H-Ray

Extensible multi-threaded raytracer in Haskell.

Architecture overview

The architecture is inpired by NVIDIA's OptiX raytracer.

Functional programming style amazingly fits the core concept of raytracing: cast individual independent rays, mutate the data using arbitrary functions and return the result. So let's take advantage of it!

Behaviour

The raytracer consists of several different kinds of functions:

Initial Ray Generation
Starts the whole raytracing process. Returned rays are traced against the scene one by one.

Intersection
Checks if a ray intersects the geometry and computes an intersection point and normal.

Bounding Volume
For a given geometry returns a container that encloses it.
Examples: AABB, Bounding Sphere

Closest Hit (a.k.a. Shading)
Takes the first intersection along the ray and returns one of the following:

• Color
• A set of new rays to compute and a function to assemble the color once they are computed

Any Hit (a.k.a. Masking)
Takes the first intersection along the ray and returns one of the following:

• Color
• Nothing
  Can be used for visibility checks.

Data

The raytracer is executed on a scene with context.

Scene

The scene is represented as a set of separate objects with functions assigned to them:

• Bounding Volume
• Intersection
• Closest Hit
• Any Hit (optional)

Context

Context holds different raytracer parameters such as:

• Physical lights
• Ambient lighting
• Setting specifying whether to use AnyHit / ClosestHit / both
• ...

Each time a set of rays is created, a context is assigned to them and passed to all the consequent functions.