicate this in Forth, and will have
to settle for a very short, extremely simplified version.

Given the extensive debate on FOOP we see on CLF I'd be
interested in comments on Walisch's primesieve project
from a Forth perspective. Is this how all C++ projects
look nowadays? What exactly does the primesieve gain
from being object oriented? Would FOOP be as good as
Walish's accomplishment?

-marcel