ProceduralGeneration

Current state of the application: Chunks being generated based on a noise map and streamed to the GPU for drawing.

Real-time terrain generation in a 3D world from a procedural density field. CUDA kernels are responsible for evaluating the density field and extracting a mesh out of it via Marching Cubes, which then can be drawn by OpenGL.

Building

With CMake Workflow

1. Run workflow

   cmake --workflow --preset <platform>-<type>

   where <platform> can be any of gcc, clang, and windows and <type> can be either release, or debug.

2. Run the Application

   ./build/<type>/src/app/ProceduralGeneration

Manually

1. Configure CMake

   cmake --preset <platform>-<type>

2. Build the Application

   cmake --build --preset <platform>-<type>

3. Run the Application

   ./build/<type>/src/app/ProceduralGeneration

NVIDIA CUDA Support

Please note: macOS does not support CUDA Please note: CUDA support has only been tested and verified by myself on a Linux system with CUDA Toolkit version 13.3, reliability and functionality on Windows may be lacking.D

If you want to use CUDA (NVIDIA GPUs only), you can pass the needed flags to CMake at configuration time like so:

cmake --preset <platform>-<type> -DPEN_ENABLE_CUDA=ON -DCMAKE_CUDA_HOST_COMPILER=<path_to_a_c++23_compatible_compiler>

For example on Linux, use /usr/bin/clang++-19 for CMAKE_CUDA_HOST_COMPILER.

And afterward, build like normal:

cmake --build --preset <platform>-<type>

CMake will then figure out where the nvcc compiler is and give it the proper settings.

Acknowledgments / Credits

• LearnOpenGL.com
• OpenGL Sphere by Song Ho Ahn (안성호)
• Polygonising a scalar field by Paul Bourke
• Marching Cubes with OpenGL by DKGR

History

First stage of the project: A torus with a lighting sphere in the center

Second stage of the project: Single chunk that is generated based on noise.