Zigrupt

Zigrupt is a high-performance event broadcasting library for Zig, heavily inspired by the original LMAX Disruptor. It lets application threads publish events once and have multiple handlers process every event, in order, on their own threads. A fixed-capacity ring buffer, lock-free coordination, and attention to CPU cache lines keep the path between producers and handlers lean.

Why broadcast?

A work queue assigns an event to one worker. zigrupt gives each registered handler the full stream. For example, one handler can update live metrics while another writes an audit trail. They work independently and may finish at different times, but neither misses an event because the other received it.

How an event moves

A producer publishes a value into the next reusable slot in the ring. Each publication has a sequence number, so handlers can follow the stream in order even after the ring wraps around. Every handler tracks which events it has processed and acknowledges its progress separately.

The ring reuses a slot only after all handlers have finished with its previous event. If the slowest handler falls a full ring behind, producers wait for capacity instead of overwriting unread data. A successful publication means the event is available to handlers; it does not wait for them to finish.

Built for busy event streams

The ring allocates a bounded set of slots up front and reuses them as events flow through. Atomic, lock-free coordination avoids a mutex on the publication path, while cache-line-aware layout reduces false sharing between threads. Choose single- or multi-producer mode, tune how handlers wait for new events, and process events in batches to match the workload and available CPU time.

Start here

• Install and run your first program.
• Understand broadcast and backpressure.
• Choose producer mode and runtime configuration.
• Look up the public API.
• See the measured performance comparison.
• Run the canonical examples.

The project uses the MIT license.