[FFmpeg-devel] [RFC] AAC Encoder, now more optimal

[Kostya]

On Sun, Sep 07, 2008 at 01:11:18AM +0200, Michael Niedermayer wrote:
> On Sat, Sep 06, 2008 at 08:15:55PM +0300, Kostya wrote:
> > On Sat, Sep 06, 2008 at 06:17:38PM +0200, Michael Niedermayer wrote:
> > > On Sat, Sep 06, 2008 at 06:36:16PM +0300, Kostya wrote:
> > [choosing codebooks for coeffients encoding]
> > > > IMO it should encode losslessly at this stage, so no rate distortion.
> > > 
> > > How much quality do we loose from this?
> > 
> > none (yet) - 
> 
> You apparently misunderstood me, i meant 
> how much quality is lost by ignoring distortion when choosing
> codebooks? Though this doesnt matter if you fix it and make it use full RD
> i just thought you wont from your first coment above. But judging from
> your next comment below it seem you do plan to fix this so its all fine
> and iam somewhat confused now :)
 
just wait and see ;)
 
> > quantization is performed the same way for all codebooks
> > I still have to introduce codebook-specific quantization (and distortion from it).
> 
> It would be great if you did this before wasting more time with
> approximating it, because the exact solution may very well be faster than
> your approximation. pow() and cbrt() are rather slow and there is no need
> to search through the codebook bruteforce, but id like you to implement the
> bruteforce variant so we can verify that the optimized code does work
> correctly.
 
it was some thought barrier to me, will be implemented.

> [...]
> -- 
> Michael     GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
> 
> Asymptotically faster algorithms should always be preferred if you have
> asymptotical amounts of data