Lightweight Semi-Dynamic Methods for EÆcient and E ective Program Slicing

When we try to debug a large program e ectively, it is very important to separate a suspicious program portion from the overall source program. Program slicing is a promising technique to extract a program portion; however, it remains diÆcult issues. Static slicing sometimes produces a large portion of the source program, especially for a program with arrays and pointers. Dynamic slicing requires unacceptably huge run-time overhead. In this paper, we discuss intermediate semi-dynamic methods between static and dynamic slicing. We propose two slicing methods named call-mark slicing and dependence-cache slicing. These algorithms have been implemented in our experimental slicing system, and execution data for several sample programs have been collected. The result shows that call-mark slicing reduces the slice size by about 10{20% from the static slice size, with very little overhead increase. Also, dependence-cache slice reduces by about 30{90%, even for programs using arrays, with a ordable run-time overhead increase. These slicing methods will be important features for e ective debugging environments. Author address: Katsuro Inoue Department of Informatics, Graduate School of Engineering Science, Osaka University, 1-3 Machikaneyama, Toyonaka, Osaka 560-8531, Japan Ph. +81-6-6850-6570 Fax +81-6-6850-6574 E-mail [email protected] This work is partly supported by Ministry of Education, Science, Sports, and Culture, grant-in-aid for priority areas \Principles for Constructing Evolutionary Software".