params: ["agent", null]
result: [null, "nonce1", "nonce2 size"]
nonce_1 is first part of the block header nonce in hex.
request:
{
"id": 1,
"method": "mining.subscribe",
"params": ["kdaminer-v1.0.0", null]
}
response:
{
"id": 1,
"result": [null, "012345", 5],
"error": null
}
params: ["username", "password"]
result: true
request:
{
"id": 2,
"method": "mining.authorize",
"params": ["900703b6dd2493696068af72957a94129e54e85f269becc665672bf4730fc6a3", "x"]
}
response:
{
"id": 2,
"result": true,
"error": null
}
params: ["32 bytes target in big endian hex"]
{
"id": null,
"method": "mining.set_target",
"params": ["0001000000000000000000000000000000000000000000000000000000000000"]
}
params: ["jobId", "header", cleanJob]
{
"id": null,
"method": "mining.notify",
"params": [
"1234",
"286 bytes header in hex",
true
]
}
https://github.com/kadena-io/chainweb-node/wiki/Block-Header-Binary-Encoding
Size Bytes Value
8 0-7 nonce
8 8-15 time
32 16-47 parent
110 48-157 adjacents
32 158-189 target
32 190-221 payload
4 222-225 chain
32 226-257 weight
8 258-265 height
4 266-269 version
8 270-277 epoch start
8 278-285 flags
old version
params: ["username.worker", "jobId", "nonce2"]
result: true / false
request:
{
"id": 102,
"method": "mining.submit",
"params": [
"900703b6dd2493696068af72957a94129e54e85f269becc665672bf4730fc6a3.worker1",
"1234",
"6789abcdef"
]
}
===========================
GPU Mining Speedup Announcement
===========================
tl;dr New GPU miners are now available that mine 5x faster than before.
Hi Everyone,
We would like to announce a new development that is of likely interest to the Kadena mining community. A new version of NoncerPro is now available [1] that can mine more than five times faster than the previous version. A similar improvement is available for the bigolchungus miner as well [2].
For those wondering about what made this speedup possible, the last 64 bits of the block header had been reserved for future protocol changes. It was discovered that miners could use these bits as the nonce. Since these bits are at the end of the block header this makes it possible to mine significantly more efficiently. We’re announcing this to make the information available as broadly as possible to encourage a fair and decentralized mining ecosystem. We expect that other mining software, pools, etc will also come out with improvements to take advantage of this development.
If you are interested in getting more detailed information about this please reach out to us and we’ll be happy to assist.
[1] https://github.com/NoncerPro/Kadena/releases/tag/2.1.1
[2] https://github.com/kadena-community/bigolchungus/pull/20
new version
# new mining submit
params: ["username.worker", "jobId", "nonce2", "extraNonce(hex encoded `flags` field in the header)"]
result: true / false
for example:
{
"id": 102,
"method": "mining.submit",
"params": [
"900703b6dd2493696068af72957a94129e54e85f269becc665672bf4730fc6a3.worker1",
"1234",
"6789abcdef",
"0123456789abcdef"
]
}
response:
accepted share response:
{
"id": 102,
"result": true,
"error": null
}
rejected share response:
{
"id": 102,
"result": false,
"error": [21, "low difficulty", null]
}
in this example:
nonce = nonce1 + nonce2 = 0123456789abcdef (hex in big endian)
we should do `reverse_bytes(hex_decode(nonce))` before writing this to the 286 header
There are now ASICs available that can compute 100-200TH/s. Those devices exhaust 6 bytes of nonce2 in less than 2-3s. (6 bytes of nonce2 corresponds to about 280TH.)
These devices must be provided with a sufficiently large nonce2 space. 7 bytes would be sufficient for almost all blocks. But that leaves only 1 byte for nonce1, which may not be enough for some pool applications. Another solution is to provide the mining device with new work before it has used up all nonces.
A stream of different work items can be produced from a single work header by increasing the timestamp of the work header according to the rate of the stream. The timestamp uses microsecond resolution. Therefore the stream could provide a device with the full nonce2 range for each microsecond. For a nonce2 of size 6, this would be about 280TH/microsecond or 280,000PH/s from a single work header. This number is increased further because chainweb-node concurrently provides headers for all available chains.
The timestamp is stored in the work header at offset 8 as a 64bit twoth complement encoded value in little endian byte order. It counts SI microseconds since POSIX epoch (leap seconds are ignored). Details about the work header format are available at this wiki page.