On the structure of efficient DFT algorithms

This paper examines the structure of the prime length discrete Fourier transform algorithms that are developed by Winograd's approach. It is shown that those algorithms have considerable structure, and this can be exploited to develop a straightforward design procedure which does not use the Chinese remainder theorem and which includes any allowed permutations. This structure also allows the design of real-data programs and the improvement of the data transfer properties of the prime factor algorithm.