Hi Joan, On Mon, Jun 6, 2011 at 9:30 AM, Joan Navarro <[email protected]> wrote: > I have a Scene with a lot of containers representeds with boxes at this > point. I want to put in the scene a ton of them and each container is a PAT > node in my scenegraph. > > I just want to know wich are the best techniques to improve the frame rate in > this example.
Using lots of seperate box geometries and lots of seperate PositionAttitudeTransforms will leads to a scene graph that has a bottleneck in both cull and draw as there are simply too manage seperate objects. In the case of transform nodes they are particularily expensive for the scene graph to traversal as each transform node requires the view frustum to be transformed into the subgraphs local coordinate frame - we've optimized this as much as we can but in the end a transform node will always be more expensive that just a straight node. As a rule of thumb one can typically scale to having thousands of transforms on screen at any one time without major performance difficulties in cull, but scaling to ten's of thousands will introduce a bottleneck. The osgforest example has one subgraph path that tests the peformance of transforms so have a look at this to see how things scale. Since transform are expensive if you have very large numbers of them it's an obvious one to optimize away. How to do it will depend a great deal on your hardware constraints, the amount of data, whether the data itself is a static geometry, and whether the transforms are static, and just how many objects you have in the scene, whether the geometries are identical etc. Only you can answer these questions so please dive in a bit deeper about these aspects and what peformance you are looking for, what you are getting now. Once others know a bit more information we might be able to guide you. Robert. _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
