Comparison of Processing Times

Table 4 shows typical average CPU times per sequence, in seconds, used by the algorithms for different numbers of points. The tests were run on an HP Visualize B132L workstation. Clearly, the execution times depend on implementation. Also, the average values do not reflect the large variations in the times used by the iterative algorithms SJ87 and SS90. Some general conclusions can be drawn anyway.

 

Table 4: Typical processing times.

Algorihm	10 points	30 points	45 points
SJ87	0.75	54.78	659.25
SS90	0.58	18.85	81.74
HW89	0.09	0.35	0.66
RS91	0.03	0.21	0.58
IP97	0.10	0.30	0.56

 

The iterative algorithms SJ87 and SS90 are much slower than the non-iterative HW89, RS91 and IP97. The difference in the worst case running times is even more striking as the iterative algorithms, especially SJ87, occasionally converge very slowly. As the number of points $N$ grows, the running times of SJ87 and SS90 increase dramatically.

The processing time of RS91 does not depend on $v_{max}$, otherwise the time grows roughly linearly with $v_{max}$. No significant dependence on the order of points was experienced.