An interval-based approach to exhaustive and incremental interprocedural data-flow analysis

Reviewer: Martin Joseph Jourdan

In this respectable research paper the author presents an original method to solve the archetypal interprocedural dataflow problem of computing the MOD and REF sets for each procedure in a complete program. His method is based on the decomposition of this problem into a hierarchical collection of subproblems, some of which reduce to mere intraprocedural problems, and on structuring the program call graph into intervals. Burke also shows how to adapt his method to perform incremental analysis, which is quite useful for program development environments. The paper is very long both because the problem is hard and because the author thoroughly explores all the details of his technique, gives numerous examples that he discusses at length, and constantly compares his ideas with other relevant works. Because the paper has been revised several times over an unusually long period (six years), it includes comments on other works based on earlier versions of itself, which gives a broad perspective on the problem and alternative ways to solve it. This makes the paper very interesting for those who are motivated enough to absorb the huge quantity of terminology and details (basic knowledge of dataflow analysis and graph theory is helpful). In spite of the weight of the formalism, the paper is clear and well structured. The numerous comparisons with other works and the good bibliography qualify the paper as a reference work, which undoubtedly deserved to be published in this archival journal. In addition, in spite of the long publication delays, it is still up to date: in a personal communication, the author told me that today he would write nearly the same paper. I recommend this paper to all who are interested in the theoretical and practical aspects of interprocedural dataflow analysis.