As I understand it, the problem is with floating point operation inaccuracies right?
So would running bullet on a 64bit platform ease the problem a bit allowing for greater mass ratios?
The mass ratios problem
-
- Site Admin
- Posts: 4221
- Joined: Sun Jun 26, 2005 6:43 pm
- Location: California, USA
Re: The mass ratios problem
No, it is not due to FPU accuracy. Iterative constraint solvers, including the one based on PGS in Bullet, don't converge fast enough towards a solution.
Improving support for larger mass ratios can be achieved with a different constraint solver, a smaller internal time step and more constraint iterations.
Thanks,
Erwin
Improving support for larger mass ratios can be achieved with a different constraint solver, a smaller internal time step and more constraint iterations.
Thanks,
Erwin
-
- Posts: 71
- Joined: Thu May 22, 2008 8:03 pm
Re: The mass ratios problem
Would continuous collision detection help this?
Also, I read on Newton's site that they dont use Iterative solvers, and that (or something else in there?) makes Newton deterministic, any idea what they use?
Also, I read on Newton's site that they dont use Iterative solvers, and that (or something else in there?) makes Newton deterministic, any idea what they use?
-
- Posts: 508
- Joined: Fri May 30, 2008 2:51 am
- Location: Ossining, New York
Re: The mass ratios problem
I gave the ODE solver a try but it didn't work at all - I'm using DiscreteCollisionWorld, and none of my object had any friction with the ODE solver!