Skip to Main content Skip to Navigation
Reports

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

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 $\fl$ and barring underflow and overflow, such models bound the relative errors $E_1(t) = |t-\fl(t)|/|t|$ and $E_2(t) = |t-\fl(t)|/|\fl(t)|$ by the unit roundoff $u$. This paper investigates the possibility 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 provide explicit and attainable bounds on $E_1(t)$, which are all less than or equal to $u/(1+u)$ and, therefore, smaller than~$u$. For $E_2(t)$ the bound $u$ is attainable whenever $t=x\pm y$ or $t=xy$ or, in base $\beta >2$, $t=x/y$ with $x,y$ two floating-point numbers. However, for division in base $2$ as well as for inversion and square root, smaller bounds are derived, which are also shown to be attainable. This set of sharp bounds is then applied to the rounding error analysis of various numerical algorithms: in all cases, we obtain either much shorter proofs of the best-known error bounds for such algorithms, or improvements on these bounds themselves.
Complete list of metadata

https://hal.inria.fr/hal-00934443
Contributor : Claude-Pierre Jeannerod <>
Submitted on : Wednesday, January 22, 2014 - 9:29:31 AM
Last modification on : Thursday, January 9, 2020 - 3:16:02 PM
Long-term archiving on: : Thursday, April 24, 2014 - 10:45:32 AM

File

JeannerodRump2014.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : hal-00934443, version 1

Citation

Claude-Pierre Jeannerod, Siegfried M. Rump. On relative errors of floating-point operations: optimal bounds and applications. [Research Report] 2014, pp.15. ⟨hal-00934443v1⟩

Share

Metrics

Record views

664

Files downloads

301