> > This helped a lot! The 634 seconds reduced to 37 seconds, a 17-fold
> > speed improvement :-)
> great! Does it mean that the speed is proportional to the number of
> digits processed in one go, i.e. about 9 times faster on a 32 system and
> about 18 times faster on a 64 bit system? I guess there must be other
> trick as the printing time was negligible using gmp.
Yes, the time increases nearly exponentially with the number length, as
it is a loop of additions over the whole number for each non-zero bit.
So the speed improvement is noticeable only for large numbers.