Re: RFC -- improved Fourier transform

[jcupitt gmail com]

On 1 April 2013 22:53, John Denker <jsd av8n com> wrote:
>  ++ There is also a   zeropad(...)   function, which makes it easy
>   for folks to pad their data so that the number of rows is a
>   power of 2.  This lays a foundation for the future.  We should
>   not be training users to expect that our FFT will always require
>   powers of 2.

I've been using fftw in my code with success:

http://www.fftw.org/

It's a GPL fft library. It's very quick, very accurate and (relevant
here) supports arbitrary-sized transforms. This might be an easy way
to remove the power-of-two issue. It also supports real->complex and
complex->real half transforms, which might be interesting. You'd need
to add extra options for that, of course.

John