Managing the granularity of constraint-based analyses by rule transformation

Abstract

This paper proposes a transformation-based approach to design efficient constraint-based analysis at a larger granularity. In this approach, we can design a less or equally precise but more efficient version of an original analysis by rule transformation. To do this, we first define or design an index determination rule for a new sparse analysis based on some syntactic properties, so that it can partition the original indices, and then transform the original construction rules into new ones by applying the partition. As applications of this approach, we presents sparse versions of side-effect analysis and exception analysis, which give equally precise information for functions as the original ones. 