The physics has issues

#1

I created a bunch of balls to bounce on a platform (following the tutorial), and some balls just float into the air. I think there is a bug in the physics. You can test the project here: https://goo.gl/XAqoYJ

Also, the rendering is very slow when physics is enabled. Yes, I even tried setting max sub steps to 2, and higher, but I’m not confident as things scale, when there was initially only one ball in the scene.

I was hoping to use this system, but I’m not confident anymore when a simple demo fails. I’ll wait for a fix, but for now I’ll uninstall, thanks.

#2

Hmm, I can somehow agree to that, the physics seem not quite ready yet, IMHO.
I tried a simple “manual bounce ball” script which simply applies a Y force to the ball as soon as he touches the ground, but he bounces back into completely different heights at times and depending on what amount of substeps I set, even more differently (sometimes just far into the sky, while other bumps are barely noticable). Then again, when the frames drop at random times (no reason why), the physics also slow down… I thought the time step is fixed?

#3

I really look forward to the physics being fixed up. I’d love to use this system in a project one day. I may have to choose another one instead for now, but I hope the developers keep up the great work. The system is looking good. Some considerations though - they really should allow double/right-clicking script components and opening the script directly from there, instead of having to search the inventory, lol. (that also goes for the graph tree view)

#4

You might wanna create another thread about that separate issue. For now I’m still working with Godot (it will get real C# support in 3.0), and the physics look pretty good in it.

#5

The issue with bouncing balls you had in your project is related to the fact that you put the last 3 spheres inside another sphere and all of them are rigidbodies.
Local transforms are supported by the physics engine but in this particular case since all of them are rigidbodies the update order of each of them is conflicting.

If you pull out and move to root level into the hierarchy the last 3 spheres everything will work fine.
If the Sphere(6) was a static collider there would be no issue tho. Here the issue is that all of them are rigidbodies managed by the physics layer.

In terms of performance I cannot really find any particular issue. Physics in Xenko uses bullet physics. It’s quite good in terms of performance unless you have many thousands of active rigidbodies.