9 releases (5 breaking)

0.6.1 Feb 27, 2025
0.6.0 Feb 27, 2025
0.5.0 Jan 30, 2025
0.4.1 Jan 30, 2025
0.1.0 Dec 19, 2024

#124 in Compression

Download history 97/week @ 2024-12-18 294/week @ 2024-12-25 51/week @ 2025-01-01 119/week @ 2025-01-08 7/week @ 2025-01-15 254/week @ 2025-01-29 22/week @ 2025-02-05 7/week @ 2025-02-19 299/week @ 2025-02-26

338 downloads per month

MIT/Apache

2MB
57K SLoC

C++ 49K SLoC // 0.0% comments Rust 4K SLoC // 0.0% comments C 2.5K SLoC // 0.1% comments Visual Studio Project 846 SLoC Visual Studio Solution 65 SLoC Just 64 SLoC // 0.3% comments Shell 21 SLoC

FastPFor for Rust

GitHub crates.io version docs.rs docs license CI build

This is a Rust wrapper for the C++ FastPFor library, as well as a pure Rust re-implementation (work in progress). Supports 32-bit and 64-bit integers, and SIMD-optimized codecs for 128-bit and 256-bit vectors. Based on the Decoding billions of integers per second through vectorization, 2012 paper.

Supported algorithms

Unless otherwise specified, all codecs support &[u32] only.

  • BP32
  • Copy
  • FastBinaryPacking8
  • FastPFor128 (both &[u32] and &[u64])
  • FastPFor256 (both &[u32] and &[u64])
  • FastBinaryPacking16
  • FastBinaryPacking32
  • MaskedVByte
  • NewPFor
  • OptPFor
  • PFor2008
  • PFor
  • SimdBinaryPacking
  • SimdFastPFor128
  • SimdFastPFor256
  • SimdGroupSimple
  • SimdGroupSimpleRingBuf
  • SimdNewPFor
  • SimdOptPFor
  • SimdPFor
  • SimdSimplePFor
  • Simple16
  • Simple8b
  • Simple8bRle
  • Simple9
  • Simple9Rle
  • SimplePFor
  • StreamVByte
  • VByte
  • VarInt (both &[u32] and &[u64])
  • VarIntGb

Usage

Crate Features

  • cpp - C++ implementation (default)
  • rust - Rust implementation (work in progress, opt-in)

Using C++ Wrapper

use fastpfor::cpp::{Codec32 as _, SimdFastPFor128Codec};

fn main() {
  let mut codec = SimdFastPFor128Codec::new();

  // Encode
  let mut input = vec![1, 2, 3, 4, 5];
  let mut output = vec![0; 10];  // must be large enough
  let enc_slice = codec.encode32(&input, &mut output).unwrap();

  // Decode
  let mut decoded = vec![0; 10]; // must be large enough
  let dec_slice = codec.decode32(&enc_slice, &mut decoded).unwrap();

  assert_eq!(input, dec_slice);
}

Build Requirements

When using the Rust implementation, no additional dependencies are required. When using the C++ implementation, you need to have a C++ compiler that supports C++14 and SIMD intrinsics. See FastPFor C++ requirements.

Linux

The default GitHub action runner for Linux has all the needed dependencies. For local development, you may need to install the following packages.

# This list may be incomplete
sudo apt-get install build-essential libsimde-dev

macOS

To build FastPFor on macOS, you'll need to install SIMDe. Since Homebrew installs packages in /opt/homebrew (for Apple Silicon), you'll also need to explicitly set the include paths.

# install SIMDe via Homebrew
brew install simde
# Ensure the compiler can find the required headers before building
export CXXFLAGS="-I/opt/homebrew/include"
export CFLAGS="-I/opt/homebrew/include"

License

Licensed under either of

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.

Dependencies

~0–0.9MB