NumPy-based implementation of Fast Fourier Transform using Intel (R) Math Kernel Library. Supports in-place and out-of-place, 1D and ND complex FFT on arrays of single and double precision with arbitrary memory layout, so long as array strides are multiples of its itemsize.