On Monday 7 June 2010, Siva <[email protected]> wrote: > Hi, > I know that GlxGears is not a benchmark, but I'm need to validate the > performance of a graphics driver that I'm currently working on. > > What does Glxgears test exactly, when it reports fps? XServer graphics > pipeline or the graphics engine capability to handle commands? > I mean if there is a drop of 100fps, is it an indication of > significant performance drop? > The best place to look is the source code for glxgears - it's not a complex program. Basically, it's not considered a useful benchmark because it pushes a tiny number of shaded (untextured) triangles - well under 1,000 per frame - then swaps buffers.
As a result, on any modern system, the time taken to process the scene is almost completely dominated by the time it takes to execute glXSwapBuffers. If you're interested in testing general 3D performance, http://dri.freedesktop.org/wiki/Benchmarking has some decent benchmarking ideas to get you started. If there's a specific area of your driver that you need to test (e.g. triangles per second, fill rate, texture handling), you're best off writing your own code to hammer on the codepaths that concern you. -- Simon Farnsworth Software Engineer ONELAN Limited http://www.onelan.com/ _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
