On Sat, Jul 16, 2011 at 09:43:56AM -0700, Ronald S. Bultje wrote:
> Hi,
> 
> On Sat, Jul 16, 2011 at 9:33 AM, Kostya <[email protected]> wrote:
> > On Sat, Jul 16, 2011 at 09:25:52AM -0700, Ronald S. Bultje wrote:
> >> On Sat, Jul 16, 2011 at 8:33 AM, Kostya <[email protected]> wrote:
> >> > On Sat, Jul 16, 2011 at 08:24:39AM -0700, Ronald S. Bultje wrote:
> >> >> On Sat, Jul 16, 2011 at 3:05 AM, Kostya <[email protected]> 
> >> >> wrote:
> >> >> > On Sat, Jul 16, 2011 at 06:53:43AM +0200, Kostya wrote:
> >> >> >> Since it would take Diego too much time to finish his work, here's 
> >> >> >> working
> >> >> >> Flash Screen Video 2 decoder which decodes frames properly.
> >> >> >
> >> >> > Here's a version updated for the latest Git revision.
> >> >> [..]
> >> >> > +static int decode_hybrid(const uint8_t *sptr, uint8_t *dptr, int dx, 
> >> >> > int dy,
> >> >> > +                         int h, int w, int stride, const uint32_t 
> >> >> > *pal)
> >> >>
> >> >> One thing I don't like about this function is that it reads 15bpp data
> >> >> but outputs it as if it were 24bpp... Since the palette itself is also
> >> >> just 7bpp, I wonder if we shouldn't make the decoder output in a
> >> >> different bpp...
> >> >
> >> > Palette data is 24bpp, and paletted values are mixed along with 15bpp 
> >> > values
> >> > in the same block. And the next block may use 24bpp too, so it's not 
> >> > possible.
> >>
> >> Wishlist item for future: either use the swscale functions for
> >> conversion, or DSP'ize it. No need to do this now, of course.
> >
> > Not quite possible here either - 7-bit palette indices and 15-bit raw colour
> > values are mixed together.
> 
> the palette index can point to a 15bit palette, merge them and output
> the two together to sws/dsp to generate 24bpp?
> 
> You can't DSP'ize the palette LUT anyway, except maybe in altivec.

Palette is RGB24 by itself anyway. And since spec says "convert to 24-bit RGB"
we do it for hybrid blocks :).
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to