[ODE] degrees of freedom in a slider - other joint types
STenyaK (Bruno Gonzalez)
stenyak at gmx.net
Thu Aug 31 15:17:42 MST 2006
On Fri, 01 Sep 2006 00:01:01 +0200, Jon Watte (ODE)
<hplus-ode at mindcontrol.org> wrote:
> What do you mean by "struggling" ? 5 joints and 5 bodies should do
> nothing. If you hook them all to the chassis, you get a 21 body island,
> which is a little more work, but no CPU built in the last 5 years should
> be "struggling" with that, by any means, at least if you're sticking to
> reasonable update rates (60-100 Hz) and handle collision well (no
> self-collisions, etc). And using a release library build :-) Maybe with
> dWorldStep(), it being O(N^3) in island size and everything, but is
> anybody still using that?
> I haven't actually measured the CPU usage of RayCar, but it's bound to
> be minimal. Of course, that's a smaller system.
In the last months, I'm getting very low framerates when my car touches
the ground (and suspensions therefore begin moving/working).
I've measured the times, and it's the physics engine that creates the
bottleneck (not graphics nor input), taking up to 90% (average is 50%) of
the cpu time used by my program.
Then I've tried to measure collision time vs. solver time, and it looks
like both take about the same amount of time. I'm not sure if it's because
of the low precission of SDL timer, but the average was 50-50%... So it
looked to me like solving the joints takes a lot of time. But maybe i'm
assumed wrong things, or i'm doing something wrongly with ODE...
Site: http://1ksurvivor.homeip.net <1kSurvivor>
Mail: stenyak AT gmail DOT net
More information about the ODE