Now I am glad I haven't invested any time in my own version of this yet.
Interesting that they chose to put the Fermat test on the GPU instead of the sieving.
Maybe I can learn something from their approach.
In our CoinShield miner it's actually the Fermat test that's more efficient on the CPU,
whereas the sieving was accelerated nicely by the GPU (higher memory bandwidth!)
Christian
Hi Christian,
On Cudaminer.... the hash rates on scrypt jane nf are lower on windows 8 by 33%. My vista computer gets about 3.6 mh/s and my 2 windows 8 computers get 2.3 mh/s. I tried everything I could even compiling Cudaminer on windows 8. All the combinations of the arguments ect. Do you are anybody here know why please. thanks
people still mine that? I took a look at the AMD kernel - looks like it could use some work.
My vista computer gets about 3.6 mh/s and my 2 windows 8 computers get 2.3 mh/s. I worked on it on and off for months. Still only 2.3
I think I found the problem. The source code is 32 bit and I can't compile it to x86.
ok just replace every uint32_t by uint64_t
try to retune cudaminer (however it is a known problem)
At this point, im ready to load vista on my asrock 81
now if you want the ultimate scrypt-jane high factor experience... install linux, it is well known the hashrate is much higher on linuex