An Approximation-Based Abstract Interpretation Framework for Formal Verification of Floating-Point Programs

Formal verification of programs involved with floating-point data type still remains a tough problem nowadays. One of the major problems faced in this area is how to reflect the rounding error caused by IEEE standard floating-point processing for verification. In this paper we present an approximation approach which basically converts all floating-point data objects into corresponding integer values. It is carried by some scale operations, which are shown equivalent to a consistent Abstract Interpretation. We have also developed a Web-based verification system to check programming works submitted by students. Experimental results show that the learning interests of students have been significantly improved since floating-point programming problems were supported in the system.