[MPlayer-users] deinterlacer performance?

[D Richard Felker III]

On Thu, Jan 16, 2003 at 09:44:23AM -0500, D Richard Felker III wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> On Thu, Jan 16, 2003 at 11:24:45AM +0100, Moritz Bunkus wrote:
> > [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> > Hi.
> > 
> > While testint A'rpi's DVD a/v synch patch I noticed that deinterlacing
> > kicks my processor to its limits. Therefore I'd like to know if this is
> > normal (well NO I won't file a complete bugreport - I'm not even
> > remotely sure that it is a bug at all...).
> > 
> > My system is a Celeron 733 MHz, Matrox MGA400 with -vo xmga,
> > SoundBlaster don'taskmethecorrectmodel with -ao oss and the kernel's
> > emu10k1. It can play everything back normally without reaching 100% CPU
> > usage. (oh, mplayer is CVS from 10:00 CET)
> 
> MGA can do deinterlacing in hardware according to the docs, but it's
> not implemented in mga_vid driver right now. I might work on adding
> support for it sometime, especially if anyone else is interested. This
> would of course mean realtime deinterlacing with no cpu cost, but I
> think it might make triple buffering difficult or impossible (=>
> shearing) so I'm not sure if it's worthwhile.

BTW, another solution is to set your video mode to interlaced and use
the interlaced mga_vid patch I recently posted to mplayer-dev-eng.
This will leave the picture interlaced but since fields are being
updated at alternating times it won't look bad like displaying
interlaced video on a progressive display does. This is very
experimental though so if you want to try it, read my message fully.

BTW2, you might try using an interlaced video mode with -vo fbdev
instead, but this will have shearing and will use more cpu time for
colorspace conversion.

Rich