How It Works¶
This section covers the mathematical and algorithmic foundations of RayON's rendering pipeline. If you want to understand why the renderer produces the images it does, start here.
The rendering equation, Monte Carlo estimation, the bounce loop, and why noise decreases with more samples.
Lambertian, mirrors, glass, GGX conductors, thin films, clear coat, lights and the volume material, with the math and a YAML snippet for each.
Surface Area Heuristic tree construction, GPU iterative traversal, and 64-byte cache-line-aligned nodes.
Why cosine-weighted sampling beats uniform sampling, and how the orthonormal basis maps directions to world space.
How 128-dimensional Joe–Kuo direction vectors and Owen scrambling deliver faster convergence than pseudo-random PCG.
Next event estimation, the power heuristic, light sampling on rectangles and spheres, and how the GPU weights emitter hits.
Four scenes with a slider: PCG vs Sobol', MIS off vs on, and both combined.
Every optimisation applied so far, from thread-block shape to hardware triangles, with its measured impact.