On Wed, 12 Sep 2012 11:12:02 +0200, Roberth Sjon??y <roberth.sjonoy at gmail.com> wrote: > Hello > > Are you psychologist? /joke I just want to know the workloads that are important to you, to make sure they are included in my benchmarking. > I wonder if it's planned to be faster, because beside the performance > it works very good for me with the right settings regarding the > tearing issue. If performance is still under heavy developement? There are a couple of kernel features in the pipeline to speed up some frequent paths, and I plan to do an attack of the gen4+ shaders. Besides a few anomalies in x11perf, the ratelimiting step at the moment is the Render protocol which doesn't provide complete or optimal offload of even the <canvas> spec. And of course, there is plenty within the Render spec that is left to optimise if it finds a use in the wild. As for the tearing issue, there is no nice solution this side of Wayland. The recent hardware has been designed around a compositor, and it is likely that more and more features of the display engine will require compositing. It appears that TearFree is unavoidable and so will need lots of love and further development. -Chris -- Chris Wilson, Intel Open Source Technology Centre
