It's pretty easy to write a simple radix-2 FFT for power-of-two sizes (e.g. the Wikipedia article on the Cooley-Tukey FFT has pseudocode). For arbitrary precision arithmetic, there's not too much point in much fancier algorithms.
(FFTW's benchmark/testing code actually contains an arbitrary-precision FFT that we use for accuracy testing.) I've been playing with a framework in Julia for adding FFT algorithms, and my plan is to include arbitrary-precision support. On Wednesday, March 5, 2014 2:19:36 PM UTC-5, John Travers wrote: > > Does anyone have any suggestions of code or sources of inspiration for > arbitrary precision FFT routines? I was thinking of a type generic FFT a > bit like the current julia quadrature routine in base can be used with any > precision. > > I might try to implement a simple generic FFT code (obviously it will be > much slower than fftw) and was wondering if anyone has any > suggestions/experience with this? > >
