TOTAL SUM CALCULATION CIRCUIT CAPABLE OF RAPIDLY CALCULATING A TOTAL SUM OF MORE THAN TWO INPUT DATA REPRESENTED BY A FLOATING POINT REPRESENTATION

摘要

Abstract of the DisclosureIn a total sum calculation circuit for the use incalculating a total sum of first through n-th input datawhich are represented by a floating point representationand which are composed of first through n-th exponentparts and first through n-th fraction parts, where n isan integer greater than two, an n-input data comparisoncircuit simultaneously compares the first through then-th exponent parts with one another to produce amaximum one of the first through the n-th exponent partsand a comparison result signal representative of whichone of the first through the n-th exponent parts is themaximum exponent part. Supplied with the first throughthe n-th exponent parts and the comparison resultsignal, a shift number calculation circuit calculatesfirst through n-th shift digit numbers between themaximum exponent part and the first through the n-thexponent parts. The first through the n-th fractionparts are shifted by first through n-th shift digitnumbers in first through n-th shifters are produced asfirst through n-th shifted fraction parts which aresummed up into an unnormalized fraction part. Theunnormalized fraction part is normalized into a totalsum fraction part by the use of normalizationinformation derived from the unnormalized fraction part.The maximum exponent part is also normalized by the(Abstract of the Disclosure continued)normalization information into a total sum exponentpart. A combination of the total sum exponent part andthe total sum fraction part is produced as the total sumrepresented by the floating point representation.