PEP: Precise and Efficient Prediction of Good Worst-case Performance for Contemporary and Future Architectures

• Applicants: Professor Dr.-Ing. Jan Reineke; Professor Dr. Reinhard Wilhelm
• Subject Area: Computer Architecture, Embedded and Massively Parallel Systems
• Term: from 2016 to 2019
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 289264719

Sound timing analysis methods for the verification of real-time requirements are routinely used in the safety-critical embedded-systems industry.However, the required effort, depending on the execution platform, may be high, and so far this prevents the application of the methods to multi-core architectures.The PEP project will extend the applicability of sound timing analysis to more complex architectures, including multi cores, by applying compositional analysis approaches to reduce complexity. To this end, the first goal of PEP is to identify processor designs that enable compositional timing analysis.This goal can be easily reached by removing all performance-enhancing architectural features.PEP will show how to enable compositional analysis without removing features like pipelining and caching, and thus without significantly sacrificing performance.Compositional analysis usually sacrifices precision for reduced analysis complexity.The second goal of PEP is to explore the tradeoff between integrated and compositional analysis approaches. PEP will identify a sweet spot that offers good precision at acceptable analysis efficiency.

• DFG Programme: Research Grants