NBody simulation examples

Various N-body implementations, demonstrating multiple CPU and CUDA optimizations

Building from source

Prerequisites

A C++ 17 compiler (GCC, Clang, MSVC)
CUDA SDK 11.0 or newer
CMake 3.25 or newer

Getting the Source Code

This source tree uses [Git submodules][5] for third party libraries, and --recursive is required when cloning:

git clone --recursive <repository>

Building on Windows

When building on Windows with MSVC, the CMake configuration must be done from a MSVC command prompt (ex. run vcvarsall.bat)

Building on Linux

A few additional system tools and libraries may be required. The CMake configuration phase might fail if they are missing, and it would log instructions on how to install them.

For example, on a recent Ubuntu / Debian system, use the following commands:

sudo apt-get install curl zip unzip tar
sudo apt-get install libxmu-dev libxi-dev libgl-dev
sudo apt install libxinerama-dev libxcursor-dev xorg-dev libglu1-mesa-dev

Building with CMake

Tip: CMake uses the CUDA_PATH environment variable. Make sure it is set and points to the right location before running cmake.

This project uses CMake, so the general build process looks like:

mkdir build && cd build
cmake .. -DCMAKE_BUILD_TYPE=Release
make -j8 # replace 8 with the number of cores

Running the N-body simulation

Usage: nbody [--help] [--version] [--width VAR] [--height VAR] [--bodies-count VAR] [--static-velocities] --plugin VAR
Optional arguments:
  -h, --help               shows help message and exits
  -v, --version            prints version information and exits
  -w, --width              Simulation window width [nargs=0..1] [default: 1024]
  -h, --height             Simulation window height [nargs=0..1] [default: 1024]
  -n, --bodies-count       Bodies count [nargs=0..1] [default: 5000]
  -s, --static-velocities  Initialize the bodies with zero initial velocities
  --plugin                 Implementation plugin, valid options:
                            - cpu_openmp
                            - cpu_threadpool
                            - cuda
                            - cuda_rsqrt
                            - cuda_soa
                            - cuda_surface_draw
                            - cuda_tiled
                            - cuda_vbo
                            - dummy [nargs=0..1] [default: "cuda"]

Running on Linux + Wayland

If the app exits with the following initialization error:

CUDA API Error: 999 (unknown error)
 > cudaGraphicsGLRegisterImage(...)

... then try setting the following environment variables first:

export __NV_PRIME_RENDER_OFFLOAD=1 __GLX_VENDOR_LIBRARY_NAME=nvidia

Math

This 2D N-body simulation is based on a simple Verlet integration using the following equations: