/mithril

Pure Rust Monero Miner

Primary LanguageRustGNU General Public License v3.0GPL-3.0

Build Status codecov

Github Sponsor Donate

Mithril

Mithril v0.10.0 is Cryptonight v8 ready.

Rust Monero Miner (pure Rust is the goal, but the project is not there yet, see help wanted section)

Current Status (2020-05-06)

RandomX is implemented, but still some more testing is needed. It is now possible to mine with the implementation. Optimisations are next.

Unique Selling Points

  • Auto-Tuning: Finds the optimal setup for your hardware itself with a bandit algorithm
  • Easy to compile: inherited from the great rust toolchain
  • cross platform: should run on every (x86) platform supported by Rust
  • [Fast: should be as fast as the C-implementations] (I want to say that, but I currently cannot prove it)

Roadmap

Feature Backlog:

  • API for statistic data (especially the hash-rate)
  • Simple statistic terminal UI
  • RandomX (compiled mode)
  • WebAssembler Support
  • integrate GPU Mining (AMD)

Future Feature Backlog

  • ARM support (Raspberry, Pine64)
  • integrate GPU Mining (NVIDIA)

DONE:

  • RandomX update (interpreter mode)
  • hard-fork v7 update
  • auto-optimisation via bandit algorithms
  • optimise cryptonight hashing function
  • implement software AES (current work)
  • measure hash performance
  • implement stratum protocol (for pooled mining support)
  • implement cryptonight hashing function

HowTo Compile, Configure and Run

You need the Rust nightly version to compile Mithril, since it uses inline assembler which is only available in the nightly version of Rust. The nightly version is best installed with rustup. Once you have the nightly version installed, type cargo build --release for an optimised binary. The binary can be found in the target/release/ folder.

Mithril expects a config.toml in the working directory. Copy the default_config.toml as config.toml to the Mithril working directory. You need at least configure your Monero address in the [pool] section for the reward and the num_threads depending on your machine (a good start is to use 2x number of your cores on your machine).

If you get a wrong instruction set kind of error you can try to disable hardware AES with the has_aes flag in the [hardware] section.

If you find any issues, please report them here: Mithril Issues

Auto-Tuning

Configuration

Auto-Tuning is enabled by default in Mithril. You can disable and configure auto-tuning in the config.toml:

[worker]
num_threads = 8
auto_tune = true
auto_tune_interval_minutes = 15
auto_tune_log = "./bandit.log"

If you set auto_tune to false, Mithril will honour your num_threads and will use the number of threads configured there. The other options are only relevant if you set auto_tune to true. The config auto_tune_interval_minutes controls, how often a new bandit arm will be drawn and a new thread count setup will be tried. I suggest picking a longer interval, to average-out some spikes on loads on the machine the miner is running on.

You can enable detailed logging by setting a path to a file in auto_tune_log. Each step in the bandit algorithm will be logged there. You can evaluate the performance of the bandit algorithm on your machine with the Bandit-Tools that have been created for exactly this purpose. You find them here: Bandit-Tools.

The current state of the bandit algorithm will always be saved to ~/.mithril/bandit_state.json. You can stop the miner and on the next startup it will continue the arm evaluation on the point were it stopped last.

Evaluation

As mentioned you can use the Bandit-Tools Web-App to evaluate the evaluation/exploitation phases of the bandit algorithm. You can upload the bandit_state.json and bandit.log file and get a visualisation of the information present in these two files.

If you want to share publicly your log and state files, please open a pull request on the bandit_data branch of this project. Discussing results should be done on Reddit: Reddit Post.

Hash-Rate Logging

Mithril has basic support for logging the hash rate of the miner (in order to tune it). Hash-Rate Logging has to be enabled in the [metric] section and is disabled in the default configuration:

enabled = false
resolution = 100 #determines how often a hash result is reported
sample_interval_seconds = 60
report_file = "/path/to/hash/report/file.csv"

The most important configuration option is report_file. You can configure an absolute path to a csv file where the hash rate is logged. Each sample_interval_seconds a new line with <unix-timestamp>;<#hashes since last sample> is appended to this file. You can calculate the average hash rate (for a given time interval) from this file with external tools (e.g. Google Drive).

The resolution option determines how often a hash count is measured internally. Every resolution hashes the result is published to a metric sub-thread in the program. Setting this to a low value will increase the overhead for measuring.

Supported Platforms

Mithril was tested on this Platform/architecture combinations so far:

  • macOS 10.13/x64
  • Windows/x64
  • Linux
    • CentOS 7 64bit

Please notify me, if you tested mithril on one other platform and it is running stable.

ARM support (Raspberry, Pine64) is a short term goal I am working on.

Help Wanted

The goal of this project is to build a pure Rust monero miner implementation. Currently the Skein and JH hash functions are used as FFI C-Bindings, because there is not Rust implementation available (or I have not found any). A pure Skein or JH Rust implentation would be very welcomed. Notify me if you did implement one of these hashing functions and I will gladly use them in mithril!

Support

If you want to donate directly to support further development, this is my Monero donation address:

48y3RCT5SzSS4jumHm9rRL91eWWzd6xcVGSCF1KUZGWYJ6npqwFxHee4xkLLNUqY4NjiswdJhxFALeRqzncHoToeJMg2bhL

or support the project on Github Sponsors:

Github Sponsor

or Paypal:

Donate