Analyzing Loop Paths for Execution Time Estimation

Statically estimating the worst case execution time of a program is important for real-time embedded software. This is difficult even in the programming language level due to the inherent difficulty in detecting infeasible paths in a program's control flow graph. In this paper, we study the problem of accurately bounding the execution time of a program loop. This involves infeasible path detection followed by timing analysis. We employ constraint propagation methods to detect infeasible paths spanning across loop iterations. Our timing analysis is exact modulo the infeasible path information provided. Moreover, the analysis is efficient since it relies on memoization techniques to avoid exhaustive enumeration of all paths through a loop. The precision of our timing analysis is demonstrated on different benchmark programs.