[Ffmpeg-devel] Cedocida DV-Codec

[Michael Niedermayer]

Hi

On Sat, Aug 27, 2005 at 01:33:47AM -0400, Rich Felker wrote:
> On Fri, Aug 26, 2005 at 10:58:13PM -0500, Mark Rages wrote:
> > On 8/26/05, Steven M. Schultz <sms at 2bsd.com> wrote:
> > > 
> > > On Fri, 26 Aug 2005, Mark Rages wrote:
> > > 
> > > > The ffmpeg DV encoder is too slow for my purpose (I want to decode,
> > > > process and re-encode in real time).  From a quick look through the
> > > 
> > >         Need a faster system :)
> > > 
> > >         I see 147 frames/sec decoding with ffmpeg (2GHz G5 PPC) - that should
> > >         leave plenty of time for encoding.  Haven't fired up the Opteron system
> > >         to see what it'd do.
> > > 
> > 
> > Decoding is fast, it's encoding that is slow.  I'm on an "AMD
> > Athlon(tm) 64 Processor 3000+" system.  Decoding takes no time at all.
> > 
> > I don't know enough assembly to know if Cedocida is AMD64 friendly or
> > not.  That's why I'm asking here. Obviously, AMD64 support would be
> > rather important to me...
> 
> I really doubt this other implementation is worth bothering with,
> except maybe to lift a few optimized functions. Porting external code
> into lavc results in lots of bloat and code duplication, since the
> toolkit of dsp functions already in lavc won't get used. I expect the
> lavc encoder just needs some optimization, and then it will be
> perfectly fine on its own.

agree, i looked at the code briefly, it looks clean but it is full of
global non constant arrays and does colorspace conversation within the 
codec, the asm code i looked at also is not optimal at all

[...]
-- 
Michael