You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Define repeatable performance baselines for channel work against semantically similar Rust implementations.
Scope
Cover ready paths and concurrent bounded and unbounded workloads for these shapes:
one producer and one consumer;
many producers and one consumer;
one producer and many competing consumers;
balanced many-producer, many-consumer contention;
broadcast fanout and producer contention when broadcast exists;
latest-state notification when watch exists.
Use comparable semantics and document exceptions such as Tokio broadcast's overwrite and lag behavior. Measure multiple participant counts so wake-all or contention cliffs are visible.
Acceptance criteria
The benchmark methodology and peer versions are documented.
Results are reproducible locally through the repository workflow or a documented companion workflow.
Sustained regressions above 3x are investigated, and order-of-magnitude regressions block an implementation from being accepted unless the semantic tradeoff is explicit.
Third-party benchmark dependencies do not become runtime dependencies of asyncband.
Activity