[ODE] Iterative solution

[Billy Zelsnack]

>>>Quite so, excellent point.  It's just a shame that the island-finding
>>>is what, O(nlogn) itself?  But a cheap one.
>>
>> Should be O(n+e), n==number of bodies, e==number of constraints, no?
Isn't it
>> just a series of simple DFS's of the graph formed by the bodies and
joints?
>I was being charitable and assuming that a magic hash lookup to
>detect cyclic graphs in the island-building from traversing joints
>would occur in log(n) time complexity for each node.

The processor time required to build islands is absolutely insignificant
compared to just about everything else in a physics engine. I would not
worry about it.

-billy