A flow analyzer can be very helpful in the process of program understanding, by providing the programmer with different views of the code. As the documentation is often incomplete or inconsistent, it is extremely useful to extract the information a programmer may need directly from the code. Program understanding activities are interactive, thus program analysis tools may be asked for quick answers by the maintainer. Therefore the control on the trade-off between accuracy and efficiency should be given to the user. This paper presents an approach to interprocedural reaching definitions flow analysis based on three levels of precision depending on the sensitivity to the calling context and the control flow. A lower precision degree produces an overestimate of the data dependencies in a program. The result its anyhow conservative (all dependencies which hold are surely reported), and definitely faster than the more accurate counterparts. A tool supporting reaching definition analysis in the three variants has been developed. The results on a test suite show that three orders of magnitude can be gained in execution times by the less accurate analysis, but 57,4% extra dependencies are on average added. The intermediate variant is much more precise (1.6% extra dependencies), but gains less in times (one order of magnitude)
Variable Precision Reaching Definitions Analysis for Software Maintenance
Tonella, Paolo;
1996-01-01
Abstract
A flow analyzer can be very helpful in the process of program understanding, by providing the programmer with different views of the code. As the documentation is often incomplete or inconsistent, it is extremely useful to extract the information a programmer may need directly from the code. Program understanding activities are interactive, thus program analysis tools may be asked for quick answers by the maintainer. Therefore the control on the trade-off between accuracy and efficiency should be given to the user. This paper presents an approach to interprocedural reaching definitions flow analysis based on three levels of precision depending on the sensitivity to the calling context and the control flow. A lower precision degree produces an overestimate of the data dependencies in a program. The result its anyhow conservative (all dependencies which hold are surely reported), and definitely faster than the more accurate counterparts. A tool supporting reaching definition analysis in the three variants has been developed. The results on a test suite show that three orders of magnitude can be gained in execution times by the less accurate analysis, but 57,4% extra dependencies are on average added. The intermediate variant is much more precise (1.6% extra dependencies), but gains less in times (one order of magnitude)I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
