HighPrecisionDivisionandSquareRoot-HPLabs.PDFVIP

HEWLETT @ PACKARD High Precision Division and Square Ro ot Alan H. Karp Peter Markstein HP Labs Rep ort 93-93-42 (R.1) June 1993 Revised Octob er 1994 Multiprecision oating p oint, quad preci- We present division and square ro ot al- sion, square ro ot, division; Computing Re- gorithms for calculations with more bits views Categories G.1.0, G.4 than are handled by the oating p oint hardware. These algorithms avoid the need to multiply two high precision num- b ers, sp eeding up the last iteration by as much as a factor of ten. We also show how to pro duce the oating p oint numb er clos- est to the exact result with relatively few additional op erations. Internal Accession Date Only c Copyright Hewlett-Packard Company 1994 1 Intro duction Floating p oint division and square ro ot take considerably longer to compute than addition and multiplication. The latter two are computed directly while the former are usually com- puted with an iterative algorithm. The most common approach is to use a division-free Newton-Raphson iteration to get an approximation to the recipro cal of the denominator (division) or the recipro cal square ro ot, and then multiply by the numerator (division) or input argument (square ro ot). Done carefully, this approach can return a oating p oint