Implementing a single one-dimensional FFT (fast Fourier transform) with S samples on P processors

The ready acceptance of parallel computers is strongly dependent on the programming systems and subroutine libraries available. In the paper the authors discuss the challenge facing the writer of a subroutine library who wishes to produce a library routine which efficiently utilizes the hardware to solve a user defined task. In particular the authors discuss some of the considerations and techniques used in implementing a subroutine for performing a single one-dimensional FFT on an AMT DAP