Abstract
We enumerate the round-off errors in fixed point multiplication both for rounding by chopping and for symmetric rounding. Then we compute the mean and variance of the round-off error. For symmetric rounding, the formulas differ for an odd base β vs an even base β. We conclude that if symmetric rounding is to be used, there is an advantage to an odd base over an even base. For prime base β, no special assumption is needed. For non-prime base β, the last non-zero element of the first factor is assumed to be relatively prime to β.
Similar content being viewed by others
References
Klaus Ecker, Verteilung der Rundungsfehler beim Produkt zweier Zahlen, “Computing”, Vol. 7: 3, (1971), pp. 188–193.
Richard Goodman and Alan Feldstein,Round-off Error in Products, “Computing”, Vol. 15: 3 (1975), pp. 263–273.
Peter Henrici,Discrete Variable Methods in Ordinary Differential Equations, Wiley, New York, 1962.
William P. Timlake,Problems in Discrete Variable Methods in Ordinary Differential Equations, (mimeographed official set of solutions to problems in reference [3]).
Minoru Urabe,Roundoff Error Distribution in Fixed-Point Multiplication and a Remark about the Rounding Rule, SIAM J. Num. An. Vol. 5: 2, June 1968, pp. 202–210.
John Von Neumann and H. H. Goldstine,Numerical Inverting of Matrices of High Order, Bulletin of the AMS, Vol. 53,: 11 November 1947, pp. 1021–1099.
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Goodman, R. On round-off error in fixed-point multiplication. BIT 16, 41–51 (1976). https://doi.org/10.1007/BF01940776
Received:
Revised:
Issue Date:
DOI: https://doi.org/10.1007/BF01940776