[ODE] Difference in efficiency between single trimesh geomandmultiple trimesh geoms?

[Erin Catto]

I suspect a single AABB tree using OPCODE would be faster. Each query should
be O(log N) on average. People usually only break up trimeshes because using
shorts for indices saves a lot of memory. Another reason to breakup
trimeshes is for gameplay reasons, such as destructible terrain.

Erin

-----Original Message-----
From: ode-bounces at q12.org [mailto:ode-bounces at q12.org] On Behalf Of Megan
Fox
Sent: Sunday, February 20, 2005 12:31 PM
To: John Miles
Cc: ode at q12.org
Subject: Re: [ODE] Difference in efficiency between single trimesh
geomandmultiple trimesh geoms?

The thing is that OPCODE already contains a separate space
partitioning system, so it's a little unclear to me if it's a better
idea to let OPCODE handle everything unmolested, or to keep sorting
things my way with the chance that ODE's various optimizations will
start conflicting with OPCODE's, causing overall slowdown.

-Megan Fox

> I'd be amazed if consolidating separate geoms into one big representation
> actually improves performance.  It's much better to keep objects separate
> and aggressively disable the ones that aren't candidates for collisions
> (potentially even removing them from the space).
> 
> dSpaceAdd() and dSpaceRemove() are pretty fast in my experience.  They are
> your first, best optimization.
> 
> -- jm
> 
> > -----Original Message-----
> > From: ode-bounces at q12.org [mailto:ode-bounces at q12.org]On Behalf Of Megan
> > Fox
> > Sent: Sunday, February 20, 2005 11:01 AM
> > To: ode at q12.org
> > Subject: [ODE] Difference in efficiency between single trimesh geom
> > andmultiple trimesh geoms?
> >
> >
> > Has anyone done any testing on this, where they started with a world
> > full of one trimesh geom per "world chunk entity" (a flight of stairs,
> > a door, a piece of wall, a table, etc), then later moved to a
> > representation where everything was merged into a single trimesh geom?
> >
> > As it stands, each chunk in my world is defined via solid geoms, with
> > said geoms associated with the proper entity - "wall", "wall corner",
> > "floor", "stairs", etc.  Moving to trimesh representation of
> > everything (in prep for terrain), I imagine I'll keep it the same,
> > with one trimesh geom per.  I'm just curious how much of a speed
> > increase I'm likely to see in moving from this to a single-geom
> > representation of the world, said figure serving as a motivator for
> > changing things in that direction.
> >
> > -Megan Fox
> >
> 
> _______________________________________________
> 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