Delay bound analysis for heterogeneous multicore systems using network calculus
Meng Li, Guchuan Zhu, Yvon Savaria · 2018
In heterogeneous multicore systems, a real challenging problem is to provide a timing performance guarantee for applications that have stringent timing constraints. One challenge is that task execution time varies. A dominant solution to this issue is to use the worst-case execution time as an upper bound. Another challenge arises from the communication delays due to task dependencies, which highly depend on task scheduling schemes. There lacks a formal approach to provide a rigorous analysis of transmission delays for heterogeneous multicore systems. In this paper, a mathematical tool, namely Network Calculus, is employed for end-to-end delay analysis. Under this framework, a virtual channel concept is first introduced for communication between CPU cores. Then a flow regulation model is proposed for network performance analysis. An upper bound of communication delays in a heterogeneous multicore system is established. A case study is provided to demonstrate the proposed approach. Based on a formal analysis, mathematical upper bounds of transmission delays are obtained to ensure the system performance.