On Jan 27, 9:47 pm, "Eric I." <rubytrain... / gmail.com> wrote:
> I made a slight modification to Paolo's solution, so it avoids
> searching coin combinations that are permutations of others (i.e.,
> [25, 1] and [1, 25]).  It does this by only allowing the system to add
> coins at or beyond the last coin used in the combination being built.
> So, if the coins are [25, 10, 5, 1], and we're building from 25, 10,
> 10, 5 then the next coin be only a 5 or a 1.  It seems to speed
> Paolo's solution by about 30% on my system.

Cool, thanks!

(By the way, the other solution was 6 times slower than breadth-first,
not 3.  The difference arose because I ran one with battery power and
the other with wall power).

Paolo