It's come to my attention that someone is cheating on ypool. That sucks, and I figure that one way to help combat that is to put a better miner in everyone else's hands. It's binary only, because I'll be you-know-whatted if I'm going to let someone else cheat, and it's too easy to just thwack the sieve until ypool applies a server-side fix.
It has a 10% dev fee that I'll reduce soon - 9.5% to me, 0.5% to jh00. Please benchmark carefully against your own binary build, however, because it's not CPU optimized. I'd love feedback. I believe it's about 2x faster than a default checkout/build of clintar's xptMiner port.
For those who care, clintar's new build is 64k-fair if I'm reading the source right (it sieves all six to 64k), so I've matched that definition of fairness in this release.
http://www.cs.cmu.edu/~dga/crypto/ric/Linux only because I'm still an idiot about windows.
Updated benchmarks:
AMD 64 core:
clintar: [03:00:00] 2ch/s: 42.3390 3ch/s: 0.9049 4ch/s: 0.0190 Shares total: 50 / 30
mine: [03:00:07] 2ch/s: 52.0408 3ch/s: 1.3137 4ch/s: 0.0349 Shares total: 92 / 90
Haswell
clintar: [03:00:00] 2ch/s: 8.5398 3ch/s: 0.1843 4ch/s: 0.0038 Shares total: 10 / 9
mine: [03:00:00] 2ch/s: 11.5648 3ch/s: 0.2871 4ch/s: 0.0057 Shares total: 15 / 14
sse4 machines look solid. This may mean I need to do an avx2 optimized build (I built clintar's with -march=native and mine's sse4), but the AVX2 crowd is still probably running MMC or something anyway.
Happy mining - let's find some record prime clusters!
sigh 
