Hello.
I’ve had an issue for quite some time which has caused me to revert to de-optimizing my application due to strange behaviors.
I have an “AtomSpatial” which is a spatial class, its simply a sphere with a material and texture.
I also have a “MoleculeNode” which is a node. MoleculeNode’s are the parents for the AtomSpatials.
Now if I call AtomSpatial.getWorldTranslation(), everything is fine (as its not optimized).
Method #1:
I change the MoleculeNode to extend from BatchedNode rather than Node, after adding all the AtomSpatials as children, I call MoleculeNode.batch().
This method doesnt seem to do ‘anything’. It doesnt screw up the coordinates of AtomSpatial.getWorldTranslation() and it doesnt increase fps either (if it does it
a minimal increase).
Method #2:
Instead of adding the AtomSpatials directly to the MoleculeNode, I create a new Node which I’ll call (“OptimizedNode”).
The AtomSpatials are then added to the OptimizedNode. Once they are all addeed, GeometryBatchFactory.optimize() is called.
Once the OptimizedNode is optimized, it is then added to the MoleculeNode as a child.
This screws up the AtomSpatial.getWorldTranslation() completly.
Method #3 (Physics Engine Unstable):
Instead of Method #2, I have tried directly adding the AtomSpatials to the MoleculeNode and then calling GeometryBatchFactor.optimize() on the MoleculeNode itself.
This caused unpredictable and odd behaviour when I attached a RigidBodyControl to the MoleculeNode and simulated physics.
I was using Method #2, because it had the best results aside from the coordinate mismatches when I used AtomSpatial.getWorldTranslation(). It seemed that AtomSpatial.getWorldTranlation while being optimized was returned getLocalTranslation instead.
Then I figured, well if I know the local translation of the AtomSpatial and I know the Transform of the parent MoleculeNode, I should be able to calculate the “BatchedWorldTranslation”:
[java]public Vector3f getBatchWorldTranslation() {
//Get the parent compoundNode Position
Vector3f batchedPos = this.getParentMoleculeNode().getWorldTranslation();
//Offset the position with the AtomModels local translation and parent compoundNode Position
Vector3f offset = batchedPos.add(this.getLocalTranslation());
//Get the rotational state of the compoundNode
Quaternion molRot = this.getParentMoleculeNode().getWorldTransform().getRotation();
//Mult the rotation with offset to get position
Vector3f newPos = molRot.mult(offset);
return newPos;
}[/java]
This seemed to be a fairly decent solution, but its odd. It’s odd because when I try to calculate the distance between 2 atoms using getBatchWorldTranslation() it seemed to give me erratic and illogical results.
Can someone please shed some light on the optimization process and what may be a valid and elegant way to retain the positions of the batched Spatials?
Furthermore can someone please explain why extending from BatchNode and calling batch() does not yield any difference in performance?
Regards,
ostenning
