Interprocedural Control Flow Analysis

Abstract

Control Flow Analysis is a widely used approach for analysing functional and object oriented programs. Once the applications become more demanding also the analysis needs to be more precise in its ability to deal with mutable state (or side-effects) and to perform polyvariant (or context-sensitive) analysis. Several insights in Data Flow Analysis and Abstract Interpretation show how to do so for imperative programs but the techniques have not had much impact on Control Flow Analysis. We show how to incorporate a number of key insights from Data Flow Analysis (involving such advanced interprocedural techniques as call strings and assumption sets) into Control Flow Analysis (using Abstract Interpretation to induce the analyses from a collecting semantics).

Faithful Translations between Polyvariant Flows and Polymorphic Types

“…Below we shall see that our descriptive framework can model both approaches (which can be "mixed", as in (Nielson & Nielson, 1999)). …”

Control-Flow Analysis of Functional Programs

“…Nielson and Nielson have championed the specification approach to program analysis [110,54,113,111]. A specification is formulated as a series of declarative demands that a valid analysis result must fulfill.…”

“…Initially Nielson and Nielson's specifications were non-compositional and defined by co-induction [110,54], but later they were reformulated compositionally [113,111] (in which case induction and co-induction coincide [112]). …”

“…The analysis framework of Nielson and Nielson's later paper on higher-order flow analysis supporting side-effects [111] is however compositional, as is Rehof and Fähndrich's [126] context-sensitive flow analysis of typed higherorder programs.…”

“…In a later paper [111], Nielson and Nielson develop a framework for controlflow analysis of a functional language with side-effects. The approach incorporates ideas from interprocedural data-flow analysis.…”