FFT Garden is a collection of Fast Fourier Transform algorithms implemented in the C programming language.

The project aims to provide easy to read and easy to study code for various FFT algorithms. Most of the proposed algorithms are described in the literature.

There are many designs choice that can be made when designing an FFT algorithm. When combined, this yield a very large number of variants. The collection of 44 algorithms provided here cover a small portion of the FFT design space. The source files naming reflect the implemented algorithm variants as described below.

Let's consider three aspects of FFT algorithms:

• The butterfly design:

  • Cooley-Tukey Radix-2 FFT [ct]
  • Mixed Radix 2/4 FFT [mr]
  • Split-Radix FFT [sr]
  • Conjugate Pair FFT [cp]

• The decimation:

  • Decimation In Time [dit]
  • Decimation In Frequency [dif]

• The traversal and algorithm structure:

  • Breadth-First Iterative [bi]
  • Depth-First Recursive [dr]
  • Depth-First Iterative [di]

More variants can be devised by considering the following additional properties:

• Butterflies using the same twiddles can be grouped by stage [G].
• Butterflies using the same twiddles can be grouped by block [S].
• Input reordering and/or butterfly scheduling may rely on lookup tables [L].
• Butterfly functions may be simplified for some twiddle factor values [P].
• The size of the twiddle table may be reduced to N/8 entries [F].
• Some algorithms can be made to work in-place [N].

Many more algorithms could be added. In particular:

• Only power of 2 size algorithms are proposed.
• Most proposed algorithms use decimation in time.
• Most proposed algorithms are out-of-place.
• Other butterfly designs could be included.

Pull requests are welcome.

The project source code is licensed under the CC0 creative commons license.

• Alexandre Becoulet - Initial work as part of the paper A Depth-First Iterative Algorithm for the Conjugate Pair Fast Fourier Transform