A BPF-Based Task Scheduling Scheme for Heterogeneous Multiprocessor Systems

Abstract

With the booming development in the IoT field, more and more embedded systems are required to support both real-time task processing and rich functionality. The need for Heterogeneous Multiprocessor Systems-on-Chip (HMPSoC) and task scheduling between operating systems on HMPSoC has arisen. This paper discusses a load balancing scheme, BPF-based Task Scheduling Scheme (BTSS), based on BPF technology and inter-core communication mechanisms, to transfer some computational tasks from the General Purpose Operating System (GPOS) to the Real Time Operating System (RTOS) for execution. A prototype of BTSS is implemented to verify the feasibility and evaluate the performance of the scheme, using Linux as the GPOS and an IoT operating system as the RTOS. The evaluation results demonstrate that with the BTSS, we can utilize the processor resources of HMPSoC more efficiently and increase the system computing capability.