Beyond the traditional analyses and resource management in real-time systems
In recent years the increasing computational power demand of applications leads to the evolution of computing platforms towards complex processor architectures and sophisticated system components. To overcome the single-core performance barrier, today’s processor manufactures have indeed introduced several advanced features like multi-/many-core, complex pipelines, multi-level caches, memory prefetcher, and many others. Unfortunately, this makes the problem of computing the WCET with traditional static timing analyses extremely difficult, thus limiting the use of modern architectures in some classes of embedded systems, e.g. in safety-critical systems. To cope with the traditional WCET static analysis problems, probabilistic real-time has been proposed in the first years of 2000s to provide a distribution of the WCET instead of a single scalar value. The topics of my PhD thesis are mainly focused on the methodology of probabilistic-WCET analyses and, in particular, on how to increase their reliability. In fact, even if several steps have been done in the recent years, we are still very far to obtain a probabilistic-WCET value usable for certification purposes.
Back to Alumni