A Critical Analysis of Incremental Iterative Data Flow Analysis Algorithms

Abstract

Recently, incremental iterative data flow analysis algorithms have been proposed for updating data flow information after a program change. Unfortunately, many of them are “faulty”; they do not yield the same solution as an exhaustive, “from scratch” fixed point iteration on the altered data flow problem. The basis of these faulty algorithms is the technique of restarting iteration. We discuss this technique, giving illustrative examples from interprocedural and intraprocedural analysis. We also summarize some incremental techniques which calculate precise data flow information. © 1990 IEEE