Fig. 22. Example of 67 * 12 multiplication using RP algorithm.

(the multiplier becoming zero), hence it is not necessary to keep a separate iteration count.

A system diagram based on the RP algorithm is given in Figure 23. The data part has roughly the same structure as the original basic algorithm. While we eliminate one DMgpa for the iteration count, we must have a separate DMgpa for both the multiplier and multiplicand, so the implementation is still a 2-DMgpa, 1- Bus implementation. The multiplicand-multiplier interchange subprocess has been included to speed up the algorithm slightly. With the change of algorithm the performance has improved over the original. The multiplication step is now a little longer, but the number of steps is less (on the average), assuming small numbers.

Variations can be played on the implementation of the RP algorithm, just as in

138