Which Physics library to use???

SwitchCase · Post by **SwitchCase** » Tue Nov 06, 2007 2:52 pm

Newton? Bullet? ODE?

I usually use Newton and is the only physics library I have used. Starting a new project is a good time to learn how to use a new library. What are the pros and cons of each library? I have found it difficult to find unbiased comparisons of these libraries. My main reason for wanting to change to an alternative library is that Newton is not open source.

Thanks for any replies.

rogerborg · Post by **rogerborg** » Tue Nov 06, 2007 3:04 pm

If you had to choose today, then Newton would be out, given that the download links and forum are both 404 (as is IrrNewt). Also, Newton and IrrNewt are both run by people that I consider to be borderline sociopaths. It takes one to know one.

The Magic 8 Ball says that all the signs point to Bullet over ODE, but it's been known to be wrong.

SwitchCase · Post by **SwitchCase** » Tue Nov 06, 2007 3:15 pm

Does the Magic 8 Ball say why all the signs point to Bullet rather than ODE?

sRc · Post by **sRc** » Tue Nov 06, 2007 8:02 pm

ODE's in the process of integrating Bullet into it, by the way

Halifax · Post by **Halifax** » Tue Nov 06, 2007 8:11 pm

Bullet receives direct updates, and optimizations via SPU programmers from Sony for the PS3, which also in turn improves the general readiness of bullet for multi-core processing.

Bullet is updated more frequently compared to ODE.

Bullet is made by an ex-Havok programmer, and if you don't know what Havok is then I think you are pretty sheltered.

Overall if you are looking for a professional, and multi-core supported, PS3 supported and optimized physics system, then all signs point to Bullet.

rogerborg · Post by **rogerborg** » Tue Nov 06, 2007 8:16 pm

IMO:

ODE is an open source project that happens to be a physics engine.

Bullet is a physics engine that happens to be open source.

SwitchCase · Post by **SwitchCase** » Wed Nov 07, 2007 2:58 pm

OK so Bullet kicks the crap out of ODE.

But... at the moment I find Newton very good, is bullet as good? or better? obviously Bullet's already ahead on points being open source and all.

Halifax · Post by **Halifax** » Wed Nov 07, 2007 8:25 pm

Bullet is way better. Everything done in Newton is done faster in Bullet, trust me. Just look at the reasons that I already provided you. Plus Newton has become/on the verge of becoming abandon-ware.

klikmaster · Post by **klikmaster** » Wed Nov 07, 2007 10:44 pm

Is Bullet easy to integrate with Irrlicht? Are there any nice tutorials for getting it to work nicely with Irrlicht?

SwitchCase · Post by **SwitchCase** » Wed Nov 07, 2007 11:21 pm

I don't doubt you halifax. lol. I guess i'm just trying to find a reason not to have to learn to use a new library. Well, actualy it'll be good fun playing around with it, guess I'll get the hang of it pretty easy if it works anything like Newton does (or did

).

Thanks for the info.

agi_shi · Post by **agi_shi** » Fri Nov 23, 2007 8:05 pm

NewtonGD is not abandon ware, it is not abandoned. Julio had to switch hosting providers, everything is back up. He's also working on version 1.6 and version 2.0 - the community is perfectly active. Nothing done in Newton is easier or faster done in another library - give me a specific example. I'd love to see my portals done in Newton done faster and easier in Bullet.

Carnes · Post by **Carnes** » Sat Nov 24, 2007 9:58 am

Newton seems like a good option, but like Rogerborg pointed out, IrrNewt is no longer updated. I have recently started working with it, very easy to use and it auto calculates a lot for you.

Recompiling the IrrNewt source for Irrlicht 1.3.1 is easy, just one edit required. I've only had one issue with it so far. The createCar function had the tires out of order? It seemed like the author made all his models on a 90 deg rotated X axis and corrected it in code. You can see some of that in his examples. But his code is very readable and it's easy to change/recompile the library&dll.

Tumle and IPhysics are some other Wrappers for Newton but i haven't messed with them yet.

sio2 · Post by **sio2** » Sat Nov 24, 2007 10:56 am

Carnes wrote:Recompiling the IrrNewt source for Irrlicht 1.3.1 is easy, just one edit required. I've only had one issue with it so far. The createCar function had the tires out of order? It seemed like the author made all his models on a 90 deg rotated X axis and corrected it in code. You can see some of that in his examples. But his code is very readable and it's easy to change/recompile the library&dll.

My driving demos on my website use Newton/IrrNewt. I spent time tweaking all the parameters in order to the vehicle driving in its current fashion - pretty good, but needing some more tweaking.

I also have a Newton demo (with source) on my website. I've been meaning to convert it to Bullet - hopefully I can get around to that soon...

Carnes · Post by **Carnes** » Sun Nov 25, 2007 11:02 am

You're not kidding about the tweaking, takes hours to get everything just right.

sio2 · Post by **sio2** » Sun Nov 25, 2007 11:51 am

If you look at the IrrNewt vehicle demo you'll see that when the vehicle is reset the wheels bounce right up through the chassis about 4 feet (IIRC). It's pretty quick so you may need to look closely to see it. That's if you can call a block with a torus on each corner a vehicle.

IrrNewt also assumes the front and rear axles are symetric, which in car models it is not.

And yes, it took a lot of work to get the SUV driving "nice". Stuff like bouncing up into the air and rolling when hitting the kerb (sidewalk) isn't good.

The biggest issue you may encounter with IrrNewt is the "single material" issue. If you look at the racetrack in my first driving demo IrrNewt treats it as one single material - no separate materials for grass, tarmac, dirt, etc. The author of IrrNewt said that the way I wanted it (kept as one single physics entity) could not be done and we had an argument where basically I said it was an easy engineering task and implied that he was incompetent. A few months later he tracked me down in another thread and started on me again; I called his bluff and explained the principles of a solution and have heard nothing since.

By all means use IrrNewt/Newton - just be aware of the limitations. For anything other than trivial/demo stuff you may want a more robust solution. Or, since you have the IrrNewt source code, you could modifiy it to your requirements.

Last I checked you couldn't download Newton or IrrNewt, though. Newton may be temporarily down; the IrrNewt author seems to be ignoring the pleas in his own thread for a valid download link.

I myself am looking to move over to Bullet but the main reason for this is that my day job is coding games on the PS3 (and XBOX) and there's a PS3 optimised version of Bullet I can grab.