Hi Kirill, I would recommend taking a step back from doing specific attempts at optimization and work out from high level what you are trying to achieve and the performance you require and the hardware/OS/OpenGL version combinations that you must target.
So could you please tell us what you application is trying to do. Also tell us a bit about the models you are trying to place in your scene, how many geometries, state, nodes etc. How many instances of models you want in the whole scene and how much you might expect on screen at one time. Also tell us about the OS, OpenGL version and hardware you plan to target. Tell us about the performance targets you have. Tell us what results you have got already. When doing benchmarking you absolutely must use a release build of the OSG and your application, doing performance profiling with a debug build is absolutely the most misleading thing you could do. Things you should be looking for in profiling is frame rate, update, cull and draw dispatch and draw GPU times, number of nodes, geometries, state in the scene and on screen - use the StatsHandler to collect stats and report them on screen (see osgviewer.cpp for how to attach a StatsHandler). Robert. _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
