I’ve been trying to use odejava for a few months now and have become fairly comfortable with it. However it’s out-of-dateness and trouble with exporting the binaries to various operatins systems have been driving me nuts so I decided to look into JOODE. I was wondering if anyone has done in benchmarks or other comparisons to see how well they compare to each other.
odejava-jni (a module in cvs) is much more ‘up to date’ than than the odejava you are probably using. Also there is a branch in odejava, which uses this new version (something between ODE 0.7 and .8 iirc).
In jME Physics 2 you can switch between ODEJava and JOODE implementation (though JOODE version is not the most recent one). The performance is quite different. Large number of objects and especially large number of collisions won’t work. Additionally JOODE still lacks quite some features (numerous joints, cylinders, trimesh-primitive collision etc.). But I did not run benchmarks actually. JOODE is still under heavy development, I think.
Would love to use JOODE as well, but it’s still not in a state I could call complete.
I would say a few of those comments are a bit out of date.
I have simulations running 200+ objects. JOODE also has configurable joints which generalise joints (though constraining their motions with bounds is not implemlemented).
Collision detection is a bit ropey. Agreed. I need some experts on that…
I would say JOODE is still under heavy development, but I think its come to the point where it is becomeing useful, and in some areas better than ODE. I am very responsive to any bugs ,problems, feature requests and general inquireries. I could do with people playing with JOODE to help me get it more user orientated. It will be difficult at first for the first few users but like I said, I am happy to help and we will all benifit in the end.
Tom
I think I might try joode since it should be fairly easy to port my existing code.
A note to on getting the more up-to-date version of odejava from cvs. I tried that, however I wasn’t able to successfully build it since the readme’s kept seeming to contradict each other. If you know of a post or tutorial that does a better job than those in cvs I’d really appreciate it.
They shouldn’t! Post about it on this board, please - it should be fixed…
hmm, if you add another 0 it would get where I’d expect it to 
interesting - with those you can implement universal, hinge 2 with stops, slider with stops etc.? cool
[quote]interesting - with those you can implement universal, hinge 2 with stops, slider with stops etc.? cool
[/quote]
Actually I am not too sure whether hinge 2 will be possible, becuase that has two axels located at different places. But all the others should be possible (once I have got stops working correctly). But yeah, the idea is to generalize the joint system, less code to maintain that way.
[quote]hmm, if you add another 0 it would get where I’d expect it to
[/quote]
Hmmm. Ok. Well I have 200+ bodies simulated and connected in one single island, and using RK4 adaptive step, so the overhead in simualting that is way higher than ODEs fixed Euler steps (but a hell of alot more stable), so you can’t really compare them. I don’t know how many bodies I can Euler step, alot more.
I am gonna try and get stops working this afternoon.
Tom