Barycuda is a tiny CUDA accelerated library with no dependencies that performs various barycentric operations. This is an attempt to speed-up a 3D renderer that I'm working on.

Currently, the library exposes the following functions:

• point_in_simplex - takes an array of points, the vertices of a 2D/3D simplex, and tells you if each point lies inside the simplex.
• bary_simplex - takes an array of points, the vertices of a 2D/3D simplex, and returns the barycentric coordinates for each point.

You can use Barycuda directly from C++ (see src/testBary.cpp for an example), or if you prefer using Python, Barycuda has a Python wrapper available. You can install it by doing

pip install pybarycuda

Installation using pip will come bundled with a prebuilt libbarycuda.so for Linux x86-64. The linker paths are correctly set out of the box, and you shouldn't have to change your LD_LIBRARY_PATH to be able to use it.

I plan to support other architectures in the future, but if you want to do it yourself, you can build the binaries (see below) and add libbarycuda.so to LD_LIBRARY_PATH to make the Python wrapper work.

You will need a CUDA capable system for any of this to work.

# inside the project directory
mkdir build
cd build
cmake ../
make

Feel free to add more geometric operations related to graphics rendering and raytracing.

Shirley, P. (2009) Fundamentals of Computer Graphics
https://en.wikipedia.org/wiki/Barycentric_coordinate_system
https://math.stackexchange.com/questions/1226707/how-to-check-if-point-x-in-mathbbrn-is-in-a-n-simplex