Accelerating Software Reliability Assessment Through Program Transformation

摘要

A theoretically sound approach for assessing software reliability is to use statistical sampling according to the operational profile. However, this method frequently requires a prohibitive amount of testing. To make this approach more feasible, it is necessary to reduce the time needed to execute a test case. In this paper, we develop transformation methods that speed up the execution of each test case so that a large number of test cases can be run in a shorter time. Essentially, a given program P is transformed into a version Q having two important properties: (1) Q will run faster than P for the same input and (2) if the output of Q is correct for a given input then it is guaranteed that the ouput of P will also be correct for that input. If Q is incorrect for an input then P has to be run to determine the behavior of P for that input. Transformations that assure this type of equivalence between two program are called monotoni-cally equivalent. We present several monotonically equivalent transformation rules.