[EDIT] 22.11.2010.

Unfinished sources of Bullet implementation has been published by developers. Binaries included.
http://www.neoaxisgroup.com/phpBB2/view ... 3974#33974

We will continue development of Bullet implementation later. After NeoAxis 1.0.


--------------

Hi here!

-------

I thinking about implementation Bullet physics. This looking good and polished. Demos are good too.

-------


Why need Bullet?

First, we need fast physics on MacOSX. As we know, PhysX is not support MacOSX. Epic fail. Then we can use only ODE.

Second, we want to have one universal, good and expandable physics library.
SoftBodies, Cloth, etc will be great i think. ODE is not support this.

Also Bullet physics soon (or already) will have OpenCL support for better performance.

PhysX fails: closed sources, no MacOSX support, vendor-lock for nVidia cards.
ODE fails: Slower, no softbody, cloth and some other advanced features support.

When Bullet physics will be implemented, we will remove support of ODE and PhysX (sources will be shared of course).

So, Community, what you think about it?
Pluses? Minuses?

I think your choice is good for developing the tool but I also think that you shouldn't remove PhysX as it is actually the "standard de facto" in physics.

It seems a necessary solution for the new development.

i like this.

Code: Select all   Expand view

SoftBodies, Cloth, etc will be great i think. ODE are is support this. 
 

and this!
but NvidiaPhysix is a also standard in phisics....
the problem is if someone has managed the physix code in NA....
I vote yes for Bullet. but i think we need other choices...

Hello,
I think Bullet is a good choice.
PhysX, with its vendor related status, always made me think twice before using it. Yet, beside the fact I had stability problem with it (Big masses and huge forces), it's still a standard (robust and validated).
ODE has always been more stable for me, but is slower.
So, I can try a new solution !

Regards,

Sounds like a good option here also, being fairly new to the scene and at the stage where were are playing with Vehicle Physics, I see some great Bullet demos out there.

Go for it.

I am not sure you should get rid of ODE and Physics Maybe just add bullet as another option, atleast until we all can test bullet and make sure it is suitable for our games.. I use ODE and takeing that out will more then likely break my game all together.. and probably a few other's games.

I'd say go for it!

PhysX is fast but has many quirks and tends to get unstable unless you are very careful about your physics usage. Nvidia has also tightened their vendor lock in with their drivers. It used to be possible to have one Nvidia GPU for PhysX acceleration and ATI for rendering. Now the driver detects if an ATI GPU is present and disables PhysX acceleration if so.

NeoAxis should also strive to be platform independent and this is another great step in that direction. Bullet Physics is also gaining popularity fast and I believe has already proven itself on commercial titles.

Edit: Of course it must be tested carefully so people who has invested a lot of time using ODE or PhysX doesn't run into severe problems. Maybe they can be kept for a while giving the user an option to select them and then at some point in the future remove them if ppl are okay with it...

So long as we don't have stuff like this happening in our games I'm all for it: http://www.gametrailers.com/video/ai-lo ... 2033/63219 (fast forward to 2:18 to see)

I didn't know this.
In this case I revert my opinion and think surely moving to a completely free library like Bullet.
Also there is a quite nice news right now on the site
http://bulletphysics.org/wordpress/?p=175

Good to send to f**k off Nvidia and their closed solution.
I wonder how still software producers can think that so widespread libraries and softwares can remain closed and still live.
They still don't understand that now or later somebody will come out with a more free solution and they'll have to open their too to not lose popularity.

Uh, there's editor only for Maya... hope they'll set up one for Max soon.

I think PhysX will mean more to people looking at NeoAxis, as it's seen as the near 'standard' for physics on Windows machines. Most people I know have never heard of 'Bullet Physics' and will just be confused and move to another engine that uses stuff they've heard of.

@GhostaCoasta : not sure about what you say. People interested in using an engine will try it before looking somewhere else. If Demos are convincing, they may look further (features etc...). Then, they will start to ask the question of the reliability of it's component (stability, support, future evolution...).
People that would turn only on the physics library name are mainly people that are searching for "curiosity" (which is good also ! I state no judgement), and they are not really making any choice, just experiment. So they wont stick to any engine for now, while in experiment phase.

If the engine do it's job, anybody "trying" the engine will be convinced, whatever the library. Then, if one asks itself about "using an engine for sometime", they will asks for some guaranty, and that's where PhysX uses its name, and then again, only "true" guaranty (not only name, but feature, seriousness of the project and related libraries) should really make somebody stays (here, I agree with you because of the wide usage of the PhysX library, that's an implicit guaranty). Yet, as this "market" is fast evolving, almost anybody now how fast a big name can become obsolete... and vice-versa.
In my opinion Indies are exactly the kind of people that will try "a not so famous name" based on actual performance and "vendor constraint-free" solution (more than a big studio, with support contract from Nvidia's and such), for the sake of its freedom or budget reason.

So I only partly agree with you, and as NA is targeted at Indies and such, it seems appropriate to try it. But indeed, this choice means that the engine will have to be convincing with a not-yet very famous solution.

Regards,

+1 for bullet. i have neither been impressed by ODE or PhysX and both have had their problems for me over the years. I also find that most of my physics models either work in ODE or PhysX, but rarely both. I personally think it would be a great idea to have a single physics lib and concentrate all future development on that, no matter what the selection is. That should equate to a more stable and robust physics solution for NA.

@GhostaCoasta, if anyone is going to drop use of an engine cause it doesn't have PhysX, or get 'confused' by the use of a different physics lib then they probably have no business developing a game anyway. it has already been stated about the vendor locks with PhysX, so if someone will drop an engine cause it doesn't have PhysX, you are also snubbing a great deal of your potential customer base.

I support any move that will increase my customer base when my game hits the shelves.

+1 for bullet!

Totally agree. Go for bullet!