About CPU PoW algorithm
This update is supposed to be about the details of official launch; and I am actually supposed to issue more details about the algo once I hear from catia. This exceptional is due to my past efforts wasted on the GPU miners which actually turned out to be not a simple "waste", as we are turning into CPU algo and looks like many people like this change.
I've sent PM to djm and catia to seek helps on cudaminer. Further discussion with catia carried on when I asked about the difficulty of implementing Nvidia miner in a situation with multiple multiplications of M7 hash functions. Since catia is an expert on GPU stuffs, he knows the hardness of getting GPU miner on a particular algo. His immediate reaction was multiple multiplications increase the difficulty, and even implementing division will be very bad on GPU. Beyond that he seems interested to such a PoW algo; with improvement on manipulating mathematical functions, GPU miner could be even harder. While this conversation is in progress, there is supposed to be the due of news today so I am here to update what's going on. I also hope this appears to be a public invitation to catia and hope we can work out this matter.
I can foresee this idea would be very interesting. Many past efforts clarify people are fond of CPU mining, an initial goal of LTC. Scrypt is memory hard and Scrypt-N is an enhanced version, but the coming of ASIC for LTC seems put forward a need of much more enhanced memory-hard algo in order to prevent ASIC; aka, "the main reason why memory hardness is important is to make the proof of work function resistant to specialized hardware". Primecoin is one of earlier tries through finding Cunningham chains of prime numbers which turns out to be "memory hard" too. The demand in memory has a common feature as a consequence of the hardness of solving the problem. BitShares PTS is based on a birthday problem in search for matching pairs of birthdays. A recent approach proposed by tromp utilizes the cuckoo cycle to solve the proof of work problem. While seeing the coming GPU miners for the prior two mentioned coins, it's quite probable of GPU for cuckoo-cycle algo once implemented. To a certain extent, the statement regarding the relationship between memory and ASIC is right, while this is incorrect if we deem GPU as ASIC too. The cryptomining has been evolved such that GPU is more like a kind of ASIC, noticing the GPU farm & cloud mining are all around. GPU mining dumps is more a serious issue for altcoins, compared to ASIC for BTC mining.
The fact lying in the difference between GPU and CPU should be pretty much utilized. One may notice the apparent incapability of GPU in some aspects compared to CPU after all it is not a "CPU". In my opinion, what we can get on CPU which is not possible for a GPU would be a possible solution to achieve a real CPU coin. Memory hardness puts forward to hardware requirement for ASIC; addressing the basic difference would be fatal to GPU implementation. Notice such a proposal remains to be confirmed; a secondary option is the combination of both.
I'm personally very keen to catia's idea. I will be second to the modification of the M7 algo to make it GPU resistant. Along with the Magi's coin distribution plan, we can see the potential of accomplishing most of minings through CPU. Magi may not be a fully CPU solution, but here going forward, we believe this is a beginning, based on the efforts undertaking.
It's great that you have more hands behind the idea, especially those such capable as catia. Good thing is that GPU resistance would not to stand too long, as soon as Magi switches to PoS it will be much easier.
Coin for the CPU-only masses implicating fair distribution - this will spur some interest.
