Timing analysis of process graphs with finite communication buffers
Chung-Wei Lin, Marco Di Natale, Haibo Zeng, Linli Thi Xuan Phan, Alberto Luigi Sangiovanni-Vincentelli · 2013
Real-Time Calculus (RTC) is a modular performance analysis framework for real-time embedded systems. It can be used to compute the worst-case and best-case response times of tasks with general activation patterns and configurations, such as pipelines of tasks that are connected via finite buffers. In this paper, we extend the existing RTC framework to analyze arbitrary graph configurations of tasks and messages, with mixed periodic and event-based activation models and finite buffers between any pair of nodes. Our extension also improves upon several sources of pessimism in the existing analysis. We present an application of the extended RTC to the Loosely Time-Triggered Architecture (LTTA) implementation of synchronous models, commonly used in the development of embedded automotive, avionics and control systems. We show how our method can be used to model scheduling and communication delays in an LTTA mapping, which gives tighter analysis bounds on the output rate and the latency compared to existing techniques. The evaluation on automotive workloads shows that our approach is scalable and outperforms existing techniques in terms of analysis accuracy.