After few self-interrogations and reading of osggpucull..it seams that variant sized TF is a feature only required for gpu culling stuff. So on, in a gl4 context this technic seams to be deprecated by the couple scatteredwrite(replacing the query)+indirectdraw...
mp3butcher wrote: > Hi, > > For the moment the transformfeedback example given is compatible with opengl > 3.3 spec... > > If i understand correctly ARB/transform_feedback2/3.txt , for opengl>=4 it's > not the correct way to handle TF as > classic VBO rely on an array_buffer (GLBufferObject), and so on it disallows > the use glDrawTransformFeedBack (because this func requires explicitly a > GL_TRANSFORM_FEEDBACK). > (Correct me if i'm wrong) > > Consequently, It forces ones to use GPU query (overhead) when the TF have > variable size... > > > Further a second issue outcomes to me reading > http://docs.nvidia.com/gameworks/content/gameworkslibrary/graphicssamples/opengl_samples/feedbackparticlessample.htm > as they use a pingpong TFBuffer to improve perf > > My guesses are to provide osg::TransformFeedBackBufferObject (inheriting > from BufferObject)and > 1)modify the GLBufferObjectManager to generate TFBuffer. (seams hard) > 2)or create a dedicated GLTFBufferObjectManager (not hard be lot of redundant > code) > 3)or create a completely dedicated TFGeometry..(no TFBuffer Factory and so) > > What do you think of these solutions? > for the moment I choose 3) cause it's seams the simpliest way to handle > pingpongTF.. > > Thank you! > > Cheers, > Julien ------------------ Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=62368#62368 _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
