Software applications in automotive systems are comprised of communicating real-time tasks, described by cause-effect chains. To guarantee functional correctness, it is mandatory to verify end-to-end timing latencies of the cause-effect chains. The analysis of end-to-end latencies highly depends on the communication method. Implicit communication is standardized in the AUTOSAR Timing Specification and ensures data consistency. To abstract communication from the actual execution behavior of tasks, logical execution time (LET) has been proposed. However, the determinism that is provided by LET comes at the cost of increased end-to-end latencies. In industry-grade systems, periodic and sporadic tasks using LET and implicit communication co-exist. Hence, end-to-end latency analyses should cover such heterogeneous cause-effect chains.
In this work, we present the first formal analysis framework for end-to-end analysis of cause-effect chains that allows heterogeneous types of recurrent tasks and different communication mechanisms, i.e., (i) a mixed setup of sporadic and periodic tasks that (ii) communicate by a mixed setup of LET and implicit communication mechanism. In this regard, we uncover the principles that homogeneous analyses are built from and discuss how these principles can be transferred to the heterogeneous case. In particular, we cut the cause-effect chain into homogeneous parts which results in 3 different analyses: one baseline approach, one that directly uses the homogeneous results, and one that reduces the pessimism for changes of communication means. Our evaluation shows that for some systems the two more sophisticated approaches outperform the baseline significantly, while for other systems the baseline is already satisfactory.