Optimal and Energy Efficient Imprecise Computation Task Deployment on Heterogeneous Multicore Platforms Combining DVFS and DPM
TL;DR
A novel QoS-aware task deployment methodology to enhance the Quality of Service (QoS) under resource limitations is introduced and results demonstrate that the proposed method achieves superior system performance compared to existing approaches.
Abstract
Heterogeneous multicore embedded systems can handle complex tasks such as data collection, processing, and wireless communication. However, these systems often face constraints due to limited computational resources, requiring efficient task deployment strategies to manage dependent and real-time tasks. Despite advancements, existing methods rarely address the deployment of Imprecise Computation (IC) tasks on heterogeneous multicore platforms that utilize both Dynamic Voltage/Frequency Scaling (DVFS) and Dynamic Power Management (DPM). This paper introduces a novel QoS-aware task deployment methodology to enhance the Quality of Service (QoS) under resource limitations. The proposed approach simultaneously optimizes frequency assignment, processor state transitions, task allocation, scheduling, and task execution cycle adjustments to maximize system QoS while adhering to time, energy, and dependency constraints. The problem is initially formulated as a Mixed Integer Non-Linear Programming (MINLP) model, then transformed into a Mixed Integer Linear Programming (MILP) formulation to obtain optimal solutions. We developed a heuristic method with low computational complexity to address scalability challenges, employing problem decomposition and a greedy algorithm. The process operates in two stages: task allocation and frequency assignment, followed by task scheduling and cycle adjustment. The simulation results demonstrate that the proposed method achieves superior system performance compared to existing approaches, particularly in QoS ( \(41.30\% \) ), task schedulability ( \(58.67\% \) ), and execution efficiency.