An implementation of the Barnes-Hut algorithm, running in real-time on the GPU, developed using CUDA C++ and OpenGL, which scales to systems of few million particles on a NVIDIA RTX 500 Ada Laptop GPU.
The program was developed and tested on Ubuntu 22.04 LTS using the CUDA Toolkit 12.9, SDL2 and OpenGL 4.3 core. The basic dependencies can be installed under Ubuntu by running
sudo apt install build-essential libgl1-mesa-dev libsdl2-dev libglew-dev
Detailed installation instructions for the CUDA Toolkit can be found here.
To build the project for a GPU with compute capability xy, run
make CUDA_ARCHS="xy"
More info on compute capability versioning here. If not specified, the program will be built for the following architectures: 50 60 70 75 80 89.
Build files will be placed under build/, running
make clean
will remove the build/ directory but keep the executable.
The program can be executed by running
./main
Drag to orbit the camera around the origin, and scroll to zoom. Pressing o toggles the octree visualization, space pauses/resumes the simulation, 1, 2 and 3 spawn the particles in different initial configurations.
 |
 |
|---|---|
| Basic visualization | Octree visualization |
 |
 |
|---|---|
| Self-gravitating disk of 262,144 particles | Cold collapse of uniform sphere |
 |
 |
|---|---|
| Disk of 524,288 particles running at ~37 FPS (θ=0.75) | Two disks of 262,144 particles each running at ~30 FPS (θ=0.75) |
 |
 |
|---|---|
| Total energy for 32,768 particles | Relative acceleration error for 32,768 particles |
 |
 |
|---|---|
| Execution time of a single step for θ=0.75 | Scalability for different values of θ (log-log) |
-
Tero Karras. 2012. Maximizing parallelism in the construction of BVHs, octrees, and k-d trees. In Proceedings of the Fourth ACM SIGGRAPH / Eurographics conference on High-Performance Graphics (EGGH-HPG'12). Eurographics Association, Goslar, DEU, 33–37.
-
Robin Cazalbou, Florent Duchaine, Eric Quémerais, Bastien Andrieu, Gabriel Staffelbach, and Bruno Maugars. 2024. Hybrid Multi-GPU Distributed Octrees Construction for Massively Parallel Code Coupling Applications. In Proceedings of the Platform for Advanced Scientific Computing Conference (PASC '24). Association for Computing Machinery, New York, NY, USA, Article 14, 1–11. https://doi.org/10.1145/3659914.3659928
-
Jeroen Bédorf, Evghenii Gaburov, and Simon Portegies Zwart. 2012. A sparse octree gravitational N-body code that runs entirely on the GPU processor. Journal of Computational Physics 231, 7 (2012), 2825–2839. DOI:https://doi.org/https://doi.org/10.1016/j.jcp.2011.12.024