[JOODE] Contribution

[cylab]

cylab: why vpp? you could use APT which comes with javac and eclipse these days anyway. LWJGL uses it for generation of their GL classes and extensions. Saves introducing another dependancy, but thats up to you guys.

Vpp was just an example and I wasn't aware, that apt could be used under this circumstances. I am still not sure, how to implement something like

1   2   3   4   5

#ifdef USE_FAST_MATH   float result= FastMath.invSqrt(value); #else   float result= 1/Math.sqrt(value); #endif

with apt. How would you do this?

[darkprophet]

you would use annotations, so you have a template and from the template, you create your sources...

so for example:

1   2

@FastInvSqrt public float invSqrt(float v);

and its up to the processor that you create to take what FastInvSqrt means and convert that into legible code.

[t_larkworthy]

Also, the LCP iterator is pretty much usless for simulations anyway, so im not sure what applications JOODE or even ODE has outside of gaming...

ODE is very popular for robotics researches wanting a cheap simulation environment. At the end of the day a high fidelity system is not going to come close to capuring a real world environment. There are too many parameters for anyone to configure (I mostly talking about surface parameters). The major opinion is that a robot shown to work in simulation won't show that it works in real life, no matter how accurate the simulation environment is. So why bother trying. ODE and JOODE basically captures the really important stuff, joints swing about like they should do, masses have momentum etc.

Nick Jakobi suggests that to increase the chances a controllor for a robot developed in smimulation will transfer to real life involes injecting huge amounts of noise into the simulation, thus kind of negating much of the gains in a high fidelity system, see:
"Evolutionary Robotics and the Radical Envelope-of-Noise Hypothesis" Nick Jakobi