Horde3D

I read in the documentation about Horde's use of bounding volume hierarchies to optimize visibility detection, but exactly which techniques are employed I have yet to find documentation on. So my question is this, are there already frustrum culling algorithms in the engine or will I have to make those myself? I have the knowledge to implement such algorithms, but I want to make sure Horde doesn't implement such algorithms already before I invest the time in doing so.

Frustum culling is implemented and used for lights and the camera. I don't know exactly how marciano has implemented the bounding volume hierarchies, but as far as I'm aware of, each node in the scene graph has an AABB that includes all child nodes. So if a parent node is culled against the frustum, all child nodes should be culled too. But since I don't have looked at the source code to answer your question, don't take me for granted.

As Volker says, Horde is doing frustum culling based on AABBs. We used to have an AABB tree that was tied to the scene graph. However, we refactored the system so that we have a separate spatial graph besides the scene tree, which can store spatial relationship independently of the transformation hierarchy. The current implementation of the spatial graph is a flat list. A more optimizing solution may be implemented in the future. See also http://www.horde3d.org/forums/viewtopic.php?f=8&t=612