Crate criterion

A statistics-driven micro-benchmarking library written in Rust.

This crate is a microbenchmarking library which aims to provide strong statistical confidence in detecting and estimating the size of performance improvements and regressions, while also being easy to use.

See the user guide for examples as well as details on the measurement and analysis process, and the output.

Features:

• Collects detailed statistics, providing strong confidence that changes to performance are real, not measurement noise.
• Produces detailed charts, providing thorough understanding of your code’s performance behavior.

Feature flags

Modules

async_executor

This module defines a trait that can be used to plug in different Futures executors into Criterion.rs’ async benchmarking support.

measurement

This module defines a set of traits that can be used to plug different measurements (eg. Unix’s Processor Time, CPU or GPU performance counters, etc.) into Criterion.rs. It also includes the WallTime struct which defines the default wall-clock time measurement.

profiler

This module provides an extension trait which allows in-process profilers to be hooked into the --profile-time argument at compile-time. Users of out-of-process profilers such as perf don’t need to do anything special.

Macros

criterion_groupNon-codspeed

Macro used to define a function group for the benchmark harness; see the criterion_main! macro for more details.

criterion_mainNon-codspeed

Macro which expands to a benchmark harness.

Structs

AsyncBencherasync and non-codspeed

Async/await variant of the Bencher struct.

BencherNon-codspeed

Timer struct used to iterate a benchmarked function and measure the runtime.

BenchmarkGroupNon-codspeed

Structure used to group together a set of related benchmarks, along with custom configuration settings for groups of benchmarks. All benchmarks performed using a benchmark group will be grouped together in the final report.

BenchmarkIdNon-codspeed

Simple structure representing an ID for a benchmark. The ID must be unique within a benchmark group.

CriterionNon-codspeed

The benchmark manager

Enums

AxisScale

Axis scaling type

Baseline

Baseline describes how the baseline_directory is handled.

BatchSize

Argument to Bencher::iter_batched and Bencher::iter_batched_ref which controls the batch size.

BenchmarkFilter

Benchmark filtering support.

SamplingMode

This enum allows the user to control how Criterion.rs chooses the iteration count when sampling. The default is Auto, which will choose a method automatically based on the iteration time during the warm-up phase.

Throughput

Enum representing different ways of measuring the throughput of benchmarked code. If the throughput setting is configured for a benchmark then the estimated throughput will be reported as well as the time per iteration.

Functions

black_box

Reexport of std::hint::black_box.