Module parallel 

Parallel execution utilities with work-stealing scheduler.

This module provides advanced parallel execution infrastructure:

• Work-stealing scheduler for dynamic load balancing
• NUMA-aware memory allocation for multi-socket systems
• Thread pool management with configurable worker counts
• Task dependencies and execution ordering
• Load balancing metrics and monitoring

Example

use tensorlogic_infer::{WorkStealingScheduler, ParallelConfig, Task};

// Create a work-stealing scheduler
let config = ParallelConfig::default()
    .with_num_workers(8)
    .with_steal_strategy(StealStrategy::Random);

let scheduler = WorkStealingScheduler::new(config);

// Submit tasks
for task in tasks {
    scheduler.submit(task)?;
}

// Execute in parallel with work stealing
let results = scheduler.execute_all()?;

// Check load balancing stats
let stats = scheduler.stats();
println!("Steal count: {}", stats.steal_count);
println!("Load balance: {:.2}%", stats.load_balance_ratio * 100.0);

Structs

LoadBalanceStats

Load balancing statistics.

NumaNode

NUMA node identifier.

ParallelConfig

Parallel execution configuration.

SchedulerStats

Scheduler statistics.

Task

A parallel task.

WorkStealingScheduler

Work-stealing scheduler.

Enums

NumaStrategy

NUMA allocation strategy.

ParallelError

Parallel execution errors.

StealStrategy

Work-stealing strategy.

TaskPriority

Task priority level.

Type Aliases

TaskId

Task identifier.