I have been reading about reducing SHA-256 to a boolean function (see
https://bitcointalk.org/index.php?topic=55888.0). The outcome seems to be that this is not very efficient compared to straight forward computation and therefore it's not a worthwhile idea.
However isnt the object of mining to produce a hashed result less than a specific value?
So wouldn't it be useful to calculate probable nonce ranges based on the other input. - rather than just rerunning the algorithm with an incrementing integer. Even if nothing could be predicted for the first few iterations- wouldn't it be useful to use the output to develop a more directed guess for later nonce values.
Maybe I'm missing something here?