Abstract
The use of high-order integration formulae in general-purpose library routines is widely discouraged in the literature. The reasons advanced for the recommended preference for the trapezoidal, mid-point and Simpson's rules are here analysed, and found to be either irrelevant to modern computation, or highly inconclusive. Attainable error bounds are presented which help to make high-order formulae equally attractive in problems for which they were formerly regarded as inefficient.

This publication has 0 references indexed in Scilit: