Scene manager design
Author
I'm at the point right now where I'm trying to design a simple but effective scene manager. I've been reading over this article on how to create a scene manager, as well as the replies on that article and it really put me into some confusion. Someone in the replies give a different model of an efficient scene manager, but I'm still having problems understanding how it works. This article on it is easy to understand, but how efficient is it compared to the model explained in the other articles thread?
Hello.
I think you'll find that most implementations of a Scene Graph that you create are going to be fairly similar in terms of performance -unless- you go out on left field and do something... odd. :) Also, it very much depends on the optimizations that you plan on from the beginning to have present within your system. For example, how do you plan on dividing the scene up?
This is an important question because you should not get confused between the Scene Graph - which is a logical ordering of objects in your scene - and a spatial partitioning/geometry management object that is responsible for culling nodes within your scene graph hierarchy, I really can't stress this enough. (You may already know this, so ignore as you see fit).
In my scenegraph, I have placed the patial partitioning within the graph hierarchy and use it to cull objects from the current cameras view frustum, this involves the SG object having access to the spatial spartitioning system. The reason I do it this way is very simple: If I ever wish to change the spatial partitioning system, I can just yank it out, create a new class that derives from the base node object and plug in it, viola, new SP system implemented in any given title.
Here's a pertinent link:
http://triplebuffer.devmaster.net/articles/scene_graph_dx/
I think the person who maintains this website goes by the name Yann on this board, and also wrote several very large and interesting posts explaining SGs in great detail a while ago. If you search for them, you should be able to find them.
Hope this helps.
I think you'll find that most implementations of a Scene Graph that you create are going to be fairly similar in terms of performance -unless- you go out on left field and do something... odd. :) Also, it very much depends on the optimizations that you plan on from the beginning to have present within your system. For example, how do you plan on dividing the scene up?
This is an important question because you should not get confused between the Scene Graph - which is a logical ordering of objects in your scene - and a spatial partitioning/geometry management object that is responsible for culling nodes within your scene graph hierarchy, I really can't stress this enough. (You may already know this, so ignore as you see fit).
In my scenegraph, I have placed the patial partitioning within the graph hierarchy and use it to cull objects from the current cameras view frustum, this involves the SG object having access to the spatial spartitioning system. The reason I do it this way is very simple: If I ever wish to change the spatial partitioning system, I can just yank it out, create a new class that derives from the base node object and plug in it, viola, new SP system implemented in any given title.
Here's a pertinent link:
http://triplebuffer.devmaster.net/articles/scene_graph_dx/
I think the person who maintains this website goes by the name Yann on this board, and also wrote several very large and interesting posts explaining SGs in great detail a while ago. If you search for them, you should be able to find them.
Hope this helps.
Yep, that's actually IFooBar's page. However, here are some links to some pretty good scenegraph topics by Yann:
Terrain and Scene Graphs
Octrees
Scenegraph Management
Terrain and Scene Graphs
Octrees
Scenegraph Management
Author
Well, what I was thinking about was creating a scene graph for ordering scene nodes, but when it comes to rendering that will all be handled by a rendering graph. The scene node when updated will pass data pointers to the right functions in the render graph. The render graph will then call a function that's derrived from the render node base class that checks what data is in the view frustum, then it's returned and sorted in an efficient order. Once everything is called in the scene manager, you call render on the render manager and it draws everything visible.
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement
