Patch: fix multiple commutative operands

[Roman Zippel]

Hi,

Richard Henderson wrote:

> > I'm testing all possible combinations, so the number of iterations is
> > 2^n, I initialize n_swapped with zero to keep the test against it
> > simpler.
> 
> I think it would be clearer elsewhere in the code if
> we kept n_swapped in log2 form then, and changed its
> use in the permutation loop to 2<<n_swapped.

But only the 2^n value is needed?! Maybe a better name(*) and/or comment
would be better?

bye, Roman

(*) I'm open to suggestions. :)