Date   : Tue, 20 Apr 2010 21:57:03 -0300
From   : mlist@... (Steven Flintham)
Subject: BBC mandelbrot generator (preview)

On 18 April 2010 14:38, Tom Seddon <tom@...> wrote:

> Thanks for the link. I'd never seen that one before. It's... much faster
> than I was expecting :)
>
> (Even after running it in its most straightforward calculation mode, it's
> about twice as fast as mine! I'm going to have to work out what it's doing
> now!)
>

I don't know much about second processors, but I wonder if it would be
feasible to write a version which actually splits the Mandelbrot
calculations between the two CPUs? My gut feeling is that just handling the
plotting can't really be taking up that much of the host processor's time.

Jumping ahead on the assumption it is possible :-), my inclination would be
to divide the rows up between the processors in the same ratio as the clock
speeds, but "alternating" to try to avoid one CPU getting all the rows which
contain lots of black pixels and hence finishing much later than the other.
(I guess "adjacent" rows will tend to be relatively similar in terms of
computational workload.) Of course, the nicest way to do it would just be to
have a pool of not-yet-done lines which both processors pick from, but I
wonder if that would be awkward to coordinate.

Just curious to know if this is a totally impractical idea. I hope I will
learn something interesting about second processors if anyone bothers to
point out the problems...

Steve