[Ffmpeg-devel] PATCH: Blackfin Accelerated CSC ( ColorSpace Converter )

[Marc Hoffman]

Robin Getz writes:
 > On Thu 26 Apr 2007 08:52, Marc Hoffman pondered:
 > > Marc Hoffman writes:
 > >  > Please review, and let me know how this is comming along?
 > >
 > > Couple of small changes, fall back support for handling no more L1
 > > available.  internal local lables changed to start with L or .L.
 > >
 > > Please review.
 > 
 > Just looking on the way back from China...
 > 
 > +static
 > +int bfin_yuv420_bgr555 
 > 
 > +static
 > +int bfin_yuv420_rgb555
 > 
 > +static
 > +int bfin_yuv420_rgb24
 > 
 > ....
 > 
 > Since these just seem to be called from one place, 
 > 
 > +SwsFunc ff_bfin_yuv2rgb_get_func_ptr (SwsContext *c)
 > +{
 > +    SwsFunc f = 0;
 > +
 > +    switch(c->dstFormat) {
 > +    case PIX_FMT_RGB555: f = bfin_yuv420_rgb555; break;
 > +    case PIX_FMT_BGR555: f = bfin_yuv420_bgr555; break;
 > +    case PIX_FMT_RGB565: f = bfin_yuv420_rgb565; break;
 > +    case PIX_FMT_BGR565: f = bfin_yuv420_bgr565; break;
 > +    case PIX_FMT_RGB24:  f = bfin_yuv420_rgb24;  break;
 > +    case PIX_FMT_BGR24:  f = bfin_yuv420_bgr24;  break;
 > +    }
 > +
 > 
 > Maybe it makes sense to declare them as "always inline" ? It might save a few 
 > cycles (5 for the jump, 5 for the return).

I don't think I understand what your saying and how this helps?

 > 
 > What are these for? I didn't see them used?
 > 
 > +#define mL1 .l1.text
 > +#define mL3 .text
 > +
 > +#define MEM mL1

Internal memory optimization for system tuning.

 > 
 > This is a style thing, that I normally don't care about - I am not sure what 
 > they ffmpeg folks think - C++ comments in assembly files is normally frowned 
 > upon...
 > 
 > +        p2 = [fp+20];    // out
 > +        r3 = [fp+20+4];  // dW


What should it look like then?

 > _______________________________________________
 > ffmpeg-devel mailing list
 > ffmpeg-devel at mplayerhq.hu
 > http://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel