On Sep 15, 2016 6:41 AM, "Emil Velikov" <<a href="mailto:emil.l.velikov@gmail.com">emil.l.velikov@gmail.com</a>> wrote: > > On 15 September 2016 at 10:56, Emil Velikov <<a href="mailto:emil.l.velikov@gmail.com">emil.l.velikov@gmail.com</a>> wrote: > > On 15 September 2016 at 02:54, Ilia Mirkin <<a href="mailto:imirkin@alum.mit.edu">imirkin@alum.mit.edu</a>> wrote: > >> On Wed, Sep 14, 2016 at 9:52 PM, Ilia Mirkin <<a href="mailto:imirkin@alum.mit.edu">imirkin@alum.mit.edu</a>> wrote: > >>> Recent changes have been made to the VDPAU state tracker to make it > >>> unusable with nouveau. Don't provide users with an awfully slow > >>> "hardware" decoding option. > >>> > >>> [To preemptively answer the question that will invariably be asked - > >>> this is due to the state tracker's use of PIPE_BIND_SHARED, which > >>> nouveau uses to force GART placement to make things with with PRIME. > >>> However when this is used for output surfaces, which are then blended on > >>> (the most common way of implementing an OSD), this results in > >>> *incredibly* slow operation.] > >>> > >>> Signed-off-by: Ilia Mirkin <<a href="mailto:imirkin@alum.mit.edu">imirkin@alum.mit.edu</a>> > >> > >> Oops, meant to add a CC to mesa-stable, since the breakage was > >> introduced in 12.0. > >> > > Just to double-check - by "breakage" and "unusable" you mean that it's > > "awfully slow" correct ? Yeah. Nowhere close to able to keep up if blending is enabled, which happens whenever an osd is displayed. > > > > Do you/others have any plan to update the state-tracker and/or nouveau > > to brings things back on par ? No plans from me. > > > Scratch this question - I've just noticed the VDPAU patch which makes > this one obsolete. Either this or a "make it work again" patch needs to land. -ilia