ARITHMETIC UNIT PERFORMING DIVISION OR SQUARE ROOT OPERATION OF FLOATING POINT NUMBER AND OPERATING METHOD

摘要

When division X/Y of floating point numbers is performed, bit string data of a mantissa x including a fraction xf or a mantissa y including a fraction yf is shifted in accordance with magnitude relation between xf of X and yf of Y, and a fraction computation is performed. Thereby, a result of division of a fraction in which the position of the most significant bit is fixed to be in a prescribed digit is generated. When a square root operation √Y is performed, a fraction square root operation result in which the position of the most significant bit is fixed to be in a prescribed digit is generated by an exception handling if all of the three conditions that all the bits in the fraction of Y have the values of one, a value of ye-b that is a difference between the exponent ye of Y and a bias value b is an odd number, and a rounding mode is a positive infinity direction are satisfied.