Parallel Execution Infrastructure

include/Utils/Parallel/ and lib/Utils/Parallel/ provide the shared thread-pool and scheduling utilities used by parallel analyses.

Main components:

• ThreadPool for task execution and parallel loops.

• CancellationToken and CancellationSource for cooperative stop.

• PipelineScheduler and ParallelSchedulerPass for structured pipelines.

• Task and thread-local reducer helpers for work coordination.

Use this layer when an analysis needs bounded parallelism without embedding its own scheduler.

Concurrency guidance

Tasks should communicate cancellation through the provided token rather than abandoning shared work abruptly. Make ownership and mutation boundaries explicit: LLVM IR objects and analysis caches are not automatically safe for concurrent writes. Use reducer helpers for accumulations, then merge results after parallel work completes to preserve deterministic reporting where it matters.

See also Utility Libraries.