Method and circuit for computing reciprocals
US4823301A · kind A · utility
27Cited by
4References
12Claims
0Family size
Assignee
Inventor
Key dates
| Filing date | Oct 22, 1987 |
| Grant date | Apr 18, 1989 |
| Priority date | — |
| Expiry date | Oct 22, 2007 |
Classification
- Technology area (CPC G)Physics
- CPC primaryG06F2207/5356
- WIPO fieldComputer technology
- WIPO sectorElectrical engineering
Abstract
A circuit produces an output binary floating point number approximating with high accuracy the inverse of an input binary floating point number D in accordance with the expression (1/D).apprxeq.[(1/A)-C]+[C-(B/A.sup.2)], where the number A is a low accuracy approximation of D, and B is substantially equal to D-A. C is a number selected for each value of A such that the exponents of quantities [(1/A)-C] and [C-(B/A.sup.2)] are equal to the exponent of the quantity 1/2A. Quantities [(1/A)-C] and [C-(B/A.sup.2)] are produced by lookup tables and summed to provide an approximation of 1/D.
Source: USPTO / EPO open patent data. Objective bibliographic and citation counts.