[ODE] Box stacking problems

[Michal Bacik]

Hi,
exactly the same problems here. I've seen the slowdown, and with stacks,
there seem to happen float overflow quite often (numbers like 1e+20 and
similar). I didn't go deeper into investigating though. Maybe using VTune
would be fine during the slowdown in order to learn what's bottleneck.

-michal


> -----Original Message-----
> From: ode-admin@q12.org [mailto:ode-admin@q12.org]On Behalf Of Joe Ante
> Sent: Thursday, February 20, 2003 8:37 PM
> To: ODE
> Subject: [ODE] Box stacking problems
>
>
> Hi,
>
> I have setup a test scene containing 4 stacks with 4 cubes each.
> Then I pushed the outmost cubestack towards the others, so that all the
> stacks touch each other. When all the 4 stacks are somehow
> connected through
> contacts to each other, ODE becomes very slow(2fps on g4/667), starts
> sending messages in the lcp solver dSolveLCP  (LCP internal error, s <=0))
> and after some time crashes when trying to allocate memory in dSolveLCP:
> line 1125, dReal *L = (dReal*) ALLOCA (n*nskip*sizeof(dReal));
>
>
> I also  tried adjusting the erp and cfm parameter.
> I am using a gravity force of 9.81. (When I use a gravity of 1
> the slowdown
> becomes less and the crash seems not to appear)
>
> I am using 4 contact points for every collision and fixed timestepping (50
> frames/sec)
> I am using ode 0.035.
>
> Is this something other people experienced as well?
> Is simulating a scene like the above just too hard to solve for a physics
> engine?
> Will it improve with further releases of ode?
>
>
> If you had no problems with scenes like this, can you think of
> anything that
> I am doing wrong?
> Why does ode crash?
> Why does it get so slow.
> What does the LCP internal error mean?
>
> Joe Ante
>
> _______________________________________________
> ODE mailing list
> ODE@q12.org
> http://q12.org/mailman/listinfo/ode
>