On Mon, 12 Mar 2001, Benjamin J. Tilly wrote:
> >proc{|k| 0.185*6.9893**(k**(1/3.0)) }
> >which can be used as a reliable approximation for large values of k
> >(add 2% for safety)
> Ah, I didn't know that.  Well with that fact you should be
> able to indeed do well.

I found that using GNUplot and with the intuition that the exponent was
k**(1/3.0) because there are three exponentials, 2, 3, 5.

> >The amount of RAM used is in O(k**(4/3.0)). I haven't calculated how much
> >yours take but it's much less...
> Is there anything you don't know about the Hamming function? :-)

Again, I found that while I was replying to your mail...

> I believe this, and I note that my solution, in order to give
> you the thousandth element, has to calculate the first thousand
> elements as well.  Therefore in both approaches calculating the
> thousandth element is essentially the same problem as calculating
> the first thousand, and since you are faster at that, you are
> simply faster. :-)

Can you adapt your solution to start at a higher value? Using the above
approximation, of course.

matju