Early WCET Prediction Using Machine Learning

For delivering a precise Worst Case Execution Time (WCET), the WCET static analysers need the executable program and the target architecture. However, a prediction – even coarse – of the future WCET would be helpful at design stages where only the source code is available. We investigate the possibility of creating predictors of the WCET based on the C source code using machine-learning (work in progress). If successful, our proposal would offer to the designer precious information on the WCET of a piece of code at the early stages of the development process. 1998 ACM Subject Classification D.4.7 Real-Time Systems and Embedded Systems