Can CollisionShape be shared among physic spaces?

Ali_RS · June 26, 2019, 8:14am

Java doc for CollisionShape indicates that a shape can be shared among multiple objects:

jMonkeyEngine/jmonkeyengine/blob/1ad324aa57f134078abaa8d2764bd940b6d2a337/jme3-bullet/src/main/java/com/jme3/bullet/collision/shapes/CollisionShape.java#L45


import com.jme3.math.Vector3f;
import java.io.IOException;
import java.util.logging.Level;
import java.util.logging.Logger;
/**
 * The abstract base class for collision shapes based on Bullet's
 * btCollisionShape.
 * <p>
 * Collision shapes include BoxCollisionShape and CapsuleCollisionShape. As
 * suggested in the Bullet manual, a single collision shape can be shared among
 * multiple collision objects.
 *
 * @author normenhansen
 */
public abstract class CollisionShape implements Savable {
    /**
     * default margin for new non-sphere/non-capsule shapes (in physics-space
     * units, &gt;0, default=0.04)
     */

Does it hold true for multiple physic spaces? Can we share a CollisionShape among multiple collision objects in different physic spaces?

oxplay2 · June 26, 2019, 9:45am

i dont know, but it dont have any “changing on physics tick variables” so i think yes.

sgold · June 26, 2019, 4:15pm

Sharing a CollisionShape instance between collision objects in different physics spaces ought to work. That’s how it’s designed.

Ali_RS · June 26, 2019, 4:57pm

Sharing a CollisionShape instance between collision objects in different physics spaces ought to work. That’s how it’s designed.

Okay, that’s good to know.
And in this case, I guess yet it would be safe to update PhysicSpaces in parallel, Yes? (To be exact, I mean using Java’s Parallel Stream API)

sgold · June 26, 2019, 7:44pm

Like @oxplay2 said, collision shapes are read-only with respect to physics ticks. I don’t understand your concern.

jayfella · June 26, 2019, 8:19pm

Not that this is the place for this discussion, but I’ve never personally found a viable use for parallel streaming. On the surface it sounds great but from what I know about it, it will introduce more problems than it solves. Parallel iterations will hog the entire CPU until it’s done, and in most cases manifest itself in random micro stutter and lockup’s. In a game or server I definitely don’t want that. I want a dedicated threadpool that isn’t affecting the Game Logic thread.

Anyway. I’m sidetracking.