Compute the N dimensional FFT of a using subroutines from Fftw. The optional vector argument size may be used specify the dimensions of the array to be used. If an element of size is smaller than the corresponding dimension, then the dimension is truncated prior to performing the FFT. Otherwise if an element of size is larger than the corresponding dimension a is resized and padded with zeros.