[ODE] performance problems when bodies lean on each other

[Hampus Soderstrom]

Hi Jon,

Thanks for your reply.

What I have is two groups of bodies with about 20 bodies in each  
connected with hinges that collide.

Setting collide and category bits helped alot. I was able to get the  
no callbacks down to 50.

thanks

Hampa

16 feb 2006 kl. 20.56 skrev Jon Watte (ODE):

>
> First, you should use the debugger to figure out why you get 180 calls
> to nearCallback 15 contacts between two bodies. Are those the only two
> bodies, or are there other bodies, too? Are you using trimeshes?
> Trimesh:trimesh collisions? (The latter is not well supported)
>
> Second, you could look into using the collide and category bits for
> geoms, if you're getting geoms tested against each other that  
> shouldn't
> be (say, two stationary trees collide against each other).
>
> Third, you could try turning on auto-disable, which will sleep bodies
> that haven't been moving for a little bit; this will make the  
> solver run
> faster.
>
> Last, what do you mean by "slow"? In release mode, have you actually
> timed the scene? What kind of performance are you getting? And does
> dWorldStepFast() help?
>
> Cheers,
>
>             / h+
>
>
> Hampus Soderstrom wrote:
>
>> Hi,
>>
>> I'm experiencing slowdown of my application when two or more bodies
>> rest on top of each other creating constant collisions?
>>
>> Every frame i get about 180 calls to nearCallback and 15 of those
>> result in a contact point. I'm using HashSpace, dSpaceCollide and
>> dWorldStep.
>>
>> Should this be slow or can I tweak my engine to handle bodies leaning
>> on each other better?
>>
>> cheers
>>
>> Hampa
>>
>>
>> _______________________________________________
>> ODE mailing list
>> ODE at q12.org
>> http://q12.org/mailman/listinfo/ode
>>
>>
>>
> _______________________________________________
> ODE mailing list
> ODE at q12.org
> http://q12.org/mailman/listinfo/ode
>
>
>