Algorithms for dependable hard real-time systems

Hard real-time code is special purpose code whose worst-case performance (worst-case execution time WCET) needs to be good and easy to predict. Despite these specific demands on the temporal properties of hard real-time code, real-time programmers often use the same algorithms and programming techniques that have proven effective for nonreal-time applications. This paper explains the different temporal requirements imposed on real-time resp. nonreal-time code, and outlines why traditional (nonreal-time) programming tends to produce code that (a) has a high WCET and (b) is hard to analyze for its WCET. Based on these observations, the paper proposes an unconventional programming strategy that avoids the shortcoming of traditional coding and yields code that is well suited for hard real-time system, i.e., its WCET is short and predictable. The evaluation of a number of examples demonstrates the advantages of the proposed programming strategy on WCET and its predictability.