Precise Modelling of Instruction Cache Behaviour

Safety critical real-time applications in aviation, automotive and industrial automation have to guarantee not only the functionality, but also the timeliness of the results. Here, a deadline is associated with the software tasks, and a failure to complete prior to this deadline could lead to a catastrophic consequences. Hence, for the correctness of real-time systems, it is essential to be able to compute the worst case execution time (WCET) of the tasks in order to guarantee their deadlines. However, the problem of WCET analysing is difficult, because of processors use cache-based memory systems that vary memory access time significantly. Any pessimistic estimation of the number of cache hits/misses will result in loose precision of the WCET analyses, which could lead to over use of hardware resources. In this paper, we present a new approach for statically analysing the behaviour of instructions on a direct mapped cache. The proposed approach combines binary representation and a new abstraction that reduces the analysis time without sacrificing the precision. This is unlike the existing cache analysing approaches where either precision or scalability (analysis time) is sacrificed. Experimental results are presented that demonstrate the practical applicability of this analysis