An Abstract Domain to Discover Interval Linear Equalities
نویسندگان
چکیده
We introduce a new abstract domain, namely the domain of Interval Linear Equalities (itvLinEqs), which generalizes the affine equality domain with interval coefficients by leveraging results from interval linear algebra. The representation of itvLinEqs is based on a row echelon system of interval linear equalities, which natively allows expressing classical linear relations as well as certain topologically non-convex (even unconnected or non-closed) properties. The row echelon form limits the expressiveness of the domain but yields polynomial-time domain operations. Interval coefficients enable a sound adaptation of itvLinEqs to floating-point arithmetic. itvLinEqs can be used to infer and propagate interval linear constraints, especially for programs involving uncertain or inexact data. The preliminary experimental results are encouraging: itvLinEqs can find a larger range of invariants than the affine equality domain. Moreover, itvLinEqs provides an efficient alternative to polyhedra-like domains.
منابع مشابه
An Abstract Domain for Bit-Vector Inequalities⋆
This paper advances the state of the art in abstract interpretation of machine code. It tackles two of the biggest challenges in machine-code analysis: (1) holding onto invariants about values in memory, and (2) identifying affine-inequality invariants while handling overflow in arithmetic operations over bit-vector data-types. Most current approaches either capture relations only among registe...
متن کاملAbstraction of Optional Numerical Values
ion of Optional Numerical Values Jiangchao Liu and Xavier Rival INRIA, ENS, CNRS, PSL*, Paris, France, [email protected], [email protected] Abstract. We propose a technique to describe properties of numerical stores with optional values, that is, where some variables may have no value. Properties of interest include numerical equalities and inequalities. Our approach lifts common linear inequality b...
متن کاملSubPolyhedra: A (More) Scalable Approach to Infer Linear Inequalities
We introduce Subpolyhedra (SubPoly) a new numerical abstract domain to infer and propagate linear inequalities. SubPoly is as expressive as Polyhedra, but it drops some of the deductive power to achieve scalability. SubPoly is based on the insight that the reduced product of linear equalities and intervals produces powerful yet scalable analyses. Precision can be recovered using hints. Hints ca...
متن کاملInterprocedural analysis of low-level code
Static analysis of machine code is employed for reverse engineering, automatic detection of low-level errors such as memory violations, malware detection, and many other application areas. Only at the level of executables can all errors introduced by programmers or even by compilers be identified. Analysis of machine code comes at a price: high-level language features such as local variables an...
متن کاملA Smooth Combination of Linear and Herbrand Equalities for Polynomial Time Must-Alias Analysis
We present a new domain for analyzing must-equalities between address expressions. The domain is a smooth combination of Herbrand and affine equalities which enables us to describe field accesses and array indexing. While the full combination of uninterpreted functions with affine arithmetics results in intractable assertion checking algorithms, our restricted domain allows us to construct an a...
متن کامل