Skip to content

README.md

Latest commit

 

History

History
209 lines (152 loc) · 5.4 KB

README.md

File metadata and controls

209 lines (152 loc) · 5.4 KB
Raw
GPUMiner

⚡ GPUMiner

A high-performance GPU-focused cryptocurrency miner built with OpenCL, featuring advanced candidate generation, intelligent filtering, and seamless integration with modern mining backends.

Designed for speed, scalability, and experimentation with GPU-accelerated hashing pipelines.

🚀 Features

⚡ GPU Accelerated Mining

  • OpenCL-based GPU hashing engine
  • Optimized kernel execution for high throughput
  • Configurable work sizes and tuning parameters
  • Multi-GPU support

🧠 Smart Candidate Generation

  • Predictive candidate filtering before verification
  • Tail-based ranking and scoring system
  • Top-K selection for efficient CPU verification
  • Reduced unnecessary hashing overhead

🔗 Backend Compatibility

  • Monero RPC integration
  • P2Pool / Stratum compatibility
  • Broker-based job support
  • Flexible job ingestion pipeline

🧮 Advanced Share Processing

  • Candidate ranking using:
    • predicted_tail
    • rank_score
    • quality metrics
  • Reduced duplicate and low-quality submissions
  • Efficient submit pipeline

📊 Performance Optimization

  • Dynamic workload tuning
  • Adaptive scan intensity
  • Queue-aware processing
  • GPU memory optimization

🖥️ Monitoring & Logging

  • Real-time GPU stats
  • Hashrate tracking
  • Candidate + share logging
  • Debug + performance logs

🏗️ Architecture Overview

        ┌────────────────────┐
        │   Job Source       │
        │ (RPC/Stratum/etc) │
        └─────────┬──────────┘
                  │
                  ▼
        ┌────────────────────┐
        │ Job Manager        │
        └─────────┬──────────┘
                  │
                  ▼
        ┌──────────────────────────┐
        │ GPU Scanner (OpenCL)     │
        │ Kernel Execution         │
        └─────────┬────────────────┘
                  │
                  ▼
        ┌──────────────────────────┐
        │ Candidate Selection      │
        │ (Top-K + Filtering)      │
        └─────────┬────────────────┘
                  ▼
        ┌──────────────────────────┐
        │ CPU Verification         │
        │ (RandomX or backend)     │
        └─────────┬────────────────┘
                  ▼
        ┌──────────────────────────┐
        │ Share Submission         │
        └──────────────────────────┘

📦 Project Structure

gpuminer/
├── gpu_scanner.py         # OpenCL GPU scanning engine
├── opencl_kernels/        # GPU kernels
├── miner_core.py          # Main mining logic
├── job_manager.py         # Job handling
├── verifier.py            # CPU verification
├── submitter.py           # Share submission logic
├── config.py              # Configuration
├── utils/                 # Utilities
└── logs/                  # Runtime logs

⚙️ Installation

Requirements

  • Python 3.10+
  • OpenCL-compatible GPU
  • GPU drivers installed
  • (Optional) RandomX runtime for verification

Setup

git clone https://github.com/nate2211/gpuminer.git
cd gpuminer

pip install -r requirements.txt

🔧 Configuration

Example configuration:

GPU_ENABLED = True
GPU_DEVICES = [0]

WORK_SIZE = 256
GLOBAL_WORK_SIZE = 65536

POOL_URL = "stratum+tcp://127.0.0.1:3333"
WALLET_ADDRESS = "YOUR_WALLET"

ENABLE_VERIFICATION = True

▶️ Running the Miner

python miner_core.py

⚡ GPU Tuning

You can tune performance using:

  • WORK_SIZE → local workgroup size
  • GLOBAL_WORK_SIZE → total threads dispatched
  • Kernel parameters inside opencl_kernels/

Tips:

  • Start small, then scale up
  • Monitor GPU temperature and usage
  • Avoid exceeding VRAM limits

🧠 Advanced Features

🔍 Predictive Filtering

Reduces CPU load by selecting only high-quality candidates before verification.

📉 Adaptive Load Control

Adjusts GPU workload dynamically based on system conditions.

⚡ Top-K Candidate Selection

Only the best candidates are passed forward, improving efficiency.

🧮 Tail-Based Ranking

Uses hash tail analysis to prioritize high-value candidates.

📊 Performance Notes

  • Best performance achieved with tuned OpenCL kernels
  • Works well alongside CPU miners
  • Designed for high-throughput environments

⚠️ Disclaimer

This project is for educational and experimental purposes.
Running GPU miners may consume significant power and hardware resources.

📄 License

MIT License

⭐ Contributing

Pull requests are welcome.
For major changes, open an issue first to discuss what you'd like to change.