⚡ zBench - A Zig Benchmarking Library

zBench is a benchmarking library for the Zig programming language. It is designed to provide easy-to-use functionality to measure and compare the performance of your code.

Content

• Installation
• Usage
• Configuration
  • Compatibility Notes
  • Benchmark Functions
  • Reporting Benchmarks
  • Running zBench Examples
  • Troubleshooting
• Contributing
• License

Installation

For installation instructions, please refer to the documentation.

Usage

Create a new benchmark function in your Zig code. This function should take a single argument of type std.mem.Allocator. The function would run the code you wish to benchmark.

fn benchmarkMyFunction(allocator: std.mem.Allocator) void {
    // Code to benchmark here
}

You can then run your benchmarks in a test:

test "bench test" {
    var bench = zbench.Benchmark.init(std.testing.allocator, .{});
    defer bench.deinit();
    try bench.add("My Benchmark", myBenchmark, .{});
    try bench.run(std.io.getStdOut().writer());
}

Configuration

To customize your benchmark runs, zBench provides a Config struct that allows you to specify several options:

pub const Config = struct {
    iterations: u16 = 0,
    max_iterations: u16 = 16384,
    time_budget_ns: u64 = 2e9, // 2 seconds
    hooks: Hooks = .{},
    track_allocations: bool = false, 
};

• iterations: The number of iterations the benchmark has been run. This field is usually managed by zBench itself.
• max_iterations: Set the maximum number of iterations for a benchmark. Useful for controlling long-running benchmarks.
• time_budget_ns: Define a time budget for the benchmark in nanoseconds. Helps in limiting the total execution time of the benchmark.
• hooks: Set before_all, after_all, before_each, and after_each hooks to function pointers.
• track_allocations: Boolean to enable or disable tracking memory allocations during the benchmark.

Compatibility Notes

Zig is in active development and the APIs can change frequently, making it challenging to support every dev build. This project currently aims to be compatible with stable, non-development builds to provide a consistent experience for the users.

Supported Version: As of now, zBench is tested and supported on Zig version 0.12.0.

Benchmark Functions

Benchmark functions have the following signature:

fn(allocator: std.mem.Allocator) void

The function body contains the code you wish to benchmark.

Reporting Benchmarks

zBench provides a comprehensive report for each benchmark run. It includes the total operations performed, the average, min, and max durations of operations, and the percentile distribution (p75, p99, p995) of operation durations.

benchmark              runs     time (avg ± σ)         (min ... max)                p75        p99        p995
---------------------------------------------------------------------------------------------------------------
benchmarkMyFunction    1000     1200ms ± 10ms          (100ms ... 2000ms)           1100ms     1900ms     1950ms

This example report indicates that the benchmark "benchmarkMyFunction" ran with an average of 1200 ms per execution and a standard deviation of 10 ms. The minimum and maximum execution times were 100 ms and 2000 ms, respectively. The 75th, 99th and 99.5th percentiles of execution times were 1100 ms, 1900 ms, and 1950 ms, respectively.

Running zBench Examples

You can run all example tests with the following command:

zig build test_examples

Troubleshooting

• If Zig doesn't detect changes in a dependency, clear the project's zig-cache folder and ~/.cache/zig.
• Non-ASCII characters not printed correctly on Windows

Contributing

The main purpose of this repository is to continue to evolve zBench, making it faster and more efficient. We are grateful to the community for contributing bugfixes and improvements. Read below to learn how you help improve zBench.

Contributing Guide

Read our contributing guide to learn about our development process, how to propose bugfixes and improvements, and how to build and test your changes to zBench.

License

zBench is MIT licensed.