[MPlayer-users] video does not scale to fullscreen with -vo xv

[Andy Furniss]

Reimar Döffinger wrote:
> On Sun, Mar 06, 2016 at 11:01:56AM +0000, Andy Furniss wrote:
>> Reimar Döffinger wrote:
>>>> Showing output from playing something would confirm but I guess it's
>>>> using x11 software output/scale/csc which is not as good as other
>>>> hardware accelerated outputs like gl.
>>>>
>>>> If you are getting -vo x11 then adding -zoom will allow it to scale.
>>>
>>> Which will work not very well either.
>>> If it is x11 vo then then also means XVideo is not working (or not
>>> compiled in is an option, too).
>>> Or if it is using xv, then it is broken.
>>> Either way it seems very likely the primary answer is "fix your graphics
>>> card drivers first".
>>
>> Yea, but if they turn out to be working/get fixed it's still intel, so
>> at the risk of being undiplomatic here I would consider trying mpv as
>> it should be able to use vaapi for h/w dec/post processing.
>
> There are wrappers to use VA-API as VDPAU (libvdpau-va-gl1) and have
> it working with MPlayer.

Yea, it's a while since I played with intel, but IIRC it only worked
for vo rather than decode, but then I didn't try very hard.

> Though for the normal use-cases (and at least if power draw is not
> a major concern) hardware acceleration is just an unnecessary pain.
> It's often slower (limited to e.g. 60 FPS when the CPU can easily do
> 300, which has some impact on seek performance for example), has
> really pathetic error concealment, lacks many of the encoder bug
> workarounds needed especially for MPEG-4 ASP content etc.

True, but, skylake can probably do hevc which is not so easy and the
OP has a laptop.

cherryview/braswell which I think are similar gen but low power can
do main level 6.2

https://01.org/linuxgraphics/documentation/hardware-specification-prms

Shame it's only 8 bit, looks like you'll need broxton (whatever that is)
to get 10bit -

http://cgit.freedesktop.org/vaapi/intel-driver/log/

I know there are vaapi patches for mplayer - it would be good if they
could be fixed. Since the thread on devel stopped I have retested (AMD)
several times in the hope that I could declare working, but no luck.

I'll retry again as time permits, because it seems a shame that they
apparently worked OK on intel.

Ignoring hw dec the post processing on recent intel chips is quite good
when I was testing a baytrail (10W TDP) it has a nice motion adaptive
de-interlace with edge interpolation which would be expensive for HD on CPU.