- speed - both SSE and AVX support
- meshes can be loaded via .obj file
- scenes are configured easily via yaml (see scenes/example_scene.yml)
git clone https://github.com/baurst/rbrt.git
cd rbrt
# important: build with avx and sse enabled
RUSTFLAGS='-C target-feature=+sse,+avx' cargo build --release
# download the stanford bunny
wget http://graphics.stanford.edu/~mdfisher/Data/Meshes/bunny.obj
# run the raytracer: adjust samples according to your needs
./target/release/rbrt --target_file out.png --height 768 --width 1024 --samples 50 --config scenes/example_scene.yamlDo not forget to set RUSTFLAGS='-C target-feature=+sse,+avx', otherwise a scalar fallback will be used and the raytracer will run very slow.
On my machine the AVX enabled version runs over 5(!) times faster than the scalar fallback. For SSE, the speedup over the scalar version is at a factor of around 3.3.
The raytracer uses a right-handed coordinate system, with negative z pointing through the camera lens towards the scene.
For an example scene configuration check out scenes/example_scene.yaml. Triangle meshes can be loaded by providing a path to an .obj file, spheres can also be added to the config yaml. For each scene element, a material definition needs to be specified in the config yaml. Currently, three materials are supported:
- matte, lambertian material with paramter albedo, e.g. (0.02,0.2,0.02)
- transparent dielectric material with parameter ref_idx, e.g. 1.8
- metallic material, with scalar parameter roughness and albedo
All of these features can be seen in the example_scene.yaml file in the config folder.
This is yet another raytracer written in Rust. Coming from C++, Rusts ownership concepts (and easy way of managing dependencies and cross-platform builds) fascinated me and so I thought it could be fun to write a lightweight but capable raytracer to get to know the language better. The inspiration for this project came from the excellent book 'Raytracing In One Weekend' by Peter Shirley as well as ssloys awesome tinyraytracer.