Autolykos CUDA-miner

Prerequisites (Linux)

(For Ubuntu 16.04 or 18.04)

To compile you need the following:

1. CUDA Toolkit: see installation guide
2. CUDA Driver compatible with installed Toolkit: see compatibility table
3. libcurl library: to install run

$ apt install libcurl4-openssl-dev

4. OpenSSL 1.0.2 library: to install run

$ apt install libssl-dev

Install (Linux)

1. Change directory to autolykos/secp256k1
2. Run make

If make completed successfully there will appear an executable autolykos/secp256k1/auto.out and (if not already present) a config file autolykos/secp256k1/config.json with stub contents.

Test (Linux)

1. Change directory to autolykos/secp256k1
2. Run make test

If make completed successfully there will appear a test executable autolykos/secp256k1/test.out.

Install (Windows 64-bit)

1. Install compatible pair of MS Visual Studio C++ toolchain and CUDA toolkit compatibility table for latest CUDA toolkit
2. Build libcurl from sources with Visual Studio toolchain instruction
3. Download OpenSSL 1.0.2 installer from slproweb.com
4. Edit secp256k1/buildwin.cmd file, change OPENSSL_DIR, LIBCURL_DIR to your libcurl and OpenSSL directories, change CUDA_COMPUTE_ARCH to GPU code architecture you want
5. Find vcvars64.bat script, it should be in VISUAL_STUDIO_INSTALL_DIRECTORY\VC\Auxiliary\Build
6. Run cmd.exe, run vcvars64.bat script, then change dir to secp256k1, then run buildwin.cmd
7. If everything went good, miner.exe should appear in secp256k1 directory
8. If miner.exe can't find nvml.dll, add C:\Program Files\NVIDIA Corporation\NVSMI to your PATH environment variable before running.

Run (Linux)

• To run the miner you should pass a name of a configuration file [YOUR_CONFIG] as an optional argument
• If the filename is not specified, the miner will try to use autolykos/secp256k1/config.json as a config
• The configuration file must contain json string of the following structure:
  { "mnemonic" : "mnemonicstring", "node" : "https://127.0.0.1", "keepPrehash" : false }

If your seed mnemonic string is protected by password, add option "mnemonicPass": "yourpassword" to your configuration.

The mode of execution with keepPrehash option:

1. true -- enable total unfinalized prehashes array (5GiB) reusage. ( Should only be used if your CUDA devices have >= 8GiB memory)
2. false -- prehash recalculation for each block. (For CUDA devices with >= 3GiB memory)

To run the miner on all available CUDA devices type:

$ <YOUR_PATH>/autolykos/secp256k1/auto.out [YOUR_CONFIG]

To choose CUDA devices change and use runner.sh or directly change environment variable CUDA_VISIBLE_DEVICES

Run (Windows 64-bit)

• Create a config.json file in miner directory with following structure: { "mnemonic" : "mnemonicstring", "node" : "https://127.0.0.1", "keepPrehash" : false }

If your seed mnemonic string is protected by password, add option "mnemonicPass": "yourpassword" to your configuration.

The mode of execution with keepPrehash option:

1. true -- enable total unfinalized prehashes array (5GiB) reusage. ( Should only be used if your CUDA devices have >= 8GiB memory)
2. false -- prehash recalculation for each block. (For CUDA devices with >= 3GiB memory)

To change CUDA devices available to the miner change environment variable CUDA_VISIBLE_DEVICES , for example set CUDA_VISIBLE_DEVICES="0,1"

Mnemonic and public key generation

• Run auto.out -G

It will generate a BIP39 192 bit entropy mnemonic phrase and public key for you.

Testnet public node

Sometimes mining problems can be related to your Ergo node problems. We recommend to test your miner build with this public testnet node:

{ "mnemonic": "noise peasant subway frozen illegal pretty oak agent train valid wash title burst column yard decide move coin gas asset pretty hire happy fuel", "node": "http://159.203.36.162:9052", "keepPrehash": true }

HTTP Info

Miner has a HTTP info page located at http://miningnode:36207 (one can change default port by adding -DHTTPAPI_PORT XXXX to Makefile).

It outputs total hashrate, and per-GPU hashrates, power usages and temperatures in JSON format (relies on NVML, can fail if NVML fails - if so, JSON contains error field).