#bench #benchmark #simple

no-std benchmark-simple

A tiny, super simple and portable benchmarking library

8 releases

0.1.7 Oct 20, 2021
0.1.6 Oct 20, 2021

#76 in No standard library

23 downloads per month
Used in 3 crates

MIT license

13KB
238 lines

benchmark-simple

A tiny benchmarking library for Rust.

  • Trivial to use
  • Works pretty much everywhere, including WebAssembly (WASI, but also in-browser)
use benchmark_simple::*;

fn test_function() {
    // ...
}

let bench = Bench::new();
let options = Options::default();
let res = bench.run(&options, || test_function());
println!("result: {}", res);

Throughput computation:

use benchmark_simple::*;

fn test_function(m: &mut [u8]) {
    // ...
}

let mut m = vec![0u8; 1_000_000];
let bench = Bench::new();
let options = Options::default();
let res = bench.run(&options, || test_function(&mut m));
let throughput = res.throughput(m.len() as _);
println!("throughput: {}", throughput);

Options:

pub struct Options {
    /// Number of iterations to perform.
    pub iterations: u64,
    /// Number of warm-up iterations to perform.
    pub warmup_iterations: u64,
    /// Minimum number of samples to collect.
    pub min_samples: usize,
    /// Maximum number of samples to collect.
    pub max_samples: usize,
    /// Maximum RSD to tolerate (in 0...100).
    pub max_rsd: f64,
    /// Maximum benchmark duration time.
    pub max_duration: Option<std::time::Duration>,
    /// Verbose output
    pub verbose: bool,
}

Benchmark results can be made verbose by setting verbose to true in the Options struct, or by defining a BENCHMARK_VERBOSE environment variable.

Dependencies

~225KB