Struct criterion::Criterion

pub struct Criterion { /* fields omitted */ }

The benchmark manager

Criterion lets you configure and execute benchmarks

Each benchmark consists of four phases:

• Warm-up: The routine is repeatedly executed, to let the CPU/OS/JIT/interpreter adapt to the new load
• Measurement: The routine is repeatedly executed, and timing information is collected into a sample
• Analysis: The sample is analyzed and distiled into meaningful statistics that get reported to stdout, stored in files, and plotted
• Comparison: The current sample is compared with the sample obtained in the previous benchmark. If a significant regression in performance is spotted, Criterion will trigger a task panic

Methods

impl Criterion

fn sample_size(&mut self, n: usize) -> &mut Criterion

Changes the size of the sample

A bigger sample should yield more accurate results, if paired with a "sufficiently" large measurement time, on the other hand, it also increases the analysis time

Panics

Panics if set to zero

fn warm_up_time(&mut self, dur: Duration) -> &mut Criterion

Changes the warm up time

Panics

Panics if the input duration is zero

fn measurement_time(&mut self, dur: Duration) -> &mut Criterion

Changes the measurement time

With a longer time, the measurement will become more resilient to transitory peak loads caused by external programs

Note: If the measurement time is too "low", Criterion will automatically increase it

Panics

Panics if the input duration in zero

fn nresamples(&mut self, n: usize) -> &mut Criterion

Changes the number of resamples

Number of resamples to use for the bootstrap

A larger number of resamples reduces the random sampling errors, which are inherent to the bootstrap method, but also increases the analysis time

Panics

Panics if the number of resamples is set to zero

fn noise_threshold(&mut self, threshold: f64) -> &mut Criterion

Changes the noise threshold

This threshold is used to decide if an increase of X% in the execution time is considered significant or should be flagged as noise

Note: A value of 0.02 is equivalent to 2%

Panics

Panics is the threshold is set to a negative value

fn confidence_level(&mut self, cl: f64) -> &mut Criterion

Changes the confidence level

The confidence level is used to calculate the confidence intervals of the estimated statistics

Panics

Panics if the confidence level is set to a value outside the (0, 1) range

fn significance_level(&mut self, sl: f64) -> &mut Criterion

Changes the significance level

The significance level is used for hypothesis testing

Panics

Panics if the significance level is set to a value outside the (0, 1) range

fn with_plots(&mut self) -> &mut Criterion

Enables plotting

fn without_plots(&mut self) -> &mut Criterion

Disabled plotting

fn can_plot(&self) -> bool

Checks if plotting is possible

fn bench_function<F>(&mut self, id: &str, f: F) -> &mut Criterion where
    F: FnMut(&mut Bencher), 

Benchmarks a function

The function under test must follow the setup - bench - teardown pattern:

use self::criterion::{Bencher, Criterion};

fn routine(b: &mut Bencher) {
    // Setup (construct data, allocate memory, etc)

    b.iter(|| {
        // Code to benchmark goes here
    })

    // Teardown (free resources)
}

Criterion::default().bench_function("routine", routine);

fn bench_functions<I>(
    &mut self,
    id: &str,
    funs: Vec<Fun<I>>,
    input: &I
) -> &mut Criterion where
    I: Display, 

Benchmarks multiple functions

All functions get the same input and are compared with the other implementations. Works similar to bench, but with multiple functions.

fn bench_seq_fib(b: &mut Bencher, i: &u32) {
    b.iter(|| {
        seq_fib(i);
    });
}

fn bench_par_fib(b: &mut Bencher, i: &u32) {
    b.iter(|| {
        par_fib(i);
    });
}

let sequential_fib = Fun::new("Sequential", bench_seq_fib);
let parallel_fib = Fun::new("Parallel", bench_par_fib);
let funs = vec![sequential_fib, parallel_fib];

Criterion::default().bench_functions("Fibonacci", funs, &14);

fn bench_function_over_inputs<I, F>(
    &mut self,
    id: &str,
    f: F,
    inputs: I
) -> &mut Criterion where
    I: IntoIterator,
    I::Item: Display,
    F: FnMut(&mut Bencher, &I::Item), 

Benchmarks a function under various inputs

This is a convenience method to execute several related benchmarks. Each benchmark will receive the id: ${id}/${input}.

use self::criterion::{Bencher, Criterion};

Criterion::default()
    .bench_function_over_inputs("from_elem", |b: &mut Bencher, &&size: &&usize| {
        b.iter(|| vec![0u8; size]);
    }, &[1024, 2048, 4096]);

fn bench_program(&mut self, id: &str, program: Command) -> &mut Criterion

Benchmarks an external program

The external program must conform to the following specification:

fn main() {
    let stdin = io::stdin();
    let ref mut stdin = stdin.lock();

    // For each line in stdin
    for line in stdin.lines() {
        // Parse line as the number of iterations
        let iters: u64 = line.unwrap().trim().parse().unwrap();

        // Setup

        // Benchmark
        let start = Instant::now();
        // Execute the routine "iters" times
        for _ in 0..iters {
            // Code to benchmark goes here
        }
        let elapsed = start.elapsed();

        // Teardown

        // Report elapsed time in nanoseconds to stdout
        println!("{}", elapsed.to_nanos());
    }
}

fn bench_program_over_inputs<I, F>(
    &mut self,
    id: &str,
    program: F,
    inputs: I
) -> &mut Criterion where
    F: FnMut() -> Command,
    I: IntoIterator,
    I::Item: Display, 

Benchmarks an external program under various inputs

This is a convenience method to execute several related benchmarks. Each benchmark will receive the id: ${id}/${input}.

fn summarize(&mut self, id: &str) -> &mut Criterion

Summarize the results stored under the .criterion/${id} folder

Note: The bench_with_inputs and bench_program_with_inputs functions internally call the summarize method

Trait Implementations

impl Default for Criterion

fn default() -> Criterion

Creates a benchmark manager with the following default settings:

• Sample size: 100 measurements
• Warm-up time: 1 s
• Measurement time: 1 s
• Bootstrap size: 100 000 resamples
• Noise threshold: 0.01 (1%)
• Confidence level: 0.95
• Significance level: 0.05
• Plotting: enabled (if gnuplot is available)