[Bug 84944] tearing on radeonsi vdpau deinterlacer

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=84944

--- Comment #23 from Andy Furniss <adf.lists at gmail.com> ---
(In reply to Christian König from comment #13)
> (In reply to warpme from comment #11)
> > Maybe Kabini's xserver log output about "EXA not supported" is clue here? Is
> > EXA supported on Kabini at all?
> > 
> > heh - I spent great time to adding glamor to carefully minimized minimyth2
> > image - just because on many places it is stated that mesa's radeonsi
> > mandatory requires glamor.... 
> 
> Sorry you misunderstood me: Glamor is mandatory on SI, there is not support
> for EXA. This is correct.
> 
> But glamor shouldn't be used in this situation because we have only one
> fullscreen application on the screen. In this situation page flipping should
> be used by the X server without any copy operation required by glamor.
> 
> Page flipping operations are tear free, so what we see here is that without
> deinterlacing the graphics stack uses page flipping and with deinterlacing
> it uses glamor with of course results in badly tearing of the video.
> 
> The question is why the heck do we fallback to a copy operation with glamor
> when deinterlacing is enabled? Deinterlacing is independent of displaying
> the video so this makes no sense.

Maybe de-interlacing just tips him over into not being fast enough - just
speculating.

A couple of observations from my setup - not kabini, but si R9270X.

With Michels 2nd patch I see mplayer and mpv both get 1920x1080 OK so I
pageflip.

My normal use case, however, is to have 2 screens - TV below monitor +
different refresh rates. Like this I don't get flips, but I also don't have any
tearing issues with deint. I always set cpus to perf - so I guess that my
system is fast enough to handle copy.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/dri-devel/attachments/20141026/8ad2cb8d/attachment-0001.html>