Explicit Preemption Placement For Real-Time Conditional Code Via Graph Grammars And Dynamic Programming

Traditional worst-case execution time (WCET) analysis must make very pessimistic assumptions regarding the cost of preemptions for a real-time job. For every potential preemption point, the analysis must add to the WCET of a job the cache-related preemption delay (CRPD) incurred due to the contention for memory resources with other jobs in the system. However, recent work has shown that CRPD can vary at each preemption point (due to the cache lines that must be reloaded for subsequent code after the preemption). Using this observation and information obtained from schedulability analysis on the maximum length of the non-preemptive region of a job, we seek to find the optimal set of explicit preemption-points (EPPs) that minimize the WCET and ensure system schedulability. Utilizing graph grammars and dynamic programming, we develop a polynomial-time algorithm that is capable of analyzing jobs that can be represented by control flowgraphs with arbitrarily-nested conditional structures. This algorithm extends previous work that could only handle sequential flowgraphs. Finally, we show via a case study that our approach potentially decreases the WCET over the previous algorithm and has low runtime.