Christoph Hormann <chr### [at] gmxde> wrote:
> This would be just a bounding box test that would certainly speed up the
> collisions a bit but would not change anything about the principal
> problem of mass interactions being O(n^2).

And here's the crux of the problem - I don't know of any way to make the
function simpler or faster, which means I'm just stuck with the speed of it
as it is right now.  I might continue writing my own system, and use sse
for speed, but that would be at the expense of accuracy.

>  You should have a look at
> what Daniel Jungmann wrote recently (a few threads ago) which was mostly
> meant for mass-face-collisions but also works on mass-mass collisions to
> some extent.

I'll take a look at this.

> And the sqrt() has to be calculated for all collisions that actually
> occur since the force depends on the distance.

Of course, and I never meant to imply that you could do without calculating
it for those collisions that do occur.  I was only looking for a way to
cull those which do not.

....Chambers

Post a reply to this message