On-the-fly detection of access anomalies for parallel programs helps you find bugs as they happen, with practical, scalable monitoring. This guide explains how to spot non-deterministic access to shared data while a program runs, using a general model that covers fork-join and synchronization primitives. It shows how to keep memory use reasonable and still catch real problems as they appear.
Readers will gain a clear picture of the algorithm, its storage strategies, and how it improves over post-mortem tracing. The text walks through merging and subtraction techniques, formal definitions, and step-by-step reasoning for real-time anomaly detection in parallel code.
- How on-the-fly detection works and what problems it targets
- Two compression approaches: merging shared-variable sets and subtracting unused data
- How the method handles complex patterns like fork-join and inter-thread coordination
- Practical considerations for storage needs, performance, and applicability to real programs
Ideal for readers of parallel debugging and runtime analysis tools seeking a practical monitoring approach for shared-memory programs.