The Effects of Optimizing Transformations on DataFlow-Adequate Test Sets

Abstract

Systematic testing during program development is typically performed on some unoptimized version of a program. When development is complete, the program is compiled and optimizing tran~ormations are often applied. While translation such as compilation alters a program’s form, optimizing tran~ormations create a new version of the program that must be retested. This paper characterizes optimizing tran#ormations according to their eflects on adequate test sets developed for data jlow testing. The main benejit of this result is that it is known if a test set remains a&quate after optimization and thus, eliminates the need for complete reanalysis of a program before retesting. When the original test set is inadequate for the optimized version of the program, an incremental testing technique is used to &termine what must be retested. The hta j?ow tester has been incorporated into Free Software Foundation, Inc’s GNU CC compiler” and experiments are being performed.


Related research categories:
(1) Data Flow
(2) Testing

Go To Publications