gocuda

GoCUDA - Professional CUDA + Go Framework

A high-performance, production-ready framework for GPU-accelerated computing in Go with intelligent CPU/GPU dispatch, automatic memory management, and comprehensive performance monitoring.

🚀 Features

Core Capabilities

• Intelligent Dispatch: Automatic CPU/GPU selection based on workload characteristics
• Memory Management: Advanced memory pooling and automatic cleanup
• Performance Monitoring: Real-time metrics collection and analysis
• Batch Processing: Efficient batch operations for improved throughput
• Concurrent Workers: Optimized worker pools for maximum performance

Professional Features

• Comprehensive Logging: Structured logging with configurable levels
• Error Handling: Robust error handling and recovery mechanisms
• Configuration Management: Flexible configuration with optimal defaults
• Resource Management: Automatic resource cleanup and leak prevention
• Metrics Export: Integration with monitoring systems

Developer Experience

• Simple API: Clean, intuitive interface for common operations
• Extensive Documentation: Comprehensive guides and examples
• Performance Optimization: Built-in performance analysis tools
• Production Ready: Tested and optimized for production workloads

📋 Requirements

• Go 1.21+
• CUDA 11.0+ (with compatible GPU drivers)
• GCC/G++ compiler (for CGO compilation)
• Linux/WSL2 (primary support)

🔧 Installation

# Install the framework
go get github.com/BasedCodeCapital/gocuda

# Install CUDA dependencies (Ubuntu/Debian)
sudo apt update
sudo apt install nvidia-cuda-toolkit build-essential

# Verify CUDA installation
nvcc --version
nvidia-smi

🎯 Quick Start

Basic Usage

package main

import (
    "context"
    "fmt"
    "log"
    
    "github.com/BasedCodeCapital/gocuda/gocuda"
    "github.com/sirupsen/logrus"
)

func main() {
    // Create configuration
    config := gocuda.DefaultConfig()
    config.LogLevel = logrus.InfoLevel
    config.CPUThreshold = 1024 * 1024 // 1M elements
    config.EnableMetrics = true
    
    // Create and initialize engine
    engine, err := gocuda.NewEngine(config)
    if err != nil {
        log.Fatalf("Failed to create engine: %v", err)
    }
    
    if err := engine.Initialize(); err != nil {
        log.Fatalf("Failed to initialize engine: %v", err)
    }
    
    defer engine.Shutdown()
    
    // Create compute engine
    compute := gocuda.NewComputeEngine(engine)
    
    // Vector addition with intelligent dispatch
    a := []float32{1, 2, 3, 4, 5}
    b := []float32{10, 20, 30, 40, 50}
    
    result, err := compute.VectorAdd(context.Background(), a, b)
    if err != nil {
        log.Fatalf("Vector addition failed: %v", err)
    }
    
    fmt.Printf("Result: %v\n", result)
}

Advanced Usage

// Matrix multiplication
matrixA := make([]float32, 1024*1024)
matrixB := make([]float32, 1024*1024)
// ... fill matrices ...

result, err := compute.MatrixMultiply(context.Background(), matrixA, matrixB, 1024)

// Batch operations
operations := []gocuda.VectorOperation{
    {A: vectorA1, B: vectorB1},
    {A: vectorA2, B: vectorB2},
    // ... more operations
}

err = compute.BatchVectorAdd(context.Background(), operations)

// Performance metrics
metrics, err := engine.GetMetrics()
fmt.Printf("GPU Utilization: %.1f%%\n", metrics.GPUUtilization)
fmt.Printf("Average Execution Time: %v\n", metrics.AvgExecutionTime)

🏗️ Architecture

┌─────────────────┐    ┌─────────────────┐    ┌─────────────────┐
│   Application   │───▶│   GoCUDA        │───▶│   CUDA Runtime  │
│   Code          │    │   Framework     │    │   + GPU         │
└─────────────────┘    └─────────────────┘    └─────────────────┘
                              │
                              ▼
                    ┌─────────────────┐
                    │   Performance   │
                    │   Monitoring    │
                    └─────────────────┘

Core Components

• Engine: Main framework coordinator
• Device Manager: CUDA device detection and management
• Memory Manager: GPU memory allocation and pooling
• Compute Engine: High-level operation interface
• Metrics Collector: Performance monitoring and analysis

📊 Performance Characteristics

CPU vs GPU Dispatch

• Small Operations: Automatically use CPU to avoid GPU overhead
• Large Operations: Leverage GPU parallelism for maximum performance
• Adaptive Thresholds: Configurable based on hardware capabilities

Memory Management

• Memory Pools: Reduce allocation overhead
• Automatic Cleanup: Prevent memory leaks
• Smart Allocation: Optimize memory usage patterns

Benchmark Results (RTX 4070)

Vector Addition (2M elements):
  CPU: 4.5ms
  GPU: 8.2ms (small overhead, uses CPU)
  
Matrix Multiplication (2048x2048):
  CPU: 32.1s
  GPU: 52.2ms (615x speedup!)
  
Batch Operations (100 operations):
  Sequential: 450ms
  Batched: 125ms (3.6x speedup)

⚙️ Configuration

Default Configuration

config := gocuda.DefaultConfig()
// Automatically optimized for your hardware

Custom Configuration

config := &gocuda.Config{
    PreferredDevice: 0,              // Use specific GPU
    CPUThreshold:    512 * 1024,     // 512K elements
    WorkerCount:     8,              // 8 concurrent workers
    BatchSize:       64,             // Process 64 ops per batch
    MemoryPoolSize:  1024 * 1024 * 1024, // 1GB pool
    EnableMetrics:   true,           // Enable monitoring
    LogLevel:        logrus.InfoLevel,
}

Optimal Configuration

// Get hardware-optimized settings
compute := gocuda.NewComputeEngine(engine)
optimalConfig, err := compute.GetOptimalConfig()

📈 Monitoring & Metrics

Built-in Metrics

• Total operations executed
• Success/failure rates
• Average execution times
• GPU utilization
• Memory usage statistics

Accessing Metrics

metrics, err := engine.GetMetrics()
fmt.Printf("Total Operations: %d\n", metrics.TotalOperations)
fmt.Printf("GPU Utilization: %.1f%%\n", metrics.GPUUtilization)
fmt.Printf("Memory Usage: %d MB\n", metrics.MemoryUsage/1024/1024)

Integration with Monitoring Systems

// Export to Prometheus, Grafana, etc.
// (Implementation depends on your monitoring stack)

🧪 Testing

# Run unit tests
go test ./...

# Run benchmarks
go test -bench=. ./...

# Run with race detection
go test -race ./...

🔧 Development

Project Structure

gocuda/
├── gocuda/           # Core framework
│   ├── engine.go     # Main engine
│   ├── device.go     # Device management
│   ├── memory.go     # Memory management
│   ├── compute.go    # Compute operations
│   └── metrics.go    # Performance monitoring
├── examples/         # Usage examples
├── docs/            # Documentation
└── tests/           # Test suites

Building from Source

# Clone repository
git clone https://github.com/BasedCodeCapital/gocuda.git
cd gocuda

# Build framework
make build

# Run tests
make test

# Run examples
make examples

📚 Documentation

• API Reference - Complete API documentation
• Performance Guide - Optimization strategies
• Architecture Overview - Technical deep dive
• Migration Guide - Upgrading from v1.x

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

Development Setup

# Fork the repository
git clone https://github.com/BasedCodeCapital/gocuda.git
cd gocuda

# Install dependencies
go mod tidy

# Run tests
make test

# Submit pull request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• NVIDIA CUDA team for the excellent GPU computing platform
• Go team for the fantastic programming language
• Open source community for contributions and feedback

🔮 Roadmap

v2.0 (Current)

• ✅ Intelligent CPU/GPU dispatch
• ✅ Advanced memory management
• ✅ Performance monitoring
• ✅ Batch processing

v2.1 (Planned)

• 🔄 Multi-GPU support
• 🔄 Streaming operations
• 🔄 Custom kernel integration
• 🔄 Enhanced metrics export

v3.0 (Future)

• 🚀 Machine learning integration
• 🚀 Distributed computing
• 🚀 Advanced optimization algorithms
• 🚀 WebAssembly support

📞 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions
• Email: support@gocuda.dev

Made with ❤️ by the GoCUDA team