[FFmpeg-cvslog] r18985 - trunk/libavcodec/lcldec.c

[Reimar Döffinger]

On Sat, May 30, 2009 at 01:20:35AM +0200, Aurelien Jacobs wrote:
> On Fri, May 29, 2009 at 10:15:38PM +0200, reimar wrote:
> > Author: reimar
> > Date: Fri May 29 22:15:38 2009
> > New Revision: 18985
> > 
> > Log:
> > Use av_clip_uint8 instead of equivalent but unoptimzed code,
> > 
> > Modified:
> >    trunk/libavcodec/lcldec.c
> > 
> > Modified: trunk/libavcodec/lcldec.c
> > ==============================================================================
> > --- trunk/libavcodec/lcldec.c	Fri May 29 15:46:46 2009	(r18984)
> > +++ trunk/libavcodec/lcldec.c	Fri May 29 22:15:38 2009	(r18985)
> > @@ -81,11 +81,7 @@ static inline unsigned char fix (int pix
> >      int tmp;
> >  
> >      tmp = (pix14 + 0x80000) >> 20;
> > -    if (tmp < 0)
> > -        return 0;
> > -    if (tmp > 255)
> > -        return 255;
> > -    return tmp;
> > +    return av_clip_uint8(tmp);
> >  }
> 
> Now the tmp variable looks pretty much useless...

I wish that was the only issue. The only thing the whole code does is to
very inefficiently in *.20 fixed-point do the YUV->RGB conversion which
shouldn't be done in the decoder in the first place.
And yes, the coefficients are within 1% of those in the second formula
in MPlayer's DOCS/tech/colorspaces.txt, so it is completely ordinary
YUV, unlike for some other formats.
Only issue is that it uses a packed format not supported by FFmpeg, so
it should probably be converted into planar YUV.