[ODE] center of mass

[Jon Watte (ODE)]

> I noticed that geom "pos" and "R" pointers always point to elements of 
> an implicit posr structure.  I combined the pointers into one, making a 
> saving of one pointers worth of space.  The geom now has two posr 

Ah, cool! So the only real cost is the "if" when the position is 
updated, to test whether to use an offset or not. That's at most the 
cost of a mis-predicted branch (i e, pipeline flush), which can be nasty 
in an inner loop, but probably not bad if it's per-geom per-frame and no 
more.

Cheers,

			/ h+