H i Nikita, Handling updates to lots of dynamic models can be done with update callbacks and needn't be a bottleneck, in your case right now it's not a bottleneck - the cull, draw dispatch are so it's these areas that you need to look at optimizing.
I would however suggest that you do realistic benchmarking, creating a regular grid of models as using this as a benchmark is very likely unrepresentative of what you application will have to manage - I'd guess it's pretty unlikely you'll have 100 aircraft on screen at any one time. It's important to do benchmarking on realistic cases as you may find yourself wasting lots of time trying to optimize for botttlnecks that never occur in real applications. For instance much of the bottleneck you see in your models is down to what I'm guessing are transforms in the scene that are positioning the models. If the scene has the 100 planes distributed over the work in a representative way most will be culled higher in the scene graph that then transforms so the cost of visiting that subgraph will disappear from the cull, draw dispatch and draw GPU. Once you have built yourself a representative test and got benchmarking results then it'd be appropriate to start looking at the relative costs of various operations and what you can do to lower these. There are so many different things you can do when optimizing scene graphs that right now it'd be premature to start throwing lots of ideas at you, plus it's just a waste of our time. Robert. Robert. _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
