| 摘要 |
PURPOSE:To form data having a variable-length exponential part from a fixed- length exponential part and a fixed-length mantissa part idependently of the size of the exponential part by outputting the fixed-length exponential part data and the fixed-length mantissa part data as processed results. CONSTITUTION:A floating-point arithmetic unit is constituted of a floating-point register 11, a mantissa part register 12, an exponential part register 13, a counter 14, a flip flop 15, and a two-input logical circuit 16. Data operation is executed by the following procedure. Namely, 1: counting of a '0' or '1' string, 2: separation of the exponential part from the mantissa part, 3: adjustment of the value of the exponential part in case of addition/subtraction, 4: operation, 5: normalization, 6: coupling the exponential part with the mantissa part, and 7: insertion of the '0' or '1' string. Consequently, a numeral approximate to '1' which may frequency appear in ordinary time can be expressed more accurately as compared to the ordinary method and data based upon the exponential part variable length expression system which can be expressed can be formed also for an extremely large value or an extremely small value. |
