On Systematic Testing for Execution-Time Analysis

TL;DR

This paper introduces a systematic approach for worst-case execution-time testing of loop-free programs, using timing measurements and linear programming to accurately estimate maximum execution times and platform repeatability.

Contribution

It presents a novel method combining timing measurements with integer linear programming to improve accuracy in worst-case execution-time analysis for loop-free programs.

Findings

More accurate WCET bounds than previous methods

Ability to achieve arbitrary accuracy with increased measurements

Quantification of platform timing repeatability

Abstract

Given a program and a time deadline, does the program finish before the deadline when executed on a given platform? With the requirement to produce a test case when such a violation can occur, we refer to this problem as the worst-case execution-time testing (WCETT) problem. In this paper, we present an approach for solving the WCETT problem for loop-free programs by timing the execution of a program on a small number of carefully calculated inputs. We then create a sequence of integer linear programs the solutions of which encode the best timing model consistent with the measurements. By solving the programs we can find the worst-case input as well as estimate execution time of any other input. Our solution is more accurate than previous approaches and, unlikely previous work, by increasing the number of measurements we can produce WCETT bounds up to any desired accuracy. Timing of a program depends on the properties of the platform it executes on. We further show how our approach can be used to quantify the timing repeatability of the underlying platform.