5 releases

0.8.6 Sep 27, 2022
0.8.5 Apr 12, 2022
0.8.4 Mar 16, 2022
0.8.2 Mar 8, 2021
0.8.0-beta.5 Nov 18, 2020

#11 in Algorithms

Download history 10032/week @ 2022-08-10 9894/week @ 2022-08-17 10406/week @ 2022-08-24 10683/week @ 2022-08-31 13642/week @ 2022-09-07 11280/week @ 2022-09-14 13472/week @ 2022-09-21 13687/week @ 2022-09-28 15729/week @ 2022-10-05 15942/week @ 2022-10-12 16943/week @ 2022-10-19 19180/week @ 2022-10-26 18053/week @ 2022-11-02 16644/week @ 2022-11-09 17635/week @ 2022-11-16 13763/week @ 2022-11-23

69,199 downloads per month
Used in 108 crates (37 directly)

BSL-1.0 license

90KB
2K SLoC

xxhash-rust

Rust Crates.io Documentation

Implementation of xxHash in Rust

Each algorithm is implemented via feature, allowing precise control over code size.

Example

  • Cargo.toml
[dependencies.xxhash-rust]
version = "0.8.5"
features = ["xxh3", "const_xxh3"]
  • main.rs
use xxhash_rust::const_xxh3::xxh3_64 as const_xxh3;
use xxhash_rust::xxh3::xxh3_64;

const TEST: u64 = const_xxh3(b"TEST");

fn test_input(text: &str) -> bool {
    match xxh3_64(text.as_bytes()) {
        TEST => true,
        _ => false
    }
}

assert!(!test_input("tEST"));
assert!(test_input("TEST"));

Features:

By default all features are off.

  • xxh32 - Enables 32bit algorithm. Suitable for x86 targets
  • const_xxh32 - const fn version of xxh32 algorithm
  • xxh64 - Enables 64 algorithm. Suitable for x86_64 targets
  • const_xxh64 - const fn version of xxh64 algorithm
  • xxh3 - Enables xxh3 family of algorithms, superior to xxh32 and xxh64 in terms of performance.
  • const_xxh3 - const fn version of xxh3 algorithm

HW acceleration

Similar to reference implementation, crate implements various SIMDs in xxh3 depending on provided flags. All checks are performed only at compile time, hence user is encouraged to enable these accelerations (for example via -C target_cpu=native)

Used SIMD acceleration:

  • SSE2 - widely available, can be safely enabled in 99% of cases. Enabled by default in x86_64 targets.
  • AVX2;

Streaming vs One-shot

For performance reasons one-shot version of algorithm does not re-use streaming version. Unless needed, user is advised to use one-shot version which tends to be more optimal.

cosnt fn version

While const fn provides compile time implementation, it does so at performance cost. Hence you should only use it at compile time.

To guarantee that something is computed at compile time make sure to initialize hash output as const or static variable, otherwise it is possible function is executed at runtime, which would be worse than regular algorithm.

const fn is implemented in best possible way while conforming to limitations of Rust const fn, but these limitations are quite strict making any high performance code impossible.

Version note

  • 0.8.* corresponds to C's 0.8.*

In order to keep up with original implementation version I'm not planning to bump major/minor until C implementation does so.

Comparison with twox-hash

Refer to my comment

No runtime deps