Cause-Effect Reaction Latency in Real-Time Systems

Abstract

In embedded real-time systems, a functionality is often implemented as a dataflow chain over a set of communicating tasks. An important requirement in such systems is to restrict the amount of time an input data requires to impact its corresponding output. Such temporal requirements over dataflow chains also known as the end-to-end latency constraints, are well-studied in the context of lock-based blocking inter-task communication. However, lock-based communication does not preserve the functional semantics and complicates latency calculation due to its reliance on response times of the communicating tasks. We propose to use non-blocking inter-task communications to preserve the functional semantics. Unfortunately a naive method to compute the reaction latency by adding worst-case delays between each write-read pair is unsafe for systems with non-blocking communication. In this paper, we study a non-blocking communication protocol. We present an algorithm to compute the exact worst-case delay in a cause-effect chain, which provides a safe estimation of the worst-case cause-effect latency for systems using this protocol for non-blocking communication.