Remote Access Mathematics of Computation
Green Open Access

Mathematics of Computation

ISSN 1088-6842(online) ISSN 0025-5718(print)



On relative errors of floating-point operations: Optimal bounds and applications

Authors: Claude-Pierre Jeannerod and Siegfried M. Rump
Journal: Math. Comp. 87 (2018), 803-819
MSC (2010): Primary 65G50
Published electronically: July 7, 2017
Full-text PDF

Abstract | References | Similar Articles | Additional Information

Abstract: Rounding error analyses of numerical algorithms are most often carried out via repeated applications of the so-called standard models of floating-point arithmetic. Given a round-to-nearest function $ \textnormal {fl}$ and barring underflow and overflow, such models bound the relative errors $ E_1(t) = \vert t-\textnormal {fl}(t)\vert/\vert t\vert$ and $ E_2(t) = \vert t-\textnormal {fl}(t)\vert/\vert\textnormal {fl}(t)\vert$ by the unit roundoff $ u$. This paper investigates the possibility and the usefulness of refining these bounds, both in the case of an arbitrary real $ t$ and in the case where $ t$ is the exact result of an arithmetic operation on some floating-point numbers. We show that $ E_1(t)$ and $ E_2(t)$ are optimally bounded by $ u/(1+u)$ and $ u$, respectively, when $ t$ is real or, under mild assumptions on the base and the precision, when $ t = x \pm y$ or $ t = xy$ with $ x,y$ two floating-point numbers. We prove that while this remains true for division in base $ \beta > 2$, smaller, attainable bounds can be derived for both division in base $ \beta =2$ and square root. This set of optimal bounds is then applied to the rounding error analysis of various numerical algorithms: in all cases, we obtain significantly shorter proofs of the best-known error bounds for such algorithms, and/or improvements on these bounds themselves.

References [Enhancements On Off] (What's this?)

Similar Articles

Retrieve articles in Mathematics of Computation with MSC (2010): 65G50

Retrieve articles in all journals with MSC (2010): 65G50

Additional Information

Claude-Pierre Jeannerod
Affiliation: Inria and Université de Lyon, laboratoire LIP (CNRS, ENS de Lyon, Inria, UCBL), 46 allée d’Italie 69364 Lyon cedex 07, France

Siegfried M. Rump
Affiliation: Hamburg University of Technology, Schwarzenbergstraße 95, Hamburg 21071, Germany — and — Faculty of Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan

Received by editor(s): April 20, 2016
Received by editor(s) in revised form: October 26, 2016
Published electronically: July 7, 2017
Article copyright: © Copyright 2017 American Mathematical Society

American Mathematical Society