A Design-Time/Run-Time Application Mapping Methodology for Predictable Execution Time in MPSoCs

摘要

Executing multiple applications on a single MPSoC brings the major challengeof satisfying multiple quality requirements regarding real-time, energy, etc.Hybrid application mapping denotes the combination of design-time analysis withrun-time application mapping. In this article, we present such a methodology,which comprises a design space exploration coupled with a formal performanceanalysis. This results in several resource reservation configurations,optimized for multiple objectives, with verified real-time guarantees for eachindividual application. The Pareto-optimal configurations are handed over torun-time management which searches for a suitable mapping according to thisinformation. To provide any real-time guarantees, the performance analysisneeds to be composable and the influence of the applications on each other hasto be bounded. We achieve this either by spatial or a novel temporal isolationfor tasks and by exploiting composable NoCs. With the proposed temporalisolation, tasks of different applications can be mapped to the same resourcewhile with spatial isolation, one computing resource can be exclusively used byonly one application. The experiments reveal that the success rate in findingfeasible application mappings can be increased by the proposed temporalisolation by up to 30% and energy consumption can be reduced compared tospatial isolation.