There is a slick way to do this when your vector is the N elements [0 1 2 ... N-1]:
N = 3;
m = dec2base(0:N^N-1, N)-'0';
Notice that dec2base gives a string result, and "subtracting" '0' from that string gives the numeric result you want.
For N=8, this takes about 5 seconds to run on my machine. (I think larger N than that is impractical from a memory point of view.)
If your vector is not so super-specialized, but the elements are unique, then I think this method would still be useful. You could generate the above first, then do a substitution to get to the elements you actually want.
