Skip to content

Home

Jump to bottom
Muhammad Awad edited this page Aug 30, 2025 · 1 revision

Welcome to the Iris Wiki! 🚀

Iris is AMD's powerful and intuitive multi-GPU programming framework built on Triton. We're here to make distributed GPU computing accessible to everyone! It's designed by experts and built for scale, ensuring you get a robust and efficient experience.

🎯 What is Iris?

Iris is a Triton-based framework for Remote Memory Access (RMA) that provides SHMEM-like APIs for multi-GPU programming. Think of it as your friendly companion for writing multi-GPU code as easily as single-GPU code! We've focused on clean abstractions and minimal dependencies to keep things simple and powerful.

✨ Why Iris?

  • 🚀 Simple & Intuitive: Write multi-GPU programs without the complexity
  • ⚡ High Performance: Inherits Triton's programmability and performance
  • 🔧 Familiar APIs: SHMEM-like patterns, Triton-style device APIs, and PyTorch-like host APIs
  • 🏗️ Expert Design: Built from scratch by GPU and distributed computing experts
  • ⚡ Minimal Dependencies: Only Triton, PyTorch, HIP runtime, and mpi4py

📚 Examples & Learning

We have comprehensive examples to help you learn Iris:

🚀 Run Examples

# Basic operations
mpirun -np 8 python examples/00_load/load_bench.py

# GEMM benchmarks
mpirun -np 8 python examples/07_gemm_all_scatter/benchmark.py --benchmark --validate

🤝 Join Our Community!

We're building something special here, and we'd love for you to be part of it! Here's how you can get involved:

💬 Have Questions? Start a Discussion!

  • GitHub Discussions: Perfect for asking questions, sharing ideas, or just chatting about multi-GPU programming
  • General Questions: "How do I implement X pattern?"
  • Best Practices: "What's the recommended way to do Y?"
  • Show & Tell: Share your Iris projects and experiences!

🐛 Found a Bug? Report It!

  • GitHub Issues: Help us make Iris better by reporting bugs
  • Feature Requests: Have an idea? We want to hear it!
  • Documentation: Spotted something unclear? Let us know!

🚀 Want to Contribute? Awesome!

  • Pull Requests: Code improvements, bug fixes, new features
  • Documentation: Help make our docs clearer and more helpful
  • Examples: Share your use cases and implementations
  • Testing: Help us ensure Iris works great for everyone

🆘 Need Help?

🚨 Stuck? We've Got You!

  1. Check our examples - We have working code for most common patterns
  2. Look at the docs - Programming Model, Setup Guide
  3. Start a discussion - Our community is friendly and helpful!
  4. Open an issue - For bugs or clear problems

💡 Pro Tips

  • Start with small examples to understand the patterns
  • Check GPU memory availability vs. your heap size
  • Don't forget to use barriers for synchronization!

🔗 Quick Links

🌟 Ready to Start Your Multi-GPU Journey?

We're excited to see what you'll build with Iris! Remember:

  • Ask questions in Discussions - no question is too basic!
  • Share your experiences - we love hearing success stories
  • Contribute back - help make Iris better for everyone
  • Have fun - multi-GPU programming should be enjoyable!

Happy coding with Iris! 🚀

This project is intended for research purposes and is provided by AMD Research and Advanced Development team.

💬 Start a Discussion | 🐛 Report an Issue | 🚀 Make a PR

Clone this wiki locally