Hi All, One of the challenges with optimizing modern graphics apps is that there is complex mix of parameters to paging/viewers/database/threading experiment with and also huge range of ways to measure performance/quality. Things like frame rates are often a metric that graphics developers use to measure progress, but for professional sims my guess is that its not the best metric we could come up with, hence this email.
So what metrics could we come up with to measure how well the system is performing? Here's a few to start with 1) Frame rate with vsync off. 2) Frame rate with vysnc off. 3) Number of dropped frames i.e. frames below a specified cut off. 4) Histogram of frame rates. 5) Latency, Number of frames/absolute time it takes to load a tile once its requested. 6) Plots of update, cull, draw, database pager times per frame against frame number. 7) A frame drop risk factor, i.e. how close did the frame come to being dropped. 8) Latency between beginning of the frame and the final swap buffers happening on the graphics card. Can you think of any more? It'd be good if we could come up with a overall quality metric that we could apply to whole test run. This would also entail a control application run. Animation paths gets us some way down this route, but only part way. There is also the issue of how to record all the stats for a run, and how to review them afterwards. I don't have quick answers to this, and I'm not to about to volunteer myself with coming up with such a framework (at least straight away :-) but I can see the need for it, so though I'd punt the topic out there for others to chip with thoughts on the way forward. Volunteers welcome too :-) Robert. _______________________________________________ osg-users mailing list osg-users@openscenegraph.net http://openscenegraph.net/mailman/listinfo/osg-users http://www.openscenegraph.org/
