Next: 6 Source Code Presentation
Up: Implementation and Comparison of
Previous: 4 Results
Overall, the radix4 implementation executes faster on the platforms
tested than the author's radix2 algorithm. The predicted number of
additions and multiplications as reported in section 2
shows that radix2, as it has been implemented here, requires far
fewer multiplications than radix4. The difference between the number
of multiplications required for the author's implementation of radix2
and radix4, for a given sequence length, N, is found in equation
(15):
 
(8) 
Additionally, the overall computation time is unsurprisingly
exponential as a function of sequence length, N; evidenced by the
linear traces on the logarithmic plot. Shown in figure
1 are comparison curves for the radix2 and radix4
algorithm implementations, plotted on a loglog plot, displaying
computation time on the vertical axis and sequence length on the
horizontal axis.
Figure 1:
Comparison of radix2 and radix4 FFT algorithms across 4 platforms.

Next: 6 Source Code Presentation
Up: Implementation and Comparison of
Previous: 4 Results
Mike Andrews
6/29/1998