Wed, 18 Jun 1902 08:00:00 GMT

Shell sort vs. Comb sort (was: sorting arrays)

I've done some tests comparing the Shell sort and Comb sort
routines recently uploaded here.  As Martin suggested, I tested
soring arrays of strings.  The arrays were uniformly distributed
string[15] of upper case letters.  I tested arrays of various
sizes from 100 to about 2,000,000 elements.

Shell sort was 9% to 26% faster than comb sort.  For both
methods, the average running time is in very good agreement with
c*n^p, where n is the number of elements and c is a constant
depending on the speed of the machine, the implemention, the
compiler, etc.  For Shell, p is about 1.12 - for comb is it
about 1.138.  The values of c for the two methods is almost
identical, but Shell was about 0.3% better.

Jud McCranie