Analysis and Optimization of Real-Time Applications Running on Heterogeneous Hardware

Abstract

This is an early stage proposal of a methodology that can be applied to the data-flow analysis of an application where software can be decomposed in task DAGs, thus it answers in part to the proposed industrial challenge. The methodology builds on two main aspects: (1) using an ARINC-like scheduler, i.e., the partitioning concept can provide applications with strong temporal and space isolation (in addition, fixed priorities are allowed inside a partition at a second scheduling level), and (2) an existing modelling and schedulability analysis technique for distributed multipath flows (DAGs). This technology can be applied to multicore processors (shared bus for global memory plus core-local memory) where the worst-case execution time (WCET) of tasks can be measured or estimated through worst-case assumptions on the memory contention representing a bounded impact in the response times.