Automatic Struct of Array generation for Rust

This crate provides a custom derive (#[derive(StructOfArray)]) to automatically generate code from a given struct T that allow to replace Vec<T> with a struct of arrays. For example, the following code

#[derive(StructOfArray)]
pub struct Particle {
    pub x: f64,
    pub y: f64,
    pub z: f64,
}

will generate a ParticleVec struct that looks like this:

pub struct ParticleVec {
    pub x: Vec<f64>,
    pub y: Vec<f64>,
    pub z: Vec<f64>,
}

It will also generate the same functions that a Vec<Particle> would have, and a few helper structs: ParticleSlice, ParticleSliceMut, ParticleRef and ParticleRefMut.

How to use it

Add this to your Cargo.toml

[dependencies]
soa_derive = "0.1"

This to your src/lib.rs

#[macro_use]
extern crate soa_derive;

And then add #[derive(StructOfArray)] to each struct you want to derive a struct of array version. If you need the helper structs to derive additional traits (such as Debug or PartialEq), you can add an attribute #[soa_derive = "Debug, PartialEq"] to the struct declaration.

#[derive(Debug, PartialEq, StructOfArray)]
#[soa_derive = "Debug, PartialEq"]
pub struct Particle {
    pub x: f64,
    pub y: f64,
    pub z: f64,
}

Caveats and limitations

Vec<T> functionalities rely a lot on references and automatic deref feature, for getting function from [T] and indexing. But the SoA vector (let's call it CheeseVec, generated from the Cheese struct) generated by this crate can not implement Deref<Target=CheeseSlice>, because Deref is required to return a reference, and CheeseSlice is not a reference. The same applies to Index and IndexMut trait, that can not return CheeseRef/CheeseRefMut.

This means that the we can not index into a CheeseVec, and that a few functions are duplicated, or require a call to as_ref()/as_mut() to change the type used.

Documentation

Please see http://lumol.org/soa-derive/soa_derive_example/ for a small example and the documentation of all the generated code.

Benchmarks

Here are a few simple benchmarks results, on my machine:

running 10 tests
test aos_big_do_work_1000    ... bench:         997 ns/iter (+/- 192)
test aos_big_do_work_10000   ... bench:      21,324 ns/iter (+/- 3,282)
test aos_big_push            ... bench:          93 ns/iter (+/- 17)
test aos_small_do_work_10000 ... bench:       8,822 ns/iter (+/- 1,459)
test aos_small_push          ... bench:          10 ns/iter (+/- 4)
test soa_big_do_work_1000    ... bench:         890 ns/iter (+/- 142)
test soa_big_do_work_10000   ... bench:      10,538 ns/iter (+/- 1,621)
test soa_big_push            ... bench:         171 ns/iter (+/- 44)
test soa_small_do_work_10000 ... bench:       8,978 ns/iter (+/- 1,538)
test soa_small_push          ... bench:          24 ns/iter (+/- 6)

Benchmarks tests exist for soa (struct of array) and aos (array of struct) versions of the same code, using a samll and a big struct.

You can run the same benchmarks on your own system by cloning this repository and running cargo bench.

Licensing and contributions

This crate distributed under either the MIT or the Apache license, at your choice. Contributions are welcome, please open an issue before to discuss your changes !