[MPlayer-users] MPlayer 0.90rc4, radeon 9000, MC, IDCT

D Richard Felker III dalias at aerifal.cx
Sun Feb 23 19:13:19 CET 2003


On Sun, Feb 23, 2003 at 05:31:49PM +0100, Balatoni Denes wrote:
> [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> On 2003. február 23. 17:24, cm5oom at post.com wrote:
> > [Automatic answer: RTFM (read DOCS, FAQ), also read DOCS/bugreports.html]
> > that was no more nvidia propeganda then the other one
> > and as far as i can tell the gatos driver does not
> > support it, maybe because it is stupid.
> >
> > so now my question is why is it a "nonsensical useless feature" and how
> > does video decoding work?
> for total lamers in video coding (eg me a few days ago) 
> http://www.vcodex.fsnet.co.uk/resources.html
> is perhaps usefull.
> imho Richard meant that if your cpu is fast enough than hardware help  doesn't 
> matter much. 

No. What I meant was the insane degree of interconnectedness required
between the codec and the video driver to use hardware MC and IDCT
makes using it prohibitively expensive (in terms of code size and
complexity). Also, it totally precludes any sort of postprocessing or
other filters, since you never have the picture in system memory to
work with.

Of course, you're also right about the speed issue. If your system is
slow enough that it can't do the MC and IDCT in software, it probably
doesn't have an AGP slot anyway... :)

> others mentioned that hardware mc/idct looks better for some reason (maybe 
> they interpolate/filter something somehow), others said it's nonsense.

That's definitely nonsense. MC and IDCT have to be implemented
according to spec or you will get random image corruption in many
movies. Most likely the drivers use some sort of filtering function in
the hardware to display the image after performing the MC and IDCT...
The same (and much better) effects can be accomplished with video
filters in software.

Rich




More information about the MPlayer-users mailing list