On Tue, Jun 16, 2009 at 1:13 PM, Kostya <kostya.shish...@gmail.com> wrote:
>
> On Tue, Jun 16, 2009 at 12:38:51PM -0400, Alex Converse wrote:
> > We discussed this once without ever reaching a consensus. I'm going to bring
> > it up again.
> >
> > if(vec[k] == 64.0f){//FIXME: slow
> > This line is verifies an escape
> > if(t >= 165140.0f*IQ){ // clipped value
> > This line is a shortcut to determine if we have a clipped value
> > di = t - 165140.0f;
> > This line attempts to inverse quantize the coefficient and subtract it from
> > the actual coefficient. Using the equation t - value^(4/3) where value =8191
> > curbits += 21;
> > }else{
> > int c = av_clip(quant(t, Q), 0, 8191);
> > This line quantizes the coefficient
> > di = t - c*cbrt(c)*IQ;
> > This line inverse quantizes the coefficient and subtracts it from the actual
> > coefficient. Using the equation t - value^(4/3) * IQ
> > curbits += av_log2(c)*2 - 4 + 1;
> > }
> >
> > The two di lines do not use the same equation.
>
> Patch ok. I forgot that with very small quantizer ven low values may hit
> that wall.
Applied
_______________________________________________
FFmpeg-soc mailing list
FFmpeg-soc@mplayerhq.hu
https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-soc
