[ODE] New Physics SDK released

[Pierre Terdiman]

> the solver seems to be taken almost verbatim from Pierre Terdiman's
Flexine.

No, we rewrote it.

My previous solver (http://www.codercorner.com/Rigid.htm) was O(n) but
needed very small timesteps to be robust (the default mode was 250 Hz). This
one is O(n) but works with bigger timesteps (default is 60 Hz), so it's
actually a lot faster.

I included one of my old test scenes on purpose, so that people can spot the
obvious speed difference. In particular, this scene from my old Flexine
physics tests:

    "Physics_StressTest06d.zcb"

has been included in the new Novodex SDK:

    "Stress test 2"

On my dev PC, my Flexine scene runs at about 20 fps before going to sleep,
while the NX scene runs at more than 100 fps all the time. This is coherent
with the theory, since I run the simulation 250/60 = a bit more than 4 times
more often in Flexine. If you change the timestep value in Flexine using the
appropriate dialog (menu Scene->Rigid body simulation), and set it to 50 Hz,
the stacks just collapse...

Short story : no, it's not the same solver.


> it's a penalty-based method with a very small time step (seems to be
around 2ms
> in the demos).

The expected timestep in the new solver is 60 Hz, i.e. console people are
supposed to run one physics step / render frame. Simple.


> but it has some inherent limitations.

It probably has the same limitations as ODE's "fast step". I guess. I don't
know, I've never used ODE.


Pierre Terdiman

- Novodex AG (www.novodex.com)
- Personal : www.codercorner.com