Expand description
diol
is a benchmarking library for rust.
§getting started
add the following to your Cargo.toml
.
[dev-dependencies]
diol = "0.6.0"
[[bench]]
name = "my_benchmark"
harness = false
then in benches/my_benchmark.rs
.
use diol::prelude::*;
fn main() -> std::io::Result<()> {
let mut bench = Bench::new(BenchConfig::from_args());
bench.register(slice_times_two, [4, 8, 16, 128, 1024]);
bench.run()?;
Ok(())
}
fn slice_times_two(bencher: Bencher, len: usize) {
let mut v = vec![0.0_f64; len];
bencher.bench(|| {
for x in &mut v {
*x *= 2.0;
}
black_box(&mut v);
});
}
run the benchmark with cargo bench
, or cargo bench --bench my_benchmark
if you have multiple
benchmarks you can also pass in benchmark options using cargo bench --bench my_benchmark -- [OPTIONS...]
╭─────────────────┬──────┬───────────┬───────────┬───────────┬───────────╮
│ benchmark │ args │ fastest │ median │ mean │ stddev │
├─────────────────┼──────┼───────────┼───────────┼───────────┼───────────┤
│ slice_times_two │ 4 │ 29.61 ns │ 34.38 ns │ 34.83 ns │ 1.62 ns │
├─────────────────┼──────┼───────────┼───────────┼───────────┼───────────┤
│ slice_times_two │ 8 │ 44.17 ns │ 53.04 ns │ 53.32 ns │ 3.25 ns │
├─────────────────┼──────┼───────────┼───────────┼───────────┼───────────┤
│ slice_times_two │ 16 │ 93.66 ns │ 107.91 ns │ 108.13 ns │ 4.11 ns │
├─────────────────┼──────┼───────────┼───────────┼───────────┼───────────┤
│ slice_times_two │ 128 │ 489.97 ns │ 583.59 ns │ 585.28 ns │ 33.15 ns │
├─────────────────┼──────┼───────────┼───────────┼───────────┼───────────┤
│ slice_times_two │ 1024 │ 3.77 µs │ 4.51 µs │ 4.53 µs │ 173.44 ns │
╰─────────────────┴──────┴───────────┴───────────┴───────────┴───────────╯
Modules§
- benchmark configuration
- re-exports of useful types and traits.
- benchmark result output.
- helper traits and types
- variadic tuple type.
Macros§
- type of a variadic tuple containing the given types.
- create or destructure a variadic tuple containing the given values.
Structs§
- main benchmark entry point, used to register functions and arguments, then run benchmarks.
- bench loop runner.
- extra-precise duration type for short-ish durations.
- argument type that marks the benchmark functions as plottable.