🚀 CUDA Online Judge

Learn GPU Programming Without a GPU

CUDA Online Judge (cudaforces) is an educational platform for learning CUDA programming. Practice GPU programming concepts without expensive hardware through our innovative CPU Transpiler technology.

🌐 Website: cudaforces.com

---

📖 About This Repository

This repository provides documentation, guides, and resources for:

• 📚 Learning CUDA — Study guides and example problems
• 🔧 Building Tools — Public API documentation for developers
• 🎯 Problem Solving — Example solutions and coding patterns

---

📁 Repository Structure

cudaforces/
├── docs/
│   ├── QUICK_GUIDE.md          # Quick start guide
│   ├── COMPLETE_GUIDE_EN.md    # Comprehensive coding guide
│   └── PUBLIC_API.md           # Public API documentation
│
├── problems/                   # Example problems with solutions
│   ├── easy/
│   ├── medium/
│   └── hard/
│
└── README.md

---

📚 Documentation

Document	Description
QUICK_GUIDE.md	Quick introduction to CUDA Online Judge
COMPLETE_GUIDE.md	Complete coding guide with examples and FAQ
PUBLIC_API.md	API reference for building external tools

---

⚙️ How It Works

CUDA Online Judge uses a CPU Transpiler that converts your CUDA code to run on CPU:

┌─────────────┐      ┌─────────────┐      ┌─────────────┐      ┌─────────────┐
│  CUDA Code  │ ───► │ Transpiler  │ ───► │  C++ Code   │ ───► │ CPU Execute │
│   (.cu)     │      │  + Validate │      │  (OpenMP)   │      │  & Judge    │
└─────────────┘      └─────────────┘      └─────────────┘      └─────────────┘

This allows you to:

• ✅ Learn CUDA syntax and concepts
• ✅ Verify algorithm correctness
• ✅ Practice without GPU hardware

⚠️ Note: Performance benchmarking is not available — the platform is for correctness verification only.

---

🔌 Public API

Build your own tools using our Public API!

Base URL: https://cudaforces.com/api/v1

Available Endpoints

Endpoint	Description
GET /problems	List all public problems
GET /problems/{id}	Get problem metadata
GET /problems/{id}/stats	Get problem statistics
GET /rankings	Get user rankings
GET /users/{username}	Get user public profile

Quick Example

import requests

response = requests.get("https://cudaforces.com/api/v1/problems", params={
    "difficulty": "Easy",
    "per_page": 10
})

for problem in response.json()["data"]["problems"]:
    print(f"#{problem['id']} - {problem['title']}")

📖 Full documentation: PUBLIC_API.md

---

💡 Tool Ideas

Here are some tools you can build with the API:

Tool	Description	Main Endpoints
Discord/Slack Bot	Daily problem recommendations	/problems
Progress Dashboard	Visualize learning progress	/users/{username}
GitHub Badge	Display rank on your profile	/users/{username}
VS Code Extension	Browse problems in editor	/problems
CLI Tool	Command-line problem browser	All endpoints

---

📝 Example Problem

Print Hello World CUDA N Times (Easy)

Print "Hello World Cuda" exactly n times using CUDA parallelism.

__global__ void helloKernel(int* dummy, int n) {
    int idx = threadIdx.x;
    if (idx < n) {
        printf("Hello World Cuda\n");
    }
}

int main() {
    int n;
    scanf("%d", &n);
    
    int* d_dummy;
    cudaMalloc(&d_dummy, sizeof(int));
    
    helloKernel<<<1, n>>>(d_dummy, n);
    
    cudaDeviceSynchronize();
    cudaFree(d_dummy);
    
    return 0;
}

📖 More problems: problems/

---

✅ Code Requirements

All submissions must follow these rules:

Requirement	Description
✅ Kernel Required	At least one __global__ function
✅ Use Parallelism	Must use threadIdx, blockIdx, etc.
✅ GPU Memory	Must use cudaMalloc, cudaMemcpy
❌ No STL	Cannot use std::vector, std::string
❌ No Convenience Functions	Cannot use qsort, strcpy, etc.

This ensures you learn authentic CUDA programming patterns!

---

🚀 Getting Started

1. Visit cudaforces.com
2. Read the Coding Guide
3. Start with Easy problems
4. Build tools using the Public API

---

🤝 Contributing

Contributions are welcome! You can help by:

• 📝 Improving documentation
• 🐛 Reporting issues
• 💡 Suggesting new features
• 🔧 Building and sharing tools

---

📧 Contact

• Email: ejpark29@gmail.com
• Website: cudaforces.com
• Q&A Board: cudaforces.com/board

---