[MPlayer-dev-eng] -vc ffmpeg2 speedup trick

[Michael Niedermayer]

Hi

On Saturday 04 September 2004 04:56, Attila Kinali wrote:
> On Thu, Sep 02, 2004 at 10:13:39AM -0400, D Richard Felker III wrote:
> > On Thu, Sep 02, 2004 at 04:56:49AM +0200, Michael Niedermayer wrote:
> > > Hi
> > >
> > > try '-lavdopts idct=4'
> > >
> > > the reason why this is faster then the default idct is that the default
> > > one checks for zero coefficients, which are _very_ common to avoid
> > > useless calculations, but the dequantization used in mpeg2 messes with
> > > the least significant bit of coefficient 63, which causes it to be non
> > > zero very often ...
> >
> > ok, how about this: could we have an option to do coeff[63]&=~1; for
> > mpeg2, or something similar? mpeg2 always uses tiny gop sizes so the
> > accumulated error probably wouldn't be visible, and it could be a huge
> > performance boost.
>
> Althoug i'm not an codec expert, this sounds like something we should
> do. Especial on borderline hardware this would give some space to
> breath. And as high res MPEG2 streams (aka HDTV) seem to spread more and
> more, we might get such borderline hardware again.
-lavdopts fast 
this does completly skip the calculations needed for fliping coeff[63] LSB, it 
also does a few other tricks, like assuming YUV 4:2:0 and the default inter 
quant matrix

[...]
-- 
Michael
level[i]= get_vlc(); i+=get_vlc();  (violates patent EP0266049)
median(mv[y-1][x], mv[y][x-1], mv[y+1][x+1]); (violates patent #5,905,535)
buf[i]= qp - buf[i-1];    (violates patent #?)
for more examples, see http://mplayerhq.hu/~michael/patent.html
stop it, see http://petition.eurolinux.org & http://petition.ffii.org/eubsa/en