chia-plotter (pipelined multi-threaded)
This is a new implementation of a chia plotter which is desinged as a processing pipeline, similar to how GPUs work, only the "cores" are normal software CPU threads.
As a result this plotter is able to fully max out any storage device's bandwidth, simply by increasing the number of "cores", ie. threads.
Usage
chia_plot <pool_key> <farmer_key> [tmp_dir] [tmp_dir2] [num_threads] [log_num_buckets]
For <pool_key> and <farmer_key> see output of `chia keys show`.
<tmp_dir> needs about 200G space, it will handle about 25% of all writes. (Examples: './', '/mnt/tmp/')
<tmp_dir2> needs about 110G space and ideally is a RAM drive, it will handle about 75% of all writes.
If <tmp_dir> is not specified it defaults to current directory.
If <tmp_dir2> is not specified it defaults to <tmp_dir>.
[num_threads] defaults to 4, it's recommended to use number of physical cores.
[log_num_buckets] defaults to 7 (2^7 = 128)
Make sure to crank up <num_threads> if you have plenty of cores, the default is 4.
Depending on the phase more threads will be launched, the setting is just a multiplier.
RAM usage depends on <num_threads> and <log_num_buckets>.
With default <log_num_buckets> and 4 threads it's ~2GB total, with 16 threads it's ~6GB total.
How to Support
XCH: xch1w5c2vv5ak08pczeph7tp5xmkl5762pdf3pyjkg9z4ks4ed55j3psgay0zh
I developed this on my own time, even though I already filled all my HDDs (~50 TiB) with the official (slow) plotter.
Results
On a dual Xeon(R) E5-2650v2@2.60GHz R720 with 256GB RAM and a 3x800GB SATA SSD RAID0, using a 110G tmpfs for <tmp_dir2>:
Number of Threads: 16
Number of Sort Buckets: 2^7 (128)
Working Directory: ./
Working Directory 2: ./ram/
[P1] Table 1 took 21.0467 sec
[P1] Table 2 took 152.6 sec, found 4295044959 matches
[P1] Lost 77279 matches due to 32-bit overflow.
[P1] Table 3 took 181.169 sec, found 4295030463 matches
[P1] Lost 62514 matches due to 32-bit overflow.
[P1] Table 4 took 223.303 sec, found 4295044715 matches
[P1] Lost 76928 matches due to 32-bit overflow.
[P1] Table 5 took 232.129 sec, found 4294967739 matches
[P1] Lost 235 matches due to 32-bit overflow.
[P1] Table 6 took 221.468 sec, found 4294932892 matches
[P1] Table 7 took 182.597 sec, found 4294838936 matches
Phase 1 took 1214.37 sec
[P2] max_table_size = 4295044959
[P2] Table 7 scan took 16.9198 sec
[P2] Table 7 rewrite took 44.796 sec, dropped 0 entries (0 %)
[P2] Table 6 scan took 47.5287 sec
[P2] Table 6 rewrite took 81.2195 sec, dropped 581301544 entries (13.5346 %)
[P2] Table 5 scan took 46.6094 sec
[P2] Table 5 rewrite took 77.9914 sec, dropped 761979000 entries (17.7412 %)
[P2] Table 4 scan took 52.427 sec
[P2] Table 4 rewrite took 75.7487 sec, dropped 828872625 entries (19.2983 %)
[P2] Table 3 scan took 54.0839 sec
[P2] Table 3 rewrite took 74.9016 sec, dropped 855088153 entries (19.9088 %)
[P2] Table 2 scan took 49.692 sec
[P2] Table 2 rewrite took 73.0273 sec, dropped 865610902 entries (20.1537 %)
Phase 2 took 721.638 sec
Wrote plot header with 268 bytes
[P3-1] Table 2 took 76.0894 sec, wrote 3429434057 right entries
[P3-2] Table 2 took 75.1076 sec, wrote 3429434057 left entries, 3429434057 final
[P3-1] Table 3 took 78.0162 sec, wrote 3439942310 right entries
[P3-2] Table 3 took 73.0284 sec, wrote 3439942310 left entries, 3439942310 final
[P3-1] Table 4 took 133.769 sec, wrote 3466172090 right entries
[P3-2] Table 4 took 76.1504 sec, wrote 3466172090 left entries, 3466172090 final
[P3-1] Table 5 took 127.125 sec, wrote 3532988739 right entries
[P3-2] Table 5 took 77.7182 sec, wrote 3532988739 left entries, 3532988739 final
[P3-1] Table 6 took 134.779 sec, wrote 3713631348 right entries
[P3-2] Table 6 took 81.9068 sec, wrote 3713631348 left entries, 3713631348 final
[P3-1] Table 7 took 69.066 sec, wrote 4294838936 right entries
[P3-2] Table 7 took 94.0157 sec, wrote 4294838936 left entries, 4294838936 final
Phase 3 took 1104.11 sec, wrote 21877007480 entries to final plot
[P4] Starting to write C1 and C3 tables
[P4] Finished writing C1 and C3 tables
[P4] Writing C2 table
[P4] Finished writing C2 table
Phase 4 took 89.0748 sec, final plot size is 108834390977 bytes
Total plot creation time was 3129.28 sec
How to Verify
To make sure the plots are valid you can use the ProofOfSpace tool from chiapos:
ProofOfSpace check -f plot-k32-???.plot [num_iterations]
Future Plans
I do have some history with GPU mining, back in 2014 I was the first to open source a XPM GPU miner, which was about 40x more efficient than the CPU miner. See my other repos.
As such, it's only a matter of time until I'll add OpenCL support to speed up the plotter even more, keeping most of the load off the CPUs.
Dependencies
- cmake (>=3.14)
- libgmp3-dev
- libsodium-dev
Install
git submodule update --init
./make_devel.sh
The binaries will end up in build/, you can copy them elsewhere freely (on the same machine, or similar OS).
Known Issues
- Doesn't compile with gcc-11, use a lower version.
- Needs at least cmake 3.14 (because of bls-signatures)