Skipping first block of SHA-256

[littleblackhat]

Since the nonce is part of the second block of the first (inner) call to SHA-256, it seems like an obvious optimization to only re-calculate the whole hash if new blocks or transactions come in. Is this practice currently being employed in mining software?

[BenTuras]

simplified:
nonce is part of data.
data leads to hash1 in call to runhash1.
hash1 leads to hash in call to runhash2. (runhash1 and runhash2 are the same method: runhash)
potential valid block if hash[7] is zero.

Code:

		runhash1(hash1, data, midstate);
		runhash2(hash, hash1, sha256_init_state);

how would you like to optimize this ?

[Schleicher]

Since the nonce is part of the second block of the first (inner) call to SHA-256, it seems like an obvious optimization to only re-calculate the whole hash if new blocks or transactions come in. Is this practice currently being employed in mining software?

Yes. The hash of the first part is the "midstate".

[littleblackhat]

Thanks. midstate was the key word to search for. I guess it was so obvious that it was (at least at one point) returned by the getwork function.