On Tue, Sep 3, 2013 at 9:19 PM, Ville Syrjälä <ville.syrj...@linux.intel.com> wrote: > On Thu, Aug 15, 2013 at 10:39:31PM +0200, Vedran Rodic wrote: >> > We do have the set_caching ioctl. It's enough to flip the PTEs to UC and >> > let MOCS manage things. I actually did a few experiments on my IVB. I >> > made all Mesa's buffers UC via PTEs by patching libdrm to change the >> > cache mode of each bo after allocation. Then I fiddled with the MOCS >> > LLC bits in various ways. It definitely has an effect, sometimes making >> > things slower, sometimes faster. xonotic again seemed to benefit. IIRC >> > leaving everything LLC uncached was actually the fastest (w/ high quality >> > at least) so we may be thrashing the LLC a bit there. But eg. reaction >> > quake regressed quite a lot if most things were left as UC. >> >> Can you share the libdrm patch? > > Sorry, forgot to reply. > > Here's the patch if you're still interested.
Thanks, Just as a data point, I tried my OpenGL test application (Dota 2 on Wine with my patches to enable fast depth clear), and performance doesn't change when I use libdrm with your patch applied. I also disabled all MOCS_L3 stuff for my IVB, still no changes. Vedran _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
