[FFmpeg-devel] [PATCH]Implement max_lowres

[Michael Niedermayer]

On Wed, Jul 07, 2010 at 03:56:22PM +0300, Ivan Kalvachev wrote:
> On 7/6/10, Michael Niedermayer <michaelni at gmx.at> wrote:
> > On Tue, Jul 06, 2010 at 12:01:26AM +0200, Carl Eugen Hoyos wrote:
> >> Hi!
> >>
> >> On Monday 05 July 2010 04:40:59 pm Michael Niedermayer wrote:
> >>
> >> > > First version attached. Is there an API-change or Changelog entry
> >> > > needed?
> >> > > Note that this should not be held back because of the obvious
> >> > > regressions: ffmpeg -lowres is currently broken for default configure
> >> > > anyway;-)
> >>
> >> > i dont think mpeg1/2 are the only that support lowres
> >>
> >> As I said, this should currently be no reason to delay this patch.
> >
> > i disagre
> > only codecs that are confirmed to fail / not support it should be limited
> > not limit by default and loose feature in all untested codecs
> 
> I don't like the whole lowres thing.
> Long time ago a friend wanted to use cheap Pentium166 laptop to play
> some mpeg4 videos.
> 
> In my tests lowres=3 (that is resolution decreases by 8x8, or 1 pixel
> per mpeg1/2/4 block) gives just 30% speedup (that is, 70% of the
> normal full resolution playback). The same amount of speedup is much
> easier achieved by gray only decoding (skip color planes), that also
> gives image of much higher quality - full resolution and no
> accumulation of decoding artifacts.
> 
> Indeed lowres was clever idea and intriguing experiment, but I
> consider that it is practically useless.
> 
> So I would like to ask, would you consider removing lowres altogether?

you want a feature to be removed?
anyway, the amount of speedup depends on the file and where the bottleneck
is. With very very high bitrate videos the gain from lowres is small
also as long as we only support sdl, its absolutely needed as workaround
for ffplay to be useable because sdl fails randomly above some resolution with
some hardware

besides i wouldnt expect much meassureable gain for the normal case by
removing it.

that said, i wont approve lowres being removed, make it compile time optional
if you like but its usefull for other people, i am one of these people btw.

[...]

-- 
Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB

Old school: Use the lowest level language in which you can solve the problem
            conveniently.
New school: Use the highest level language in which the latest supercomputer
            can solve the problem without the user falling asleep waiting.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20100707/6c687834/attachment.pgp>