From this thread (1. post):
I don't have insights into this program, though. Only thing I know is that George Woltman had a lot to do with it. Louie (jjjjl) is pretty busy right now, so I'm not sure he has the time to help you besides providing the source...Originally posted by jjjjL
basically, i hacked up the GIMPS P-1 factoring code and made it do proth numbers. well, maybe.
[...]
you'll probably have to know a bit about P-1 factoring to test this correctly. if you feel you know a lot and have visual studio installed, email me and i'll send you the code for you to look though. it's a derivative of what's contained in emc.c in the GIMPS source located at http://www.mersenne.org/gimps/source23.zip
edit:
Just fiddled around a little...
Basically, you can (almost) use the program as is by just changing SoB.dat, results.txt and lowresults.txt.
The latter ones should contain already found factors which then can be excluded from factoring. The first file needs following layout:
<number of ks> <-- in your case: 1
<starting n>
<upper n bound>
k=3
<first n>
<difference to next n> <-- for example +624
<difference to next n>
<difference to next n>
...
Now to the "(almost)" part:
It seems like the program refuses to work on n's < 1M, as this would make no sense for this specific project. Plus, I had problems with a range around 2M, too. But maybe that is due to the fact that I made up a faked (arbitrary) SoB.dat.
The 3M range does indeed work with k=3.
So I think all you have to do is putting in a new SoB.dat file and getting rid of the low n value "protection".
From your forum:
Maybe it would be worth looking at the proth_sieve program Mikael Klasson and Paul Jobling wrote.We are through with sieving. There is no point sieving further because we reached the limit of effectiveness of sieving.
Since the begin of public sieving early this year, performance sped up by ~5000% (exactly, it's around 50x as fast!). And at that time, it already was 10x faster than NewPGen - for our purposes.
This is partly due to the multitude of k values we have, but even without it, the speed increase is most noticable.
Back then, Louie thought of maybe 10-20T as the optimal sieving depth IIRC.
Right now, we're just cross 100T and still have a lot of headroom - I guess 200-300T will be most effective - probably even more...