Struct RuntimeConfig

Source

pub struct RuntimeConfig {Show 22 fields
    pub worker_threads: usize,
    pub thread_stack_size: usize,
    pub thread_name_prefix: String,
    pub global_queue_limit: usize,
    pub steal_batch_size: usize,
    pub blocking: BlockingPoolConfig,
    pub enable_parking: bool,
    pub poll_budget: u32,
    pub cancel_lane_max_streak: usize,
    pub logical_clock_mode: Option<LogicalClockMode>,
    pub root_region_limits: Option<RegionLimits>,
    pub on_thread_start: Option<Arc<dyn Fn() + Send + Sync>>,
    pub on_thread_stop: Option<Arc<dyn Fn() + Send + Sync>>,
    pub deadline_monitor: Option<MonitorConfig>,
    pub deadline_warning_handler: Option<Arc<dyn Fn(DeadlineWarning) + Send + Sync>>,
    pub metrics_provider: Arc<dyn MetricsProvider>,
    pub observability: Option<ObservabilityConfig>,
    pub cancel_attribution: CancelAttributionConfig,
    pub obligation_leak_response: ObligationLeakResponse,
    pub leak_escalation: Option<LeakEscalation>,
    pub enable_governor: bool,
    pub governor_interval: u32,
}

Expand description

Runtime configuration.

Fields§

§worker_threads: usize

Number of worker threads (default: available parallelism).

§thread_stack_size: usize

Stack size per worker thread (default: 2MB).

§thread_name_prefix: String

Name prefix for worker threads.

§global_queue_limit: usize

Global queue size limit (0 = unbounded).

§steal_batch_size: usize

Work stealing batch size.

§blocking: BlockingPoolConfig

Blocking pool configuration.

§enable_parking: bool

Enable parking for idle workers.

§poll_budget: u32

Time slice for cooperative yielding (polls).

§cancel_lane_max_streak: usize

Maximum consecutive cancel-lane dispatches before yielding to other lanes.

§logical_clock_mode: Option<LogicalClockMode>

Logical clock mode used for trace causal ordering.

When None, the runtime chooses a default:

No reactor: Lamport (deterministic lab-friendly)
With reactor: Hybrid (wall-clock + logical)

§root_region_limits: Option<RegionLimits>

Admission limits applied to the root region (if set).

§on_thread_start: Option<Arc<dyn Fn() + Send + Sync>>

Callback executed when a worker thread starts.

§on_thread_stop: Option<Arc<dyn Fn() + Send + Sync>>

Callback executed when a worker thread stops.

§deadline_monitor: Option<MonitorConfig>

Deadline monitoring configuration (when enabled).

§deadline_warning_handler: Option<Arc<dyn Fn(DeadlineWarning) + Send + Sync>>

Warning callback for deadline monitoring.

§metrics_provider: Arc<dyn MetricsProvider>

Metrics provider for runtime instrumentation.

§observability: Option<ObservabilityConfig>

Optional runtime observability configuration.

§cancel_attribution: CancelAttributionConfig

Limits for cancellation attribution cause chains.

Used to bound memory growth when cancellation cascades across deep region trees or large cancellation graphs.

§obligation_leak_response: ObligationLeakResponse

Response policy for obligation leaks detected at runtime.

§leak_escalation: Option<LeakEscalation>

Optional escalation policy for obligation leaks.

When set, the runtime escalates from obligation_leak_response to escalation.escalate_to after escalation.threshold leaks.

§enable_governor: bool

Enable the Lyapunov governor for scheduling suggestions.

When enabled, the scheduler periodically snapshots runtime state and consults the governor for lane-ordering hints. When disabled (default), scheduling behavior is identical to the ungoverned baseline.

§governor_interval: u32

Number of scheduling steps between governor snapshots (default: 32).

Lower values increase responsiveness but add snapshot overhead. Only relevant when enable_governor is true.