[ODE] stack requirements

Anselm Hook anselm at hook.org
Mon Jul 22 18:44:02 2002


Hi Dave, this is the dumb answer, but can you partition the problem into
pieces dynamically?  Shouldn't that be happening automatically?  Or do you
really get worst case situations?

-a

On Tue, 23 Jul 2002, David McClurg wrote:

> In the debugger I find that with 40 bodies, dWorldStep() is taking 256K of memory (which I've switched to using dStack) to get down to dInternalStepIsland_x2().  That means about 6k per body.
>
> All my ode bodies are spheres!  I'm trying to create a ball simulation where the balls can lie around, roll, and bump into each other.  I expected dSpaceCollide() to be the memory hog and not dWorldStep().  Is there anything I can do to reduce the memory requirements of dWorldStep()?
>
> -----Original Message-----
> From: David McClurg [mailto:dpm@efn.org]On Behalf Of David McClurg
> Sent: Tuesday, 23 July 2002 9:31 AM
> To: ode@q12.org
> Subject: RE: [ODE] stack requirements
>
>
> ode/src/stack.cpp/h seems to be a *unused* replacement for alloca.  is this something that never got finished or didn't work out?  Here's the interface...
>
> class dStack
> {
>   void init(int max_size);
>   void pushFrame();
>   void popFrame();
>   char *alloc(int size)
> };
>
> Can I help with this?  Do you prefer a ansi-C interface with function overrides like the other memory routines to make it public?
>
> I need to have something like this to lower my stack requirements.  Also-- "The alloca() function is machine dependent; its use is discouraged."
> _______________________________________________
> ODE mailing list
> ODE@q12.org
> http://q12.org/mailman/listinfo/ode
>