Error Detection for Floating-Point Program via Branch and Bound Method

DOI

10.2991/isrme-15.2015.247How to use a DOI?

Keywords

floating-point; error detection; program rewriting; branch and bound;GNU Scientific Library

Abstract

It is well-known that writing an error-free floating-point program is very difficult. Thus, detecting unacceptable errors of a floating-point program is important. In this paper, we develop a system named SpaceAED. The main function of this system is to automatically detect unacceptable errors of a floating-point program written in C programming language. The key insight of this work is to use interval arithmetic in conjunction with branch and bound technique. The implementation of SpaceAED is to rewrite a floating-point program to one that can run on interval arithmetic, and then use branch and bound technique to find all inputs that can trigger unacceptable errors. We choose a great many of functions in GNU Scientific Library (GSL) to test SpaceAED, including matrix computations and evaluation of special functions etc. Numerical results show that SpaceAED is available for accurately detecting unacceptable error-triggering inputs of numerical functions.

Copyright

© 2015, the Authors. Published by Atlantis Press.

Open Access

This is an open access article distributed under the CC BY-NC license (http://creativecommons.org/licenses/by-nc/4.0/).

Download article (PDF)

TY  - CONF
AU  - Kai Song
AU  - Xia Zeng
AU  - Min Tang
PY  - 2015/04
DA  - 2015/04
TI  - Error Detection for Floating-Point Program via Branch and Bound Method
BT  - Proceedings of the 2015 International Conference on Intelligent Systems Research and Mechatronics Engineering
PB  - Atlantis Press
SP  - 1206
EP  - 1215
SN  - 1951-6851
UR  - https://doi.org/10.2991/isrme-15.2015.247
DO  - 10.2991/isrme-15.2015.247
ID  - Song2015/04
ER  -