Getting Started #

Prerequisites #

• Zig 0.16.0
• Git submodules initialized if you want direct Tokio/Go comparisons:

git submodule update --init --recursive

• Rust and Cargo for the Tokio comparison binary
• A bootstrap Go installation in PATH for the first vendored-Go comparison run

Build The Library #

make build

Or with Zig directly:

zig build build

Run The Example #

zig build example

The example in examples/basic.zig shows the smallest useful flow: create a pool, submit tasks with a wait group, wait for completion, then read a stats snapshot.

Minimal Usage #

const std = @import("std");
const thread_pool = @import("thread_pool");

fn task(ctx: *anyopaque, runtime: *thread_pool.TaskContext) void {
    _ = ctx;
    _ = runtime;
}

pub fn main() !void {
    var gpa = std.heap.GeneralPurposeAllocator(.{}){};
    defer _ = gpa.deinit();

    var pool = try thread_pool.ThreadPool.init(gpa.allocator(), .{
        .min_threads = 2,
        .max_threads = 2,
        .queue_capacity = 128,
        .local_queue_capacity = 32,
        .scheduler_variant = .work_stealing,
    });
    defer pool.deinit();

    var wg = pool.waitGroup();
    _ = try pool.submit(task, undefined, .{ .wait_group = &wg });
    wg.wait();
}

Choose A Scheduler #

• global_queue: simplest behavior and easiest baseline to reason about
• work_stealing: preferred for mixed workloads and current performance work
• lane_extension: extends the work-stealing model with a dedicated blocking lane

Next Reads #

• API Reference
• Examples
• Testing
• Runtime Comparison