On Fri, Mar 18, 2016 at 07:16:37PM +0000, Chris Wilson wrote: > Not sure how best to put this to use yet, we need a set of benchmarks to > track changes as well as a few hard fail criteria. But at least this is a > start towards measuring how long we block signals whilst in the driver. I think I'm leaning towards using an ioctl preloader to measure siglatency of real applications and asking who actually cares about siglatency within the render thread? X and other single threaded monoliths. struct_mutex contention being far more important. So write a tool to poke around applications and wait some future fireworks. -Chris -- Chris Wilson, Intel Open Source Technology Centre _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
