Practical Multicore Extension of Functionally and Temporally Correct Real-Time Simulation for Automotive Systems

Abstract

Existing simulation methods cannot provide functionally and temporally correct simulations for the cyber-side of automotive systems since they do not correctly model temporal behaviours such as varying execution times and task preemptions. To address such limitations, our previous work proposes a novel simulation technique that guarantees the functional and temporal simulation correctness. However, the simulation technique is designed assuming a single core simulator. In this work, we extend the single core simulator targeting a multicore simulator to enhance the simulation capacity. In this multicore extension, a major challenge is the inter-core interferences in a multicore environment, which causes unpredictability of simulated job execution times, which in turn makes it hard to model the timings of the real cyber-side of an automotive system. To overcome the challenge, this paper empirically analyzes the inter-core interferences for typical automotive workloads and proposes a practical multicore extension approach, which can still provide a functionally and temporally correct simulation, without using complex inter-core isolation mechanisms. Our experimental study shows that the proposed multicore extension approach can significantly improve the simulation capacity over the previous single core simulator while still preserving simulation correctness.