[ODE] General collision detection question

[Pierre Terdiman]

> Why would I use Opcode for CD, instead of ODE

Because they don't do the same thing. ODE only supports collision detection
between primitives (e.g. cube-vs-cube, cube-vs-sphere, etc). With ODE alone
you can't do sphere-vs-mesh, for example. And if you want to add support for
meshes, you must write something like Opcode. So you really don't use one
instead of the other, you use both.

> other than the fact that the tri-collider was written for Opcode (for that
matter, why was tri-collider written for Opcode instead of ODE)?

The so called tri-collider is just a small piece of code, the missing link
between ODE and Opcode (it creates correct contact data out of Opcode). You
can't use ODE alone, which lacks collision detection between primitives and
meshes. You can't use Opcode alone, which lacks physics-compliant results
(it's only "naked" CD, if you want).

> In short: What makes some of you prefer Opcode over ODE when it comes to
collision detection?

There's no overlap between the two. They don't do the same thing. So you
don't "prefer" one or the other, actually.

> And the other question is, is there a tri-collider that can be used just
with ODE?

Yep, Opcode :) Phagocyt relevant code into ODE, done. But then you'll miss
the new optimizations / features in next Opcode versions (or you'll have to
re-do the refactoring instead of just recompiling).

To make it clearer, there are three very different tasks involved :
1) Collision detection
2) Contact generation
3) Rigid body simulation

Opcode does 1) for primitive-vs-mesh
Erwin's tri collider does 2) for primitive-vs-mesh
ODE does 3) and also 1) and 2) for primitive-vs-primitive

The hardest part is 2) IMHO.

- Pierre
www.codercorner.com