Hmmm... but how can something that does not add any value in any way be called something else than "useless"?
Come on ktimesG, that statement is windier than a sackful of farts lol.
Does selecting random subranges, then searching sequential in that subrange, add any value over starting at the beginning key of the entire range and searching until the last key of the range?
Of course it doesn't really matter HOW you pick the subranges to scan. This is the entire point. The sequential scan speed-up was probably understood wrong: it was about not skipping individual keys, and doing a full start-end subrange scan, not about what subrange you intend to scan. That is a totally personal preference that does not help or interfere with anything.
Since each worker needs to start from some point, obviously it does not matter from what point it starts from - all ranges are totally equal. So choosing sequential ranges or not is the same thing, since I don't see any way to speed up the initial setup, especially when dealing with tens / hundreds of parallel workers.
[quote author=WanderingPhilospher link=topic=1306983.msg65291187#msg65291187
Stats, data, tests, tell me that it is 99.99999999999% not gonna happen lol. Run the public data that Bram provided for his PoW keys. How many were back to back in sequential order? Stop beating this dead horse. It's a risk/reward. For the 69 puzzle (in a 68 bit range), you would find maybe 2 h160 matches with 67 matched bits, in sequential order. Pretty small chance that it would be the actual h160 I am looking for. A risk I would definitely take 7 times each day and 45 times on Tuesday.
[/quote]
It's also 88.88888...% not gonna happen no matter what key you choose next, either if it's the k + 1 or some other key in some other range. There is no disagreement here.