zoop

A benchmarking library for zig.

• Uses built in monotonic, high performance timer.
• Warms up to remove function call overhead time.
• Supports lifecycle hooks like beforeAll, afterEach, etc.
• Can export raw benchmark data to JSON.
• Vibrant terminal output with easy to read data.
• Highly customizable.

Installation

Zig master version is required to use zoop.

To install zoop in your own project,

1. Add the dependency to the build.zig.zon of your project.

   .dependencies = .{
       .zoop = .{
           .url = "https://github.com/tr1ckydev/zoop/archive/0ce2e985206e28fe13ed3c6dcb6b88650688238e.tar.gz",
           .hash = "1220bad2ffae51552957e6255289dd452ee1a5e687d9f4c8c9427008edb0a802d874",
       },
   },
   

2. Add the dependency and module to your build.zig.

   const zoop_dep = b.dependency("zoop", .{});
   const zoop_mod = zoop_dep.module("zoop");
   exe.addModule("zoop", zoop_mod);
   

3. Import it inside your project.

   const Benchmark = @import("zoop").Benchmark;
   

or, if you just want to try out the benchmark from the above image, clone this repository and run zig build run.

Documentation

CallbackFn refers to the type *const fn () anyerror!void.

• Benchmark.init(allocator, config)

  Create a new Benchmark instance with the provided config options. (Provide .{} to use the default configuration.)

  var bench = Benchmark.init(allocator, .{});
  

  • allocator(Allocator): The allocator to use inside the benchmark.
  • config(struct): Configuration for the benchmark.
    • show_cpu_name(bool): If true, shows the name of host device cpu. (Default is true)
    • show_zig_version(bool): If true, shows the version of zig currently used to run the benchmark. (Default is true)
    • show_summary(bool): If true, shows the summary after the entire benchmark is finished, i.e. the fastest test and comparison with other tests. (Default is true)
    • show_summary_comparison(bool): If true, shows the comparison of the fastest test with the other tests else summary only displays the fastest test. (Default is true)
    • show_output(bool): If false, no output is printed to stdout. (Default is true)
    • enable_warmup(bool): If true, measures a noop function to calculate function call overhead which will be subtracted from results for more accurate data. (Default is true)
    • iterations(u16): The maximum number of iterations to perform if the budget expires. (Default is 10)
    • budget(u64): The maximum time (in nanoseconds) allotted to measure if the maximum iterations expire. (Default is 2 secs)
    • hooks(struct): Hooks to execute during the benchmark lifecycle. (Defaults are noop)
      • beforeAll(CallbackFn): Runs once before starting the benchmark.
      • afterAll(CallbackFn): Runs once after finishing the benchmark.
      • beforeEach(CallbackFn): Runs before measuring each test.
      • afterEach(CallbackFn): Runs after measuring each test.
    • export_json(?[]const u8): Exports the raw benchmark data in nanoseconds to the provided JSON file path. (Default is null)

• Benchmark.deinit()

  Release all allocated memory.

  defer bench.deinit();
  

• Benchmark.add(name: []const u8, function: CallbackFn)

  Add a function to the test suite.

  try bench.add("my function", myFunction);
  // ...
  // ...
  fn myFunction() !void {
      // some code here to be measured
  }
  

• Benchmark.run()

  After adding all the tests, start the benchmark. (After finishing, results are sorted from fastest to slowest.)

  try bench.run();
  

Some of the internal functions are exposed for user's convenience.

• getCpuName(allocator)

  Returns the name of the host device cpu, such as 12th Gen Intel(R) Core(TM) i7-12700H, Apple M1 Max, etc.

• fmtIntu32(u32_int)

  Returns a formatter which can convert u32 integers to short forms like 26.312k, 2.906M.

If you have any questions, feel free to join the discord server here.

Credits

zoop is inspired from hyperfine and mitata.

License

This repository uses MIT License. See LICENSE for full license text.