CPU-based Ray Tracer

Project is a C++17 implementation of a ray tracer. Guidance was used from the Ray Tracing Challenge. Everything was written from scratch, so no external libraries are used other than Catch2 for testing.

I wanted to do this, so I can:

• relearn some concepts that I explored in computer graphics
• do some performance engineering!
  • I used perf, valgrind, flamegraph, hyperfine!
  • caching the inverse matrix improved program speed by insane amounts
• learn and use C++17
• use Nix for a C++ project!

Getting Started

Prerequisites

This project requires cmake and catch2, so be so to have those installed. This project has been set up using Nix, so if you have it installed, you can easily use nix develop and have the environment I used to build this project.

Building

Run make debug or make release to build debug or release binaries respectively in a build folder. You can also use cmake to generate a way to build this project in a way you want.

Running

There are two targets that are generated: tester and Rt. Both currently needs to be run in top most directory as it references relative paths in obj. That should be changed in the future, but I didn't want to deal with it.

tester just runs tests. Rt prints to stdout the contents of a ppm file. Redirect the output to save it. Right now, Rt just renders a scene described in src/rt.cpp. The rendered scenes are currently hardcoded instead of being configurable, so rendering another scene requires changing src/rt.cpp and recompiling. Hopefully, I will change that in the future.

Improvements that I can do

• Configurable way to render different scenes without recompiling
• Read from some data format or DSL to render different scenes
• Multithreading
• Webassembly, so you can see some renders on the web
• SIMD/AVX2 optimizations for my math primitives
• BVH (Bounding volume hierarchy)
• CSG (Constructive solid geometry)
• Adding more patterns

Sample images

Balls

Teapot