If you or your company find Divan valuable, consider sponsoring on GitHub or donating via PayPal. Sponsorships help me progress on what's possible with benchmarking in Rust.
A guide is being worked on. In the meantime, see:
Divan 0.1.21 requires Rust 1.80.0 or later.
-
Add the following to your project's
Cargo.toml:[dev-dependencies] divan = "0.1.21" [[bench]] name = "example" harness = false
-
Create a benchmarks file at
benches/example.rs1 with your benchmarking code:fn main() { // Run registered benchmarks. divan::main(); } // Register a `fibonacci` function and benchmark it over multiple cases. #[divan::bench(args = [1, 2, 4, 8, 16, 32])] fn fibonacci(n: u64) -> u64 { if n <= 1 { 1 } else { fibonacci(n - 2) + fibonacci(n - 1) } }
-
Run your benchmarks with
cargo bench:example fastest │ slowest │ median │ mean │ samples │ iters ╰─ fibonacci │ │ │ │ │ ├─ 1 0.626 ns │ 1.735 ns │ 0.657 ns │ 0.672 ns │ 100 │ 819200 ├─ 2 2.767 ns │ 3.154 ns │ 2.788 ns │ 2.851 ns │ 100 │ 204800 ├─ 4 6.816 ns │ 7.671 ns │ 7.061 ns │ 7.167 ns │ 100 │ 102400 ├─ 8 57.31 ns │ 62.51 ns │ 57.96 ns │ 58.55 ns │ 100 │ 12800 ├─ 16 2.874 µs │ 3.812 µs │ 2.916 µs │ 3.006 µs │ 100 │ 200 ╰─ 32 6.267 ms │ 6.954 ms │ 6.283 ms │ 6.344 ms │ 100 │ 100
See #[divan::bench] for info on benchmark function registration.
If you want to run each benchmarked function only once (to check that they don't panic without measuring their performance), you can run the benchmark runner crate with the --test parameter. You can do so by running either
cargo bench -- --testor
cargo test --benchesPractical example benchmarks can be found in the examples/benches
directory. These can be benchmarked locally by running:
git clone https://github.com/nvzqz/divan.git
cd divan
cargo bench -q -p examples --all-featuresMore thorough usage examples can be found in the #[divan::bench] documentation.
Like the Rust project, this library may be used under either the MIT License or Apache License (Version 2.0).
Footnotes
-
Within your crate directory, i.e.
$CARGO_MANIFEST_DIR↩