The simplest transforms to describe are those performed on sequences of complex data. Such data are stored as arrays of type complex. The result of a complex FFT is also a complex sequence of the same length and, for the simple interfaces, is written back to the original array. Where multiple (m, say), same-length sequences (of length n) of complex data are to be transformed, the sequences are held in a single complex array; in the simple interfaces the array will be of length m*n containing m end-to-end sequences and the results of the m FFTs are returned in the original array. Expert interfaces are provided which give: greater flexibility in the storage of the original data and results, user provided scaling, and whether results should be written to a separate array or not.
For the simple interfaces, a two dimensional array of complex data, with m rows and n columns is stored in the same order as a set of n sequences of length m (as described above). That is, column elements are stored contiguously and the first element of the next column follows the last element of the current column. In the expert interfaces, column elements may be separated by a fixed step length (increment) while row elements may be separated by a second increment; if the first increment is 1 and the second increment is m then we have the same storage as in the simple interface.