Skip to content

Dalamar42/rayt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Rayt

This is a Rust implementation of the ray-tracer from Peter Shirley's excellent books, Ray Tracing in One Weekend, Ray Tracing: The Next Week, and Ray Tracing: The Rest of Your Life.

This is not by any means a fully featured ray-tracer. Rather, this is something I built in order to learn Rust and ray-tracing. I am sharing it in case it is useful to others trying to do the same.

Functionality

Rayt stores and reads scene data in its own format using yaml files. Those are nothing more than a yaml serialisation of the corresponding Rust objects.

There are two modes of operation described in more detail in the next section.

  • generate which will generate the yaml file for a scene based on presets in the code
  • render which will render the scene in the given yaml

Various examples from the books are provided in the form of scenes. The generate mode can be used to quickly regenerate the config files after making changes to them in the code, but for small scenes the files can be easily directly edited as well.

How to run

This code was developed and tested using Rust 1.42, however it is not using any bleeding edge features of the language so any reasonably recent stable version should be enough to compile it.

To install the binary in your local environment:

cargo install --path ${PATH_TO_THIS_REPO} 

Help:

rayt --help
rayt generate --help
rayt render --help

Generate the Cornell box scene yaml file:

rayt --config config/cornell_box.yaml generate --scene CornellBox

To render the Cornell box:

rayt --config config/cornell_box.yaml \
    render --width 512 --rays 1000 --threads 8 --output output/cornell_box.png

Some scenes use assets as textures, for example the final scene from the Ray Tracing: The Next Week book. These assets can be passed in using --asset:

rayt --config config/next_week_final.yaml \
    render --width 512 --rays 1000 --threads 8 --asset assets/earth.jpg --output output/next_week_final.png

The option --threads can be used to control how many threads the renderer should use and the option --rays will determine how many rays (samples) will be taken for each pixel. Approximately 1000 samples should be enough to produce a decent image with some noise from the provided scenes, but more are needed for a clear image. The rendering times will increase quite significantly with the number of samples selected and the size of the image. On my machine rendering next_week_final.yaml with 5000 samples, and a width of 1024 pixels took around 9 hours. Conversely, the cornell_box.yaml with a 1000 samples, and a width of 512 finishes in a few minutes.

A Makefile is included with some convenience targets:

  • make regenerate-scenes will create all the scene config yaml files
  • make render-test will render all scenes using a moderate resolution and number of rays and put the generated images in output/test
  • make cornell-test is the same as make render-test, but only for the Cornell box
  • make regenerate-samples will create high-resolution / high-ray-count versions of the Cornell box and the final image from book 2 and put them in output/samples

Samples

The directory samples contains images generated with a high number of rays, specifically 5000 rays per pixel, to showcase what kinds of images rayt can render.

Assets

The included assets were taken from the following sources:

License

See the LICENSE file that is included with this repository.

About

Monte Carlo ray tracer developed using Rust

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published