Adaptive Spotted Hyena Optimizer for Latency-Aware Task Scheduling in Heterogeneous Multicore Systems

Heterogeneous multicore architectures encompassing Central Processing Units (CPUs), Graphics Processing Units (GPUs), and Field-Programmable Gate Arrays (FPGAs) have emerged as the dominant computational paradigm for high-performance and embedded workloads. Task scheduling across such architectures presents a formidable challenge: the assignment of computational tasks to heterogeneous processing elements must satisfy precedence constraints whilst minimising overall workflow completion latency. Classical scheduling heuristics, including Earliest Finish Time (EFT) and Heterogeneous Earliest Finish Time (HEFT), offer polynomial-time approximations yet demonstrate limited adaptability under dynamic runtime conditions, frequently yielding suboptimal execution latency in large-scale task graphs. This paper proposes the Latency-Aware Adaptive Spotted Hyena Optimizer (LA-ASHO), a novel metaheuristic scheduling framework grounded in the social hunting behaviour of spotted hyenas. The framework encodes task-to-core assignments as discrete permutation vectors and evaluates fitness exclusively through a mathematically rigorous execution latency model that integrates computation time, memory overhead, and inter-core communication delays. Comprehensive simulation experiments conducted over 100 to 1,000 tasks across 8 to 64 heterogeneous cores, each repeated across 30 statistically independent runs demonstrate that LA-ASHO achieves statistically significant reductions in workflow completion latency relative to established baseline schedulers such as; Min-Min, Heterogeneous Earliest Finish Time (HEFT).  The principal contribution of this work is the formulation of an execution-latency-focused metaheuristic framework that is both theoretically grounded and practically scalable for real-world heterogeneous computing deployments.