Markus Hein wrote: > > > I will have this in mind, it seems that this wasn't needed for my tests > right now. Why did you need the custom drawable ? Could you explain > this? Thanks. > >
I had to do this so that the Compute code was always run. Sometimes, specifically with multiple compute nodes, only one of the nodes would execute. I'm not really sure why. It was not compute program specific, it did the same thing if it was a frag/vertex program as well. If I attached the program to the drawable node I could assure that it ran every time. Markus Hein wrote: > > btw.: if you are sitting on some nice higher level ParticleSystem > implementation, using osg::SSBO and osg::SSBB, maybe this could be a > good starting point for a new GPU-based ParticleSystem Library for the OSG ? > > Over more than a decade.. we have used the osg::Particle library a lot > for our needs , but now it seems that some GPU-based approach must take > over.. osg:Particle library was sometimes complicated to work with. > > Unfortunately I don't have this working yet. There are some problems that you run into, specifically the number of SSBOs that you can create is finite, and you run out fairly quickly. I got around this by creating one big buffer for all systems of a particular time. I had to create a memory buffer manager to pass open blocks of memory from particle system to particle system. Very complex and unfortunately very specific to my implementation. I plan on abstracting this out soon but it will take some time to do so. I will say that in terms of speed, this approach is light years faster than using osg::Particle. I'm using roughly 100 times the number of particles I was using before and its about 10 times faster. Not too shabby. ------------------ Read this topic online here: http://forum.openscenegraph.org/viewtopic.php?p=63226#63226 _______________________________________________ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org