Module guff::good

A set of “good” optimised routines in native Rust

This module provides a set of alternative implementations of GaloisField that are faster than the reference implementation. This is provided for those who are interested in having a reasonable set of fairly fast routines, but who don’t particularly care about how they’re implemented.

This module is intended to provide a reasonable balance between:

• speed;
• memory footprint;
• range of field sizes; and
• supported polynomials

The actual implementation used here may change over time. In fact, some of the fields may even default to using the reference implementation for some calculations. It can be assumed, however, that at least the mul routines implemented by objects here will be faster than the corresponding reference implementation, since this is usually the method that will be called most often in an application.

SemVer implications

Once I have added a constructor here, it will continue to be available across all minor versions of the library. In other words, in terms of semantic versioning, removing a constructor here counts as breaking the application interface.

However, no guarantees are made as to how the optimised version works internally. In particular, no guarantee of speed is made. Nor should the contents of the struct implementing the method be considered part of the application interface. That can change at any time.

Benchmarking

I’ve implemented some basic benchmarks using Criterion. After downloading the source for this project, simply run:

     cargo bench

Tests marked ref are run using the default, reference implementations, while those marked with good are for fields constructed using the functions below.

Functions

new_gf4_0x13

Optimised GF(2⁴) with the (primitive) polynomial 0x13

new_gf8_0x11b

Optimised GF(2⁸) with the (non-primitive) polynomial 0x11b

new_gf16_0x1002b

Optimised GF(2¹⁶) with the (primitive?) polynomial 0x1002b