[ODE] Fixed stepping
Jon Watte (ODE)
hplus-ode at mindcontrol.org
Sat Aug 19 13:56:16 MST 2006
Yes, there is still reason to use fixed step sizes, because otherwise
your simulation (or the performance of space ships) is not consistent
between runs. A first order integrator behaves very differently (i e,
how gravity affects an object) depending on step size. This is why
higher frame rates made you jump higher in Quake, for example.
We've already documented, in many places, that fixed step size is
desirable. However, it's desirable so much that we really should make it
a hassle to NOT use fixed step sizes.
Bill Sellers wrote:
> I'd agree that there's often a lot to do and hiding simulation steps
> isn't a good idea. Also if you have a system where there are no
> collisions (simulating a fixed mechanism for example, or a spaceship)
> is there actually a problem with changing the stepsize? Rather than
> spending a lot of effort changing the API all that needs to be done
> is better documentation and making sure there are good examples of
> how to do things properly...
More information about the ODE