#reed-solomon #erasure


Rust implementation of Reed-Solomon erasure coding

19 stable releases (4 major)

4.0.2 Apr 6, 2020
4.0.1 Sep 25, 2019
3.1.1 Oct 28, 2018
3.1.0 May 20, 2018
0.9.1 Dec 17, 2017

#29 in Encoding

Download history 1385/week @ 2021-06-06 1615/week @ 2021-06-13 2086/week @ 2021-06-20 1856/week @ 2021-06-27 2156/week @ 2021-07-04 2635/week @ 2021-07-11 1654/week @ 2021-07-18 2684/week @ 2021-07-25 2493/week @ 2021-08-01 2705/week @ 2021-08-08 2754/week @ 2021-08-15 2341/week @ 2021-08-22 2467/week @ 2021-08-29 2657/week @ 2021-09-05 3186/week @ 2021-09-12 3595/week @ 2021-09-19

10,116 downloads per month
Used in 35 crates (10 directly)

MIT license



Build Status Build status codecov Coverage Status Crates Documentation dependency status

Rust implementation of Reed-Solomon erasure coding

WASM builds are also available, see section WASM usage below for details

This is a port of BackBlaze's Java implementation, Klaus Post's Go implementation, and Nicolas Trangez's Haskell implementation.

Version 1.X.X copies BackBlaze's implementation, and is less performant as there were fewer places where parallelism could be added.

Version >= 2.0.0 copies Klaus Post's implementation. The SIMD C code is copied from Nicolas Trangez's implementation with minor modifications.

See Notes and License section for details.

WASM usage

See here for details

Rust usage

Add the following to your Cargo.toml for the normal version (pure Rust version)

reed-solomon-erasure = "4.0"

or the following for the version which tries to utilise SIMD

reed-solomon-erasure = { version = "4.0", features = "simd-accel" }

and the following to your crate root

extern crate reed_solomon_erasure;


extern crate reed_solomon_erasure;

use reed_solomon_erasure::galois_8::ReedSolomon;
// or use the following for Galois 2^16 backend
// use reed_solomon_erasure::galois_16::ReedSolomon;

fn main () {
    let r = ReedSolomon::new(3, 2).unwrap(); // 3 data shards, 2 parity shards

    let mut master_copy = shards!(
        [0, 1,  2,  3],
        [4, 5,  6,  7],
        [8, 9, 10, 11],
        [0, 0,  0,  0], // last 2 rows are parity hards
        [0, 0,  0,  0]

    // Construct the parity shards
    r.encode(&mut master_copy).unwrap();

    // Make a copy and transform it into option shards arrangement
    // for feeding into reconstruct_shards
    let mut shards: Vec<_> = master_copy.iter().cloned().map(Some).collect();

    // We can remove up to 2 shards, which may be data or parity shards
    shards[0] = None;
    shards[4] = None;

    // Try to reconstruct missing shards
    r.reconstruct(&mut shards).unwrap();

    // Convert back to normal shard arrangement
    let result: Vec<_> = shards.into_iter().filter_map(|x| x).collect();

    assert_eq!(master_copy, result);

Benchmark it yourself

You can test performance under different configurations quickly (e.g. data parity shards ratio, parallel parameters) by cloning this repo: https://github.com/darrenldl/rse-benchmark

rse-benchmark contains a copy of this library (usually a fully functional dev version), so you only need to adjust main.rs then do cargo run --release to start the benchmark.


Version 1.X.X, 2.0.0 do not utilise SIMD.

Version 2.1.0 onward uses Nicolas's C files for SIMD operations.

Machine: laptop with Intel(R) Core(TM) i5-3337U CPU @ 1.80GHz (max 2.70GHz) 2 Cores 4 Threads

Below shows the result of one of the test configurations, other configurations show similar results in terms of ratio.

Configuration Klaus Post's >= 2.1.0 && < 4.0.0 2.0.X 1.X.X
10x2x1M ~7800MB/s ~4500MB/s ~1000MB/s ~240MB/s

Versions >= 4.0.0 have not been benchmarked thoroughly yet




Contributions are welcome. Note that by submitting contributions, you agree to license your work under the same license used by this project as stated in the LICENSE file.


Library overhaul and Galois 2^16 backend

Many thanks to the following people for overhaul of the library and introduction of Galois 2^16 backend

WASM builds

Many thanks to Nazar Mokrynskyi @nazar-pc for submitting his package for WASM builds

He is the original author of the files stored in wasm folder. The files may have been modified by me later.

AVX512 support

Many thanks to @sakridge for adding support for AVX512 (see PR #69)

build.rs improvements

Many thanks to @ryoqun for improving the usability of the library in the context of cross-compilation (see PR #75)


Many thanks to the following people for testing and benchmarking on various platforms

  • Laurențiu Nicola @lnicola (platforms: Linux, Intel)

  • Roger Andersen @hexjelly (platforms: Windows, AMD)


Code quality review

If you'd like to evaluate the quality of this library, you may find audit comments helpful.

Simply search for "AUDIT" to see the dev notes that are aimed at facilitating code reviews.

Implementation notes

The 1.X.X implementation mostly copies BackBlaze's Java implementation.

2.0.0 onward mostly copies Klaus Post's Go implementation, and copies C files from Nicolas Trangez's Haskell implementation.

The test suite for all versions copies Klaus Post's Go implementation as basis.


Nicolas Trangez's Haskell Reed-Solomon implementation

The C files for SIMD operations are copied (with no/minor modifications) from Nicolas Trangez's Haskell implementation, and are under the same MIT License as used by NicolasT's project


All files are released under the MIT License