[ODE] Propagating auto-enable

[Geoff Carlton]

I've got a simulation that consists of a large number of boxes, and I'm 
looking to optimise it.

The domain is very much like a cookie jar full of cubes, where a new 
cube is dropped onto the top of the pile.  The new cube ends up sitting 
on top, and a few others are jiggled around, but the lower level ones 
just really sit there.  In ODE, it works but the period of "jiggling" 
causes a slowdown because the objects are auto-enabled for a brief time.

One idea would be to propagate the "enabled-ness" outward from the 
initial body, only proceeding if the action has caused enough 
movement/force in the intermediate bodies.  In a situation like a 
lopsided pyramid, the top bodies wouldn't be stable, and thus the 
movement would be propagated as far as necessary.  In a situation like 
my cookie-jar, the packed in bodies wouldn't move so the propagation 
would halt.  The propagation per iteration would probably need to depend 
on the transmitted forces and their direction, so a lightly touching 
cube at the foot of a pyramid does little, but one speeding like a 
bullet immediately turns on the lot.

Anybody got any ideas about whether this would work?

Geoff