Some Numerical Experiments on Round-off Error Growth in Finite Precision Numerical Computation

Abstract

Finite precision computations using digital computers involve the following inherent errors: (1) Round-off error of finite precision computations (2) Binary computer arithmetic precludes exact number representation of traditional decimal system used in data input stage to the computer. Though the round-off error is as small as in the seventh decimal place (single precision) in the beginning, it can enter the mainstream computation within 50 iterations in iterative computations, such as that used in numerical integration schemes, for example, the commonly used fourth order Runge-Kutta method. Growth of round-off error in recursive vis-a-vis iterative computing is described in the text. In this paper several computational experiments are presented to demonstrate the rapid growth of round-off error in iterative computations.