Flush to disk more consistently by accounting memory usage of serials/anchors in cache.
Closes#626.
It's important that this at least *approximates* the memory usage, so that we flush the cache to disk as expected. It's okay that we overestimate. The serials are stored in keys in the `boost::unordered_map`, so we can simply use that map's `DynamicMemoryUsage`. The anchors are another story.
Only compare the first n/(k+1) bits when sorting
We only need to sort based on the bits we are colliding. In earlier solver rounds, this speeds up the comparison considerably (calling `memcmp()` on 3 bytes instead of 12 in the first round for the current parameters).
Randomise the nonce in the block header
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
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
Zcash address encoding
We need to encode Zcash addresses so they aren't as large and unweildy. We're using Base58Check just like upstream does, and to ensure the first character is "z" in our addresses we must use two bytes for the version string. Two bytes gives us an extra character for free, so this PR targets the beginning of addresses to have "zc".
```
$ ./src/zcash-cli zcrawkeygen
{
"zcaddress" : "tnvaj4ZbZG83tj4RwZcFeLgJoSt8nw1ZvSCG8EMyowAsXTQgJPat77Y43BVdVCrwrbLy7GG9msJDYdn5hmreHmkXAkX17hb",
"zcsecretkey" : "SKzkxCRWvscKnroSFyhCqhY332KcDMH4LLNdK2TsSvbmr3CGAB8B",
"zcviewingkey" : "10aa74046f31cbe5eaa8965d1e104853234c3d6c6e45f9c497ca3a025d159755"
}
```
This PR also encodes the spending keys with a prefix that targets "SK". The spec needs to be updated with these changes.
Testnet addresses will start with "tn".
Closes#572
Final changes to the circuit
* Remove [redundant](https://github.com/scipr-lab/libsnark/issues/37) bitness constraints that I pointed out in #908.
* The depth is increased from 20 to 29. I chose 29 because the QAP degree ends up being a power of two, which is probably going to be useful for the FFT in our MPC. (Closes#16)
If we're happy with depth 29 (over 268 million possible joinsplits), this will probably be the final change to the constraint system before 1.0.
zkbot
Sean Bowe
Daira Hopwood
Jack Grigg
Simon
aniemerg
Taylor Hornby