Introduction

benchFFT is program to benchmark FFT software, written by Matteo Frigo and Steven G. Johnson at MIT. The benchmark incorporates a large number of publicly-available FFT implementations, both in C and in Fortran, and measures their performance and accuracy over a wide range of transform sizes. The latest version of benchFFT, along with benchmark results on a variety of machines and compilers, can be found at the benchFFT home page:

http://theory.lcs.mit.edu/~benchfft

This manual describes the usage and methodology of the benchmark, and also explains the process by which new FFTs may be added. (If you do add a new FFT, or have results on a new machine, please let us know.)

Contents

The documentation is organized into the following sections:

Feedback

If you have comments, questions, or suggestions regarding this software, don't hesitate to email us at benchfft@theory.lcs.mit.edu.