I can work on faster GPU miner. Wolf's is not that much faster imho. 1.4 vs 2.2mh is a little bad, but unlikely he has enough customers to force price out of profitability for people running OSS miner. Would be nice to get some help from community. Maybe somebody can run NV profiler and figure out where time is spent. I developed this on my laptop which cannot actually run it, so not sure where all the time goes. Obviously optimizing the kernel that takes the most time will buy us the most results.
Yes, we need the optimisation desperately. On average I am getting 1.2-1.3mh over a longer period of time (from initial 1.4mh) versus 3mh+ Wolf is getting (he has improved over 2.2mh and I am not sure if he is sharing his private miner to a few privileged). That is less than half of the hash rate of what Wolf's private miner can do. This is by no mean a small amount! With the current exchange rate of 0.000024, we are all losing a lot when mining with the public miner.