Sun Microsystems, Inc.

Sergey R Shevtsov

Valery Y Gorshtein

Anatoly I Grushin

A floating point addition unit includes two subunits each of which performs the addition. One subunit ('rounding subunit') rounds the addition result, and the other subunit ('non-rounding subunit') does not. The result of the rounding subunit is selected as the addition result when one of the following conditions (R1), (R2), (R3) is true: (R1) the operation is an effective addition; (R2) the operation is an effective subtraction, the magnitude ED of the difference between the exponents of the operands is 1, and normalization of the result is not required; (R3) the operation is an effective subtraction and ED&gt;1. The addition result is selected from the non-rounding subunit in the remaining cases. In some embodiments, the rounding subunit overlaps rounding with adding the operands, significands. In some embodiments, the addition unit satisfies ANSI/IEEE Standard 754-1985.

Growing community of inventors

Moscow, Russia

Sergey R Shevtsov