摘要 |
PURPOSE:To provide the floating point arithmetic unit which can execute complicated function arithmetic deciding the number of valid digits at high speed. CONSTITUTION:The mantissa part of floating point data is inputted to a mantissa register 1, and the exponent part is inputted to an exponent register 2. A floating point position instructor 4 detects the position of a floating point from the contents of the exponent register 2 and based on that position, a floating point position corrector 3 shifts the contents of the mantissa register 1 and generates fixed point data. Corresponding to the fixed floating point data, a mantissa arithmetic unit 5 executes the operation processing. Thus, high-speed operation is realized by omitting the exponent operation processing required for the conventional floating point operation. |