This is my take on Peter Shirley's Ray Tracing in One Weekend book.
This project uses the OCaml programming language, which is a general-purpose, functional programming language.
- Install OCaml (including the opam package manager)
- Switch to a OCaml version that has the flambda compiler enabled (the
ocaml-option-flambdaoption) Check the releases page for the newest versonopam switch create flambda ocaml-variants.5.1.0+options ocaml-option-flambda
- Clone the repository
- Optional: Change sample and thread count in
bin/ray_tracing.ml - Install dependencies
opam install . --deps-only - Build the project
dune build --release
- Run the executable
./bin/ray_tracing.exe
I've already implemented Peter Shirley's ray tracing in various programming languages running on CPU & GPU and compared their performance.
The performance was measured on the same scene (see image above) with the same amount of objects, the same recursive depth, the same resolution (1920 x 1080). The measured times are averaged over multiple runs.
Reference system: AMD Ryzen 9 5900X (12 Cores / 24 Threads) | AMD Radeon RX 6800 XT
| 1 sample / pixel | 100 samples / pixel | |
|---|---|---|
| Elixir | 67,200 ms | N/A |
| JavaScript - Node.js | 4,870 ms | 308 s |
| Go | 1,410 ms | 142 s |
| OCaml | 795 ms | 75 s |
| Java | 770 ms | 59 s |
| C++ | 685 ms | 70 s |
| Rust | 362 ms | 36 s |
| C | 329 ms | 33 s |
| GPU - Compute Shader | 21 ms | 2 s |
| GPU - Vulkan Ray Tracing Extension | 1 ms | 0.1 s |