adler32fast

Fast, SIMD-accelerated Adler-32 checksum computation

Usage

use adler32fast::Adler32;

let mut adler32 = Adler32::new();
adler32.update(b"foo bar baz");
let checksum = adler32.as_u32();

Performance

This crate contains multiple Adler-32 implementations:

• A fast baseline implementation which processes up to 16 bytes per iteration
• An optimized implementation for modern x86/x86_64 using either avx2 or ssse3 instructions

Calling the Adler32::new or Adler32::from constructor at runtime will perform a feature detection to select the most optimal implementation for the current CPU feature set.

The following was tested on a Razer Book 13 RZ09-0357, i7-1165G7 @ 2.80 GHz with 100kb of data.

Benchmarks using criterion can be run on stable Rust with cargo bench.

Contributions are welcomed for more SIMD variants!

Memory Safety

Due to the use of SIMD intrinsics for the optimized implementations, this crate contains some amount of unsafe code.

adler32fast is fuzz-tested with cargo-fuzz. (NOTE: cargo-fuzz does not work on Windows, but can run under WSL.)

cargo-fuzz currently requires the nightly toolchain.

$ cargo install cargo-fuzz
$ rustup toolchain install nightly

$ cargo +nightly fuzz run ssse3

Credits

This work is based on crc32fast as inspiration.

The AVX2 implementation was derived from simd-adler32

The SSSE3 implementation has been derived from Google's Wuffs implementation.

License

Licensed under either of

• Apache License, Version 2.0 (LICENSE-APACHE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE-MIT or http://opensource.org/licenses/MIT)

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.