RRad was a piece of software I programmed as part of a computer graphics course at university, and was the direct lead-up to my Maser's Thesis, the more sophisticated RTRad.
It is a GPU-based implementation of the Radiosity algorithm made with the open-source API OpenGL and it highlights with clarity how visibility is the only major hurdle that prevents the widespread adoption of GPU-based radiosity.
RRad approximates a scene through basic geometric shapes (spheres and triangles) and then loops over each shape and performs a simple, discrete ray-intersection on each. This geometric approximation is hard-coded into the shader’s code itself, making the application simple and lightweight, but entirely unsuitable to complex environments. Fig. 4.2 shows the default RRad scene with a lightmap of 512 × 512 pixels (~260k elements), for which a single bounce of light requires approx. 1.5 seconds of computation time on a GeForce RTX 2070S GPU.