A Realistic Simulation Environment for Evaluating Dependency-Aware Scheduling in Multicore Systems

Efficient task scheduling in multicore systems becomes increasingly complex when inter-task dependencies are introduced. Existing operating system schedulers prioritize fairness and responsiveness but lack mechanisms for evaluating dependency relationships during task dispatching. This leads to premature thread activation and resource inefficiencies, especially in dependency-rich workloads. While some solutions address this limitation through compiler-level or middleware-based strategies, they often assume full knowledge of the task structure in advance, making them impractical for dynamic or real-time environments. To address this gap, a configurable simulation environment has been developed to evaluate and compare task scheduling models in scenarios with evolving dependencies. The study focuses on a lightweight user-space execution strategy known as the Dependency-Aware Model (DAM), which uses a signal-based mechanism to defer task execution until all prerequisites have been met. The simulator supports configurable workload generation, real CPU-bound task execution, and thread-level scheduling compatible with both Windows and Linux systems. Execution metrics are collected in CSV format and visualized using matplotlib charts. Preliminary results demonstrate that the DAM approach significantly reduces idle processor occupation and improves total execution time in cases with dense dependency graphs. The environment offers a reproducible platform for testing advanced scheduling techniques and lays the foundation for future integration with adaptive AI/ML-based task management strategies.