Towards a hybrid probabilistic timing analysis
| Autor: | Peter Hellinckx, Haoxuan Li, Paul De Meulenaere, Siegfried Mercelis, Ken Vanherpen |
|---|---|
| Jazyk: | angličtina |
| Rok vydání: | 2021 |
| Předmět: |
Computer. Automation
Physics and Astronomy (miscellaneous) business.industry Computer science Management of Technology and Innovation Probabilistic timing analysis Artificial intelligence business Machine learning computer.software_genre Engineering (miscellaneous) computer Engineering sciences. Technology |
| Zdroj: | Advances in Science, Technology and Engineering Systems Journal |
| ISSN: | 2415-6698 |
| Popis: | Real-time embedded systems are widely adopted in applications such as automotive, avionics, and medical care. As some of these systems have to provide a guaranteed worst-case execution time to satisfy the time constraints, understanding the timing behaviour of such systems is of the utmost importance regarding the reliability and the safety of these systems. In the past years, various timing analysis techniques have been developed. Probabilistic timing analysis has recently emerged as a viable alternative to state-of-the-art deterministic timing analysis techniques. Since a certain degree of deadline miss is still tolerable for some systems, instead of deriving an estimated worst-case execution time that is presented as a deterministic value, probabilistic timing analysis considers execution times as random variables and associates each possible execution time with a probability of occurrence. However, in order to apply probabilistic timing analysis, the measured execution times must be independent and identically distributed. In the particular case of hybrid timing analysis, since the input and the initial processor state of one software component are influenced by the preceding components, it is difficult to meet such prerequisite. In this article, we propose a hybrid probabilistic timing analysis method that is able to (i) reduce the dependence in the measured execution times to facilitate the application of extreme value theory and (ii) reduce the dependence between software components to make it possible to use convolution to calculate the probabilistic WCET of the overall system. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|