Skip to content

Latest commit

 

History

History
178 lines (109 loc) · 7.18 KB

README.md

File metadata and controls

178 lines (109 loc) · 7.18 KB

Game of Life Exercise

An intro to Rust using Conway's Game of Life (heavily based on the SDL2 Game of Life example).

Install Rust

# install Rust by following the prompts
curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

# install the Rust Language Server (for IDE assistance), Rust code formatter, and the stdlib source
rustup component add rls rustfmt rust-src rust-analysis

If you want IDE support:

Repo Setup

Ubuntu

Get some SDL onto your box:

sudo apt install libsdl2-dev

MacOS

Install SDL2 via HomeBrew:

brew install sdl2

Then set up the header locations by adding /usr/local/lib to your path (you can skip this if it's already in your path):

echo 'export LIBRARY_PATH="$LIBRARY_PATH:/usr/local/lib"' >> ~/.bash_profile

Windows

SDL 2.0.22-devel libraries are provided in the gnu-mingw and msvc directories; combined with the power of the build.rs script, they should "just work".

Running

Run the simulation with

cargo run -- solution

then hit Space to pause/unpause.

You can also toggle the state of cells using the mouse's left click.

Hit Escape or Q to quit.

You can also experiment with different sizes, and running in release mode to get a speed boost (turns on various compile-time optimizations at the cost of slower compiles):

cargo run --release -- solution --cell-size 4 --fps 120 -w 300 -h 300

Pass the --help flag to get a full rundown of the options.

Development tips

Get faster type errors with cargo check (it skips actually building the resulting binary).

Run tests with cargo test -p conway (there aren't any useful tests right now, but you could add some; see https://doc.rust-lang.org/book/ch11-01-writing-tests.html or the shorter https://doc.rust-lang.org/rust-by-example/testing/unit_testing.html).

Format your code with cargo fmt.

You can also install the cargo-watch tool if you like continuous feedback in your terminal:

cargo install cargo-watch
cargo watch                         # continuously runs `cargo check`
cargo watch --exec test --exec fmt  # continuously runs `cargo test && cargo fmt`

Exercises

Part 1 - Get the simulation working

Goal: get started with Rust by implementing the game logic rules for Conway's Game of Life.

When you run

cargo run -- mine

then each game tick should cause the following behavior:

  • Any live cell with fewer than two live neighbors dies, as if by under population.
  • Any live cell with more than three live neighbors dies, as if by overpopulation.
  • Any live cell with two or three live neighbors lives on to the next generation.
  • Any dead cell with exactly three live neighbors becomes a live cell, as if by reproduction.

To get you started, there's a broken implementation of Game of Life which you can run with:

cargo run -- broken

You can also start the simulation in a paused state, which is useful for setting up test states (by toggling cells on/off using left mouse click):

cargo run -- mine --paused

More getting started tips:

  • conway/src/game_of_life_mine.rs is provided as a scaffold implementation for you.
  • Read conway/src/game_of_life.rs to understand the interface you're implementing.
  • You can store an array of data using the Vec data type.

Helpful doc links:

Part 2 - Implement loading patterns from files

Goal: get more comfortable with Rust by implementing a basic file format parser.

It's common to share Game of Life patterns via run length encoded text strings, so let's implement a parser which loads these files into the simulation.

  • Lines starting with # are comments and so can be ignored.
  • The first line of content is a header of the form x = m, y = n to indicate the dimensions of the pattern (this is sometimes suffixed by , rule = abc to support other cellular automata but you can ignore that).
  • Subsequent content lines are formatted as repetitions of ab (possibly separated by whitespace) where a is the number of times that b repeats and b is an indicator of either alive cells (o), dead cells (b), or start-new-line ($).
  • The content always ends with ! as an end-of-file separator.

For example content of

bo$2b
o$3o!

gives (with O indicating live cells and whitespace indicating dead cells):

 O
  O
OOO

Here's a sample of a full file format:

#C This is a glider.
x = 3, y = 3
bo$2bo$3o!

This pattern file is at patterns/gosper-glider-gun.txt, and there are more patterns in the same directory.

More info on the file format is at http://www.conwaylife.com/w/index.php?title=Run_Length_Encoded

You can also find far more patterns at http://www.conwaylife.com/wiki/Category:Patterns

Tip: You will probably get into error handling with Some (Rust's version of returning null) and Result (Rust's alternative to exceptions) types. The quick and dirty way to get at the contents of each is to call unwrap() or expect() (which will panic your app if there is no content); read the book's error handling section for more info.

Part 3 - Performance tuning

Goal: learn about performance tuning by competing to see who can build the fastest implementation of the game of life update step.

To see benchmarks powered by the Criterion benchmarking library, run:

cargo bench

The results for your implementation won't be automatically included; add it over in benches/benchmark.rs. Then try to beat the solution's speed! It shouldn't be too hard :)

NB: If you don't want to use Criterion's benchmarks, note that Rust by default builds & runs unoptimized debug code - pass the --release flag to Cargo for a 10-100x speedup.

Tips:

  • On Linux, perf and Valgrind's cachegrind are good profiling tools, and I hear MacOS has something similar in the form of dtrace. Perhaps the flamegraph crate might be worth using on Linux and MacOS? But you're on your own for Windows.
  • Parallelism is easiest to do via Rust's support.
  • If you need to communicate between threads, use channels (mpsc::channel or the chan crate) or std::sync's primitives (like Mutex).
  • A vector (Vec) of booleans is fairly efficient but a bit vector might be more efficient.
  • You might want to look into Rust's SIMD support if you have experience with that.