Newton or ODE

[technomage]

Hi everyone

I don't want to start a long thread rhere, but I'm just about to implement the physics in a little simulation I'm doing. I have been using ODE, but now that Newton targets Mac OS and Linux I was thinking of swaping.

My main question is which system will handle the most number of objects in an accurate manner, I'm talking 1000's of objects with simple sphere collision detaction.

I will be doing sometests in the near future myself, but I'd like ot get some input based on the experiance people have.



Dean

[WILL]

An interesting question Dean. I wonder what other Pascal physics engine options there are aswell on top of this.

[size=9px]Mind you I fail to see the tie-in to JEDI-SDL here though.[/size]

[{MSX}]

i never programmed with Newton, but looking at the demos i could say that they have roughly the same performances.

Remember anyway that ODE is open source while Newton is proprietary.

[savage]

Btw, ODE also works on MacOS X. Instructions can be found @ http://ode.org/ode-latest-userguide.html

Also note that Sascha or I have not had the chance to add the cross-platform stuff to the Newton headers yet.

HTH

[Sascha Willems]

Newton or ODE?
Newton won't be able to handle too many objects. This is because Newton is the only physics-engine around that has a correct solver (even commercial engines like Havok only use some cheating solvers). So you can't stack dozens of boxes on top of each-other when using Newton, but if you compare physics behaviour, you'll notice that Newton behaves much more realistic. Although with newer versions you can set the models for the solver and friction to speed simulation up, which in the end allows for more objects but also means more instability. But I see that you said "spheres", so if you only need sphere you could go with Newton, even for a huge amoung of spheres since sphere-collisions are cheap in term of contact-numbers.

Newton cross-plattform?
As savage already said, the current header hasn't been tested under other OS'es than windows. Macs are very rare here around in Europe (I'm repairing and setting-up computers for over ten years now and never had to deal with a Mac) so I won't have the possibility to get the headers working for Mac. As for Linux : The headers support Free-Pascal, so it shouldn't be too hard to get them running under Linux (maybe someone is even using them with Linux, dunno). I have Linux on my system installed, but don't have the time to try out the headers there (maybe I'll get to it at some time). And as for the "Mac-situation" : If any of you Mac-users would make the header compatible for Macs that would be great (plz. contact me then, via PM or E-Mail).

[technomage]

Thanks for the input, I think this just needs testing and find out which is better with sphere collision detection. I'll try and put a demo together which does the same stuff in ODE and netwon and see what we get.

As for Mac OS I'm in the process of setting up Pear PC (http://pearpc.sourceforge.net/) on my machine, I'm hoping that as soon as I have the network working I'll install Free Pascal and start testing some of this stuff under Mac OS. I also have a friend who has agreed to confirm if the app compiled under Pear PC works on a Real Mac, I'll let you all know how this goes, it might be a cheap way to test for Mac OS compatability.

WILL - I put this in the JEDI area because I couldn't seem to find a good 'home' for the question...

Dean

[savage]

Btw, Just tried the latest headers on Linux. I ported Sascha's Vehicle demo to JEDI-SDL and when Kylix tries to link to the libNewton.a file I get a "bad object file format" error.

Probably something silly I am doing. I should have tried it with FPC, but ran out of time. Will try that tomorrow.

[savage]

Oh and one more thing, Apple annouced today that future Macs will be powered by Intel processors.

Dean, please let the JEDI-SDL community know how your go with getting PearPC working.