Quote:
Originally Posted by TheCount
OK, so the "n per 25000 at 1e12" value in the Unproven Conjectures table is derived by calling:
> srsieve n 100001 N 110000 P 1e6 "k*b^n+/1"
For each k, multiplying the results by 1.25, and adding them all together.
The point is that P=1e6 gives a more accurate weight than P=511, so I will use that from now on.

To clarify, he runs srsieve only one time for each base, which sieves all k's at once usually in just a few seconds. There is no multiple runs for all k's followed by adding them all together. Actually, srsieve gets run once only for bases that have changed since the last time the page was updated. It's very efficient.