This version works, but generates the initial rows in a way that is not what we
want to specify. See #1175 for resolving this.
Co-author: Daira Hopwood <daira@jacaranda.org>
When the difficulty adjustment algorithm was altered, the special testnet
min-difficulty case was maintained, but the difficulty adjustment for the
following block then adjusted from min-difficulty instead of from the last
non-min-difficulty block. This caused the difficulty on the testnet to sawtooth
instead of stabilising. The intended behaviour is restored here.
The top and bottom 16 bits of the nonce are left clear for local use as thread
flags and counters. This does not leak any more local information about the
miner than is currently exposed.
The cleared bits should not be considered a consensus rule, as miners are free
to set all bits of the nonce however they wish.
Closes#1033
Includes a tweak to set the fixed-width of the rows based on whether the first
or last round is widest. This is necessary for some parameters, these ones
included.
This field has no defined semantics. While it was added as a result of
discussions about merged mining in #724, this field will not necessarily ever be
used for that purpose.
This reduces the peak number of lists in-memory from 2^k to k, and enables the
solver to eliminate most duplicates before it has instantiated the full set of
leaves.
After a new block is found or after a few nonces have been tried (currently
after every nonce), the miner checks for global changes. If any of these are
triggered, a new block is built from scratch, which re-initialises the Equihash
input. But if none of the checks are triggered, the miner just updates nTime and
continues mining - without updating the Equihash input to account for the new
block header. This bugfix corrects the behaviour by regenerating the Equihash
input in both situations.
Changing the order of difficulty calculation operations to divide first doesn't
affect the result significantly, but ensures we never overflow the arith_uint256
during multiplication and get an artificial jump in difficulty.
Changes to bloom tests were done by running the following commands:
sed -i 's/\(CDataStream stream(ParseHex(".\{152\}\)\(.\{8\}\)/\100000000000000000000000000000000000000000000000000000000\200/' src/test/bloom_tests.cpp
sed -i 's/\(vector<unsigned char> vch = ParseHex(".\{152\}\)\(.\{8\}\)/\100000000000000000000000000000000000000000000000000000000\200/' src/test/bloom_tests.cpp
and then reverting the single change to the transaction line.
The main and test networks are configured to use parameters that are currently
low-memory but usable with the basic solver; they will be increased once the
solver is optimised. The regtest network is configured to have extremely low
memory usage for speed.
Note that Bitcoin's double-hasher is used for the difficulty check. This does
not match the paper, but is simpler than changing the block header
serialization. Single hashing is kept for the EquiHash solver because there is
no requirement on execution time there, only on memory usage.
Jack Grigg