A Signal-Based Dependency-Aware Execution Model for Task Scheduling in Multicore Systems

Modern multicore systems rely on operating system schedulers to allocate processor time among concurrently running threads. While these schedulers are effective at balancing workload priorities and preserving fairness, they typically lack visibility into logical task dependencies at the application level. As a result, threads may be prematurely dispatched and enter a passive waiting state, occupying CPU resources unnecessarily and degrading overall system efficiency. The proposed study addresses this limitation by introducing a dependency-aware execution model that integrates with existing schedulers without modifying kernel behavior. The model intercepts task dispatch requests and ensures that execution occurs only when all declared dependencies have been satisfied. This is achieved through a reactive, signal-driven mechanism that dynamically evaluates task readiness using externally managed dependency vectors. The framework supports both static and dynamically arriving workloads and is agnostic to the underlying operating system, making it adaptable to a wide range of environments. The approach is theoretically justified by analyzing thread behavior and dispatch logic in Linux and Windows systems, with emphasis on minimizing idle resource contention. The model aims to improve processor utilization in complex scheduling scenarios by preventing wasteful thread activation and enhancing execution order fidelity in the presence of inter-task constraints.