rust-benchmarks

This repository presents different Rust benchmarks in an A/B format, showing which one is faster. Each benchmark reports execution time and includes an explanation of the underlying reasons on which approach is better.

The examples follow these principles:

• Presents two approaches (A/B) that solve the same problem.
• The examples are small and targeting a very specific functionality.
• Being faster doesn't always mean it's the right choice, each use case has its needs.
• Code is kept minimal, no extra boilerplate, just the benchmark.
• One file, one benchmark. Each file contains two functions to be benchmaked.
• Avoid files longer than 100 lines.
• The code of each file is self contained, no need to set up a project. See cargo install rust-script.
• Benchmarks aim to address real problems when possible, but some are for purely educational purposes.
• Each code file should include documentation explaining why one benchmark is slower or faster than the other.
• Benchmarks are run with criterion. Use black_box to prevent the compiler from optimizing away computations.
• Both Rust core language features and external crates can be benchmarked.
• Preferably shall not require a lot of domian specific knowledge of the problem.
• Some benchmarks may depend on the compiler and optimization level being used.

run

First, install the rust-script crate. It allows you to embed cargo manifests within the same .rs file, enabling a single self-contained code with everything you need.

cargo install rust-script

And now you can run any of the benchmarks as follows. Replace the name by any file in the benchmarks folder.

rust-script ./benchmarks/branching_vs_xor.rs

• TODO: Measure memory footprint.
• TODO: Measure CPU cycles.