[ANN] [XMG] MAGI | CPU mining | mPoW | mPoS

Quote from: truckythin on September 05, 2014, 01:33:37 AM

joe, any news from cudaminer and official release day?

Unfortunately cudaminer is not working yet; I've spent a few days on this, my mac didn't work out well initially so I used an Amazon EC2 spot instance and spent sometime too on the driver, cuda stuff. After some tests on cudaminer with complainings, I turned to ccminer. My initial thinking is that all about the block header but I was wrong; actually the block header does lead to a very different implementation of M7 using cuda. This basically means a different cuda version of M7 to be done for XMG, although we are using the same M7 as XCN, that'll be like recalculation of all the stuffs. This looks like unlikely for me to accomplish within a short time. Somehow it has been nearly two weeks since 1st testnet, and my focus was totally on the miners instead of the PoS-II. I think from now on I should place my emphasis on other stuffs. In any cases, the standalone cpuminer is ready.

Realizing the actual difference of the M7 implemented in standard block chain compared to the minichain, I think it won't be necessary to keep the same form of M7 as in XCN. An interesting thing about M7 is the use of multiplication of hashes. The multiplication is done by using GMP which allows arbitrary precision, only limited by available memory of the machine. This makes me thinking of an interesting feature to add into XMG: we can repeat many times of multiplications which literally increase the memory needed; the memory relies on the repeating times. This repeat is not just to get a pointless dataset to 'eat' memory, instead these 'huge' numbers are hashed by sha256 as the final. One of the consequence is that we can make GPU miner hard to be implemented. So it will be like tending to be a CPU coin, but I don't see any unlikely of GPU miners in the future. Another consequence is it could be memory hard (that also means not energy efficient by mining with GPU, efficiency will depend on the multiplication times), so the CPU/GPU mining could have a higher efficiency ratio compared to the pure M7, though it remains to be confirmed. It will also be possible to use varying multiplication times with time/block height etc to make GPU even harder. I am open to suggestions.

Since a few days on the miners, I will need sometime back to the coin itself; it will be assured my next update will be about the official release plan.