On Tue, Dec 14, 2010 at 11:56 AM, Robert Osfield <robert.osfi...@gmail.com>wrote:
> Hi Steve, > > I find it very frustrating that the OSG isn't scaling as it should. > The basic design should scale, so I believe there will be some issue > with the implementation, most likely in high level context/swap buffer > management. > > Of all your findings the following statement is one I think two are > particularly significant and worthy of further exploration: > > > For example if the frame rate is 100 using one card, the frame rate > > drops to 50 for 2 cards and 25 for 4 cards in use. If the > > application worked properly, the FPS would be 100 in all cases. > > The interval nature of the performance drop is suggests to me that > something like the swap buffers is being serialized so that each swap > is happening on a different frame boundary. > Presumably that would be clear in the statistics bar graph. If not, the patch I submitted a little while ago to color the bars by frame might be useful. For example, with that patch it becomes immediately clear that, with vsync enabled on recent Nvidia cards, The GPU work for a frame doesn't start until after SwapBuffers is called. Tim
_______________________________________________ osg-users mailing list osg-users@lists.openscenegraph.org http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
