Timing Implications of Sharing Resources in Multicore Real-Time Automotive Systems 2010-01-0454
The topic of timing has already been recognized as a major challenge when designing safety-critical automotive architectures. Consequently the availability of appropriate performance and timing analysis methods is key to building reliable automotive electric and electronics (E/E) and software architectures. Due to the potential performance increase, power reduction and cost-efficiency multicore solutions for automotive real-time environments receive growing attention. But the prediction of the timing behavior for multicore electronic control unit (ECU) systems becomes more complicated. Even in setups with static task-to-processor mapping, the execution of the tasks is usually not independent. The use of the same physical hardware, such as memories, coprocessors, or network components, makes inter-core interference unavoidable and may introduce hard-to-find timing problems including missed deadlines that can finally make the entire system fail. With this work we anticipate new issues on the timing prediction in the upcoming automotive multicore ECUs, we discuss difficulties and solutions in the evolution from distributed single-core systems to multicore systems and also provide a modeling and analysis approach of complex real-time multicore systems with shared resources.