OSAMIC

Abstract

Many embedded real-time systems have dynamic computation workloads to interact with physical processes. Combining imprecise computation and run-time mode change provides both flexible and effective computation outcomes. However, it requires complex schedulability analysis to guarantee its robustness. In this paper, we study the workload and online schedulability analysis for realtime workload for safety critical applications on heterogeneous multi-core platforms. We extend the traditional schedulability analysis and develop a new analysis for the multi-mode systems, called Online Schedulability Analysis of Real-Time Mode Change on Heterogeneous Multi-Core Platforms (OSAMIC). By generalizing the deadline based schedulability analysis, we developed an online sufficient schedulability analysis to reduce the time complexity. Two algorithms are developed to compute the offset to minimize the delay for CPU and GPU workloads. The evaluation results show that the proposed algorithm can shorten the offset up to 82.27% for preemptive workloads and to 339 ms when the task utilization is 0.5 for non-preemptive workloads.