GPU Boids Simulation

A real-time flocking simulation using a compute shader in Godot 4, based on Craig Reynolds' classic Boids algorithm from the 1987 paper "Flocks, Herds, and Schools: A Distributed Behavioral Model".

Overview

This project demonstrates emergent flocking behavior on the GPU, leveraging compute shaders for high performance. The simulation is capable of handling 10,000+ boids in real time.

The three core rules of flocking are implemented:

• Separation: Avoid crowding neighbors
• Alignment: Steer towards the average heading of neighbors
• Cohesion: Steer towards the average position of neighbors

Features

• GPU-accelerated boid simulation using Godot 4's compute shaders
• Real-time performance with large numbers of boids (10,000+)
• Efficient batch rendering with MultiMeshInstance2D
• Simple, clean Godot project structure

Controls

• Adjust simulation parameters in the script for different behaviors
• No interactive UI (for interactive controls, see the CPU version)

Performance Optimizations

• All boid logic runs on the GPU via a compute shader
• Minimal data transfer between CPU and GPU
• Single draw call for all boids
• Designed for smooth performance even with very large flocks

How It Works

• Boid positions and velocities are stored in GPU buffers
• The compute shader updates all boids in parallel each frame
• Rendering is handled by a MultiMeshInstance2D for efficiency

Requirements

• Godot 4.0 or newer
• GPU with compute shader support (WebGL2 compatible for web export)

License

MIT License - See LICENSE for details

References

• Reynolds, C. W. (1987). Flocks, Herds, and Schools: A Distributed Behavioral Model
• Red3d.com - Craig Reynolds' Website
• Godot Docs: Compute Shaders