ILP_c: A novel approach for scalable timing analysis of synchronous programs

Abstract

Synchronous programs have been widely used in the design of safety critical systems such as the flight control of Airbus A-380. To validate the implementations of synchronous programs, it is necessary to map the program's logical time (measured in logical ticks) to physical time (the execution time on a given processor). The static computation of the worstcase execution time of logical ticks is called Worst Case Reaction Time (WCRT) analysis. Several approaches for WCRT analysis exist: max-plus algebra, model checking, reachability and integer linear programming (ILP). Of these approaches, reachability, model checking and ILP provide reasonably precise worst case estimates at the expense of longer analysis time. Apart from max-plus based approaches, which can produce large overestimates, the existing approaches suffer from the state space explosion problem. In this paper, we develop a new ILP based approach, called ILP c , which exploits the concurrency explicitly in the ILP formulation to avoid the state space explosion problem. Through extensive benchmarking we demonstrate the efficacy of the approach: for complex programs, ILP c is often orders of magnitude faster compared to the existing approaches, while achieving same level of precision. Thus, this paper paves the way for scalable WCRT analysis of complex embedded systems designed using the synchronous approach.

Model-Driven Design Using IEC 61499

“…Unlike general-purpose programs, computing the worst-case execution time for synchronous programs is much simpler and has been demonstrated to be feasible even for large programs [69,101,133], making this approach suitable for real-time systems.…”

Towards Interactive Timing Analysis for Designing Reactive Systems

“…Raymond et al [32] propose an approach to improve WCET analysis by checking feasibility of computed longest paths using model checking at the modeling level. Want et al propose an ILP-based approach that exploits concurrency explicitly in the ILP formulation to avoid state space explosion [33]. These techniques could be combined with our proposal regarding WCET-feedback at the modeling level.…”

Time for Reactive System Modeling

“…Also, we trace the representation of model elements for a graphical control-flow based language, whose compilation involves a chain of model transformation steps down to code generation and use the concept of timing program points to identify the corresponding code parts. Wang et al propose an ILP-based approach that exploits concurrency explicitly [27]. These techniques could be combined with our proposal regarding WCET-feedback at the modeling level.…”