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.
For installation instructions, please refer to the documentation.
Create a new benchmark function in your Zig code. This function takes a single argument of type std.mem.Allocator and runs 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());
}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: Setbefore_all,after_all,before_each, andafter_eachhooks to function pointers.track_allocations: Boolean to enable or disable tracking memory allocations during the benchmark.
Zig is in active development, and its APIs can change frequently. The main branch of this project now targets the latest Zig master build to take advantage of new features and improvements. For users who prefer the stability of official releases, dedicated branches are maintained for older Zig versions (e.g., zig-0.13.0, zig-0.12.0, etc.). This ensures you can choose the branch that best fits your stability and feature requirements.
It's important to acknowledge that a no-op time of ca. 15 ns (or more) is expected and is not an issue with zBench itself (see also #77). This does not reflect an inefficiency in the benchmarking process.
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 ยฑ ฯ) (min ... max) p75 p99 p995
---------------------------------------------------------------------------------------------------------------
benchmarkMyFunction 1000 1200ms ยฑ 10ms (100ms ... 2000ms) 1100ms 1900ms 1950msThis 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.
You can build all examples with the following command:
zig build examplesExecutables can then be found in ./zig-out/bin by default.
- If Zig doesn't detect changes in a dependency, clear the project's
zig-cachefolder and~/.cache/zig. - Non-ASCII characters not printed correctly on Windows
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.
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.
zBench is MIT licensed.