I asked this question on another forum but did not get any responses so I thought I'd try here.
I've been researching bitcoin mining for a while and I have a work-in-progress hypothesis that could potentially speed up the mining process. I am keen to seek some feedback on the viability of this approach.
I've identified a way that allows asynchronous pre-computation of a number of values that can be computed well in advance of knowing the hashPrevBlock. These are billions of pre-computed 32bit values that do not need to be re-calculated when the block height changes; i.e. they can be reused indefinitely. The only caveat being that ASIC miners would now rely on memory to fetch these values instead of having to calculate it for every new hashPrevBlock.
In my opinion, this saves a lot of energy and chip area as the ASIC miner no longer needs to calculate these values. These values can be calculated once in software and then 'fed' to memories that the ASIC miners have access to.
Any feedback on how this approach is flawed or a non-starter? I have read about memory hard PoW algorithms that are ASIC-resistant. Am I making Bitcoin mining ASIC-resistant by such an approach?
Surely there have to be benefits of being able to pre-compute gigabytes of data ahead of time that will simplify/speed up the Bitcoin mining?
