| 摘要 |
PURPOSE:To provide a reciprocal arithmetic unit capable of performing a highly accurate reciprocal computing operation even when the bit width of input data is larger than the usable bit width of the address of a ROM. CONSTITUTION:A first reciprocal computing means 1 computes the reciprocal of a number for which a constant N (where N is an integer satisfying ¦N¦<2<a-b>) is added to the upper (b) bits {where (b) is the positive integer of (b)<(a)} of the element number A of an (a) bit width {where (a) is the positive integer} by a table. A subtractor 5 subtracts the constant N from the lower (a-b) bits of the element number A, a first multiplier 3 multiplies the result of the first reciprocal computing means 1 and the result of the subtractor 5 together and a second reciprocal computing means 2 computes the reciprocal of the number for which '1' is added to the multiplied result of the multiplier 3 by the table. A second multiplier 4 multiplies the result of the first reciprocal computing means 1 and the result of the second reciprocal computing means 2 together and the multiplied result becomes the reciprocal of the element number A. |
