So, 1 million is not an interesting orbit but 1 million minus one is, which is still pretty funny.

The logs are there : gPEAC_scans_1M.tbz if you need the raw data or build scripts. I stopped the scan at 1000615 for now, and I will continue slowly in the future but the returns diminish dramatically so the effort will be low on this front.

The histogram didn't change much, with the maximum still at 119 so there is some headroom for now.

The new sieve has 1132 entries, going up to 23251, which hints at a mostly linear progression : reaching 2K entries will require scanning up to about 2 million orbits, for example. But the selectivity increase is 7% only: the previous version

108628 sieve tests OK (21.29961%)

(with 500K orbits) now gives

101793 sieve tests OK (19.95941%)

Yet the scans show that the density is in the 7% ballpark around 1M, or 1/3  of the candidates that passed the sieve. Not many candidates give long orbits though. Well, it's a sieve, not a perfect filter.

You can get the new code there : gPEAC_1M.tbz
Enjoy.

Now I have to find why the sieve misses so many numbers.

Oh and 2147483648 % 7559=1984  so w31 is out. w30 still passes the sieve.