- New Proof-of-Work scheme, Autolykos – a variant of Equihash with better ASIC-resistance
How does it have better ASIC resistance, when Equihash is claimed to have superlinear time-memory tradeoffs (TMTO)
while Autolykos can be solved with no memory and constant slowdown?
IMHO it's closer to Ethash than to Equihash.
Autolykos is pretty close to Equihash, it is based on k-sum problem , but also adds a signature component for somewhat pool-resistance. Parameters are chosen in order to be much more memory-intensive (and memory-bandwidth intensive) than ZCash's and other popular Equihash coins PoWs.
Autolykos is very different from Equihash in memory requirements.
Equihash searches for an arbitrary size 2^k subset of indices,
whose corresponding items must xor to 0 (as well as satisfy
other conditions on the ordered binary tree of indices and items).
Memory is needed to sort and combine partial solutions.
Without memory, finding a solution would take longer than the age of the universe.
Autolykos instead generates the possible subsets as genIndexes(m||nonce),
as shown in Algorithm 1 of https://docs.ergoplatform.com/ErgoPow.pdf
The element at an index j is trivially computed as H(j||M||pk||m||w)
No memory at all is required for that.
It could be used to cache the hash computation of elements, but that is at best
a constant factor speedup.
Ah I see. We're discussing it, see https://github.com/ergoplatform/ergo/issues/1100 (are you BigEvilSoloMiner aka AsicResearcher maybe though? )
Yeah, that's definitely very helpful and useful!
I totally agree with you
When Ergo pump
