Table B-1 provides benchmark numbers for 18 common DSP programs implemented on
the 27-MHz DSP56001.
The four code examples shown in Figures B-1 to B-4 represent the benchmark programs
shown in Table B-1.
Figure B-1 is the code for the 20-tap FIR filter shown in Table B-1. Figure B-2 is the code
for an FFT using a triple nested DO LOOP. Although this code is easier to understand
and very compact, it is not as fast as the code used for the benchmarks shown in Table
B-1, which are highly optimized using the symmetry of the FFT and the parallelism of the
DSP. Figure B-3 is the code for the 8-pole cascaded canonic biquad IIR filter, which uses
four coefficients (see Table B-1). Figure B-4 is the code for a 2N delayed least mean
square (LMS) FIR adaptive filter, which is useful for echo cancelation and other adaptive
filtering applications.The code example shown in Figure B-5 represents the Real FFT
code for the DSP56002, based on the Glenn Bergland algorithm.
The code for these and other programs is free and available through the Dr. BuB elec-
tronic bulletin board.
B - 3