FFTW is a problem, since it is GPL (not LGPL as codec2). Nonfree licenses
are available for a fee from MIT. I'm not sure they'd license it for
distribution as LGPL, even for a fee, because that would pretty much ruin
their income venue as people could then grab FFTW from codec2 as LGPL.
8<8<
*Ronan Paixão*
2012/1/20 Brian Smith <brian@...>
> Rick van Rein wrote:
> > Hello,
> >
> > This would seem to be of interest to Codec2: a FasterthanFast
> > Fourier Transform designed especially for sparsely filled spectra:
> >
> > News article:
> > http://web.mit.edu/newsoffice/2012/fasterfouriertransforms0118.html
> >
> > Paper (free download, as publicfunded research ought to be):
> > http://arxiv.org/abs/1201.2501v1
> >
> > The algorithm achieves O(k.log(n)) if the input signal has at
> > most k nonzero Fourier coefficients. In terms of Codec2, we
> > could expect k=10 over 320 samples, so O(2663) becomes O(133).
> >
> > Roughly  and background noise would worsen these figures.
> > The Codec2 encoder spends about 1/3 of its time in FFT.
> >
> >
> > Cheers,
> > Rick
> >
> >
> An alternative would be to use FFTW instead of KissFFT, which runs
> roughly twice as fast, presumably by precomputing the required
> sine/cosine values used in the transform.
>
>
> 
> Keep Your Developer Skills Current with LearnDevNow!
> The most comprehensive online learning library for Microsoft developers
> is just $99.99! Visual Studio, SharePoint, SQL  plus HTML5, CSS3, MVC3,
> Metro Style Apps, more. Free future releases when you subscribe now!
> http://p.sf.net/sfu/learndevnowd2d
> _______________________________________________
> Freetelcodec2 mailing list
> Freetelcodec2@...
> https://lists.sourceforge.net/lists/listinfo/freetelcodec2
>
