期刊名称：International Journal of Advanced Computer Science and Applications(IJACSA)
印刷版ISSN：2158-107X
电子版ISSN：2156-5570
出版年度：2019
卷号：10
期号：10
DOI：10.14569/IJACSA.2019.0101058
出版社：Science and Information Society (SAI)
摘要：Numerical accuracy is a critical point in safe computations when it comes to floating-point programs. Given a certain accuracy for the inputs of a program, the static analysis computes a safe approximation of the accuracy on the outputs. This accuracy depends on the propagation of the errors on the data and on the round-off errors on the arithmetic operations performed during the execution. Floating point values disposes a large dynamic range. But the main pitfall is the inaccuracies that occur with floating point computations. Based on the theory of abstract interpretation, in the paper an upper bound to the precision of the results of these computations in program have been demonstrated.
