module FFTW3

Fast Fourier Transforms by using FFTW Ver.3.

Features

Sine / cosine transforms
User choice of optimization levels (i.e., FFTW_MEASURE etc in addition to FFTW_ESTIMATE).
Multi-threaded FFT3 support -- don't know whether it's really feasible.

Install FFTW Ver.3.
- NOTE: To activate the single-float transform, you have to install FFTW3 with the single-float compilation, in addition to the default double-float version. This can be done by configuring FFTW3 with the --enable-float option, and install it again. The single-float version will coexist with the double-float version. If you do not install the single-float version, FFT is always done with the double precision, which is not bad if you are not time- and memory-conscious.
Install NArray.
Then, install this library as follows (replace "version" with the actual version number):
```
% tar xvzf fftw3-version.tar.gz
% cd fftw3-version
% ruby extconf.rb
% make
% make site-install
```
Or
```
% make install
```
(If you are using Ruby 1.8, make install is the same make site-install.)

Complex FFT.

The 3rd, 4th,... arguments are optional.

ARGUMENTS

narray (NArray or NArray-compatible Array) : array to be transformed. If real, coerced to complex before transformation. If narray is single-presision and the sigle-precision version of FFTW3 is installed (before installing this module), this method does a single-precision transform. Otherwise, a double-presision transform is used.
dir (-1 or 1) : forward transform if -1; backward transform if 1.
optional 3rd, 4th,... arguments (Integer) : Specifies dimensions to apply FFT. For example, if 0, the first dimension is transformed (1D FFT); If -1, the last dimension is used (1D FFT); If 0,2,4, the first, third, and fifth dimensions are transformed (3D FFT); If entirely omitted, ALL DIMENSIONS ARE SUBJECT TO FFT, so 3D FFT is done with a 3D array.

RETURN VALUE