[ODE] Increasing number of particales changes physics

[Roland Kindermann]

I have no performance problems. Currently I use ODE only for collision
detection, not for simulating physics. I do not have one sphere for each
particle but only one sphere for all particles. When I move a particle,
I move the sphere to the particle's new position and perform a collision
detection. Thus collisions are only detected, when a particle moves into
an object, not when an object moves into a particle (which means in
worst case, that the collision is detected one frame too late).
I currently have up to 1000 particles simultaneously.
Cheers,
Roland Kindermann

Jacob Ole Juul Kolding wrote:
> Are you experiencing a similar problem with slowdown?
> And by slowdown I don't mean that the app will simply run slower, for me 
> the slowdown seems to be exponential eventually leading to the app 
> hanging and not responding.
> 
> Another weird thing. Most of the time my particles (using spheres as 
> geom) wont acquire torque when colliding with the floor.
> I have however seen torque being applied in some situations just before 
> the app dies. which can happen when exceeding 350 particles but also 
> when i increase the detail of the particle mesh (using cylinders right 
> now) which as far as I see shouldn't affect ODE since I'm not using 
> trimesh?
> 
> /Jacob Kolding
> 
> On 8/12/06, *Roland Kindermann* <iyo at gmx.de <mailto:iyo at gmx.de>> wrote:
> 
>     Hi,
>     I work with particles too. I think it would be nice to have a Point
>     Geometry (which only consists of one point) for particles. It would
>     probably be faster than a small Sphere.
>     Cheers,
>     Roland Kindermann
> 
>     Jacob Ole Juul Kolding schrieb:
>      > OK I have a fixed step time and that works fine.
>      >
>      > The next problem is when I enable collision (So that the particles
>      > bounce against a floor). When the simulation has up to 350
>     particles it
>      > seem to run smoothly and stable, but above 350 it will eventually
>     suffer
>      > a slowdown death causing the app to be unresponsive.
>      >
>      > I switched to HashSpace and set the category/collide bits so that
>     the
>      > particles should not collide with each other but only the floor.
>      >
>      > Is this a bottleneck?
>      >
>      > /Jacob Kolding
>      >
>      >
>      > On 8/11/06, *Jon Watte* < hplus at mindcontrol.org
>     <mailto:hplus at mindcontrol.org>
>      > <mailto:hplus at mindcontrol.org <mailto:hplus at mindcontrol.org>>> wrote:
>      >
>      >     Correct. You need to use a fixed time step. Such as in
>      >     http://www.mindcontrol.org/~hplus/graphics/game_loop.html
>      >     <http://www.mindcontrol.org/%7Ehplus/graphics/game_loop.html>
>      >
>      >     Jacob Ole Juul Kolding wrote:
>      >     >  I'm implementing a particle system in ODE and have
>     encountered the
>      >     >  problem that when the number of particles increase the
>     physics change
>      >     >  in the the way the even though the same force is applied,
>     the more
>      >     >  particles are in the simulation the longer they will travel.
>      >     >
>      >     >  My guess is that since i step the world each frame, when
>     the frame
>      >     >  rate drops the step time increase which changes the
>     outcome of the
>      >     >  simulation?
>      >     >
>      >     >  Anyone have an idea how to help this problem?
>      >     >
>      >     >  /Jacob Kolding
>      >     >
>      >    
>     ------------------------------------------------------------------------
> 
>      >     >
>      >     >  _______________________________________________
>      >     >  ODE mailing list
>      >     >  ODE at q12.org <mailto:ODE at q12.org> <mailto: ODE at q12.org
>     <mailto:ODE at q12.org>>
>      >     >  http://q12.org/mailman/listinfo/ode
>      >     >
>      >
>      >
>      >
>      >
>     ------------------------------------------------------------------------
> 
>      >
>      > _______________________________________________
>      > ODE mailing list
>      > ODE at q12.org <mailto:ODE at q12.org>
>      > http://q12.org/mailman/listinfo/ode
>     <http://q12.org/mailman/listinfo/ode>
> 
>     _______________________________________________
>     ODE mailing list
>     ODE at q12.org <mailto:ODE at q12.org>
>     http://q12.org/mailman/listinfo/ode
> 
> 
> 
> ------------------------------------------------------------------------
> 
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode