Summary: Making Set­Constraint Program Analyses Scale
Manuel F¨ahndrich \Lambda Alexander Aiken \Lambda
EECS Department
University of California, Berkeley
Berkeley, CA 94720­1776
fmanuel,aikeng@cs.berkeley.edu
July 2, 1996
1 Introduction
Constraint­based program analyses are appealing because elaborate analyses can be described with a concise
and simple set of constraint generation rules. Constraint resolution algorithms have been developed for
many kinds of constraints, conceptually allowing an implementation of a constraint­based program analysis
to reuse large pieces of existing code. In practice, however, new analyses often involve re­implementing
new, complex constraint solving frameworks, tuned for the particular analysis in question. This approach
wastes development time and interferes with the desire to experiment quickly with a number of different
analyses.
We believe that implementing an analysis should require writing only the code to generate the con­
straints, and that a well engineered­library can take care of constraint representation, resolution, and
transformation. Writing such a library capable of handling small programs is not too difficult, but scaling
to large programs is hard. Toward this goal, we are developing a scalable, expressive framework for solving
a class of set constraints. Scalability is achieved through four techniques: polymorphism, simplification,

Source: Aiken, Alex - Department of Computer Science, Stanford University

Collections: Computer Technologies and Information Sciences