[ODE] Fwd: Re: question about game scenario using ODE

[Megan Fox]

It isn't terribly possible to answer that question.  Depending on the
rendering of those bricks, you would have significant lag on just the
graphical side of things, never mind that the performence would depend
largely on how you configured your environment (and how likely
whatever final arrangement of bricks is to settle and auto-disable).

I can, however, say that ODE is not terribly well-optimized for
stacking behavior.  For that, you need what's called island disabling,
which has been added to ODE by commercial entities in the past but has
never found its way into public CVS.  When I have time, I intend to
use the method posted by one of the commericla devs to give ODE that
functionality.

Basically - objects in a stack will have a great deal of trouble
coming to rest, due to the nature of the impulse-based corrective
forces.  As it can't come to rest, it'll tend to jiggle itself down
over time, and just be a generally slow sim example (since fast sims
require only small amounts of the world to be active at a time).


You'll probably find that collision is ALWAYS going to take a great
deal more of the time than the solver.  I've never had a scenario
where I found the solver was lagging, it's always been that the
collision has been lagging.

> What kind of destructive game scenario would "stress" out both collision
> detection and iterative LCP solver to a comparable degree, so that ODE
> spends similar amount of time in each.
>
> For example, we designed an example a  single castle (made up of bricks
> resting on top of each other) destroy ted by the cannon ball. In this
> example, most of the time (close to 90%) is spent LCP solver. So I would
> like to design a game scenario which will strike a more even balance between
> collision detection and the solver.
>
> Just to give a right background, my focus is on performance analysis and
> optimization of game physics, rather than game design. So I would like to be
> able to quickly build the game scenario and focus the rest of the time on
> the above two issues.