Boids Simulation with CUDA

This application is an implementation of a 3D boids simulation. Each boid is represented by a wireframe of a tetrahedron and is enclosed in an aquarium. User can add box obstacles and adjust the following parameters in real time:

• Separation: steer to avoid crowding local flockmates, Alignment: steer towards the average heading of local flockmates,
• Cohesion: steer to move towards the average position (center of mass) of local flockmates,
• View radius: determines the radius of a sphere containing all visible boids by every boid,
• Noise: determines the amount of random acceleration applied each frame,
• Minimum speed,
• Maximum speed.

User can reset the simulation with different aquarium size, boids count and choose one of the following algorithms:

1. CPU naive algorithm,
2. GPU naive algorithm,
3. grid based GPU algorithm in the 1st variant,
4. grid based GPU algorithm in the 2nd variant.

Both 3 and 4 algorithms are based on a grid approach with sorting, which is described here (page 6).

The difference is that algorithm 3 capitalizes on the fact that many boids within a singular thread block, are within the same grid cell. To speed up the boid acceleration update process for these boids sharing a cell, shared memory is used. Conversely, Algorithm 4 overlooks this observation.

All 3 GPU methods speeds are compared on the following graph:

The conclusion drawn from the above graph is that the theoretical observation used for algorithm 3 is slowing the algorithm down.

Usage

1. Use W, S, A, D to rotate the camera and Q, E to zoom in/out.
2. Use Simulation window to
   • restart the simulation with different aquarium size, boids count and algorithm,
   • modify the simulation parameters in real time,
   • add box obstacles.

Requirements

You need NVIDIA CUDA GPU to run the application. This application has been tested on the following GPU's:

GPU	Memory	Compute Capability
NVIDIA GTX 1650Ti	4GB	7.5
NVIDIA GTX 1060	3GB	6.1
NVIDIA GTX 1070	8GB	6.1
NVIDIA RTX 3070	8GB	8.6

Compilation Guide

In order to compile the project you need to download and install NVIDIA CUDA Toolkit and CMake first.

Linux

1. Clone the repository to the desired location git clone https://github.com/migoox/boids-simulation,
2. Navigate to the cloned directory and run cmake .,
3. Run make.

Windows

1. Clone the repository to the desired location git clone https://github.com/migoox/boids-simulation,
2. Navigate to the cloned directory and run cmake .,
3. Open boids_simulation.sln using Visual Studio,
4. Set boids_simulation as the Startup Project (right click on boids_simulation and choose Set as Startup Project),
5. Compile using Debug/Release mode.