Approximation-Aware Task Deployment on Heterogeneous Multicore Platforms With DVFS

Heterogeneous (HE) multicore platforms, such as ARM big.LITTLE, are widely used to execute embedded applications under multiple and contradictory constraints, energy consumption real-time (RT) execution. To fulfill these constraints optimize system performance, application tasks should be efficiently mapped on platforms. Embedded usually tolerant approximated results but acceptable quality of service (QoS). Modeling-embedded by using the elastic task model, namely, imprecise computation (IC) can balance QoS, consumption, RT performance during deployment. However, state-of-the-art approaches seldom consider problem IC deployment HE They typically neglect migration, which improve solutions due its flexibility process. This article proposes a novel QoS-aware method maximize QoS where frequency assignment (FA), allocation (TA), scheduling, migration optimized simultaneously. The is formulated mixed-integer nonlinear programming. Then, it linearized linear programming find optimal (OPT) solution. Furthermore, based structure decomposition, we propose heuristic (HEU) with low computational complexity. subproblems regarding FA, TA, adjustment considered solved in sequence. Finally, simulation show that proposed improves 31.2% average (up 112.8%) compared methods designed HEU achieves about 53.9% (on average) OPT solution negligible computing time.