Cellular Automata

Synopsis

The purpose of this project is:

to provide an easily reusable interface for creating videos from a series of 2D-vectors describing each frame
to adapt such an interface with the goal of simulating discrete bidimentional cellular automata on rectangular or torus grids
to implement specific such automata

Preview

Brian's brain

Game of life

Abelian sandpile

Langton's ant

How to build and run

make sure that ffmpeg is installed
edit main to choose automata and setup (it will eventually be possible to load a configuration file to avoid this step)
compile and run with cargo: cargo run --release (release mode is recommended since the time lost optimizing is easily compensated during the execution, see NOTE #1)
open the resulting .avi video

The front page of LifeWiki contains a collection of patterns. It is recommended to download the archive and extract it into data/* or data/patterns/*. This will give you access to 15.6MB of patterns, from which 1446 *.cells and 2381 *.rle can be read by this project.

Documentation is available for this project with $ cargo doc --open.

Work in progress and future improvements

Work is currently in progress on:

Abelian sandpiles (https://en.wikipedia.org/wiki/Abelian_sandpile_model)
Conway's game of life and any other life-like automata (https://en.wikipedia.org/wiki/Life-like_cellular_automaton)
Brian's brain (https://en.wikipedia.org/wiki/Brian%27s_Brain)
Langton's ant (https://en.wikipedia.org/wiki/Langton%27s_ant)

TODO:

enable reading from text files to initialize grid (beta available for game of life)
create scanner to load a screenshot of a game of life state into an initializer (beta available in lifescan/)

It should be noted that this project relies heavily on ffmpeg. Luckily ffmpeg is available for all OS, but the commands may need tweaks to execute properly on non-Linux distributions.

NOTE #1: About runtime vs compile-time

We first compare build times:

$ cargo clean -p cellular_automata
$ cargo build
min/max/avg5 of 7 : 0.54/0.62/0.56

$ cargo clean --release -p cellular_automata
$ cargo build --release
min/max/avg5 of 7 : 1.93/2.00/1.97

Compiling in dev rather that release can spare us a bit over a second.

But what of runtime ?

Both builds were run with the following configuration:

Lifelike(23/3) rules
a 200x300 field
a single bricklayer loaded at the top right corner (data/patterns/bricklayer.rle)
5000 generations

The results speak for themselves:

dev:       12m28s    (incl. 45s running ffmpeg)
release:    1m29s    (incl. 45s running ffmpeg)

Thus this project should always be run with cargo run --release.

Vanille-N / cellular_automata