Optimizing a parallel fast Fourier transform

Thesis (M. Eng.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2004.Includes bibliographical references (p. 40).Parallel computing, especially cluster computing has become more popular and more powerful in recent years. Star-P is a means of harnessing that power by eliminating the difficulties in parallelizing code and by providing the user with a familiar and intuitive interface. This paper presents methods to create a parallel FFT module for Star-P. We find that because calculating a parallel FFT is more communication-intensive than processor-intensive, clever planning and distribution of data is needed to achieve speed-up in a parallel environment.by Richard Hu.M.Eng