FLOATING POINT ARITHMETIC UNIT, ERROR CORRECTION METHOD THEREFOR, LINEAR PROGRAMMING PROBLEM OPTIMIZING DEVICE AND METHOD FOR OBTAINING OPTIMUM SOLUTION THEREFOR

摘要

PURPOSE: To prevent the error of an arithmetic result from influencing succeeding arithmetic operations with respect to a floating point arithmetic unit, an error correction method therefor, a linear programming problem optimizing device provided with the floating point arithmetic unit and a method for obtaining the optimum solution. ;CONSTITUTION: This unit is provided with the exponent holding part A10 of a normalized floating point number, the exponent holding part B13 of the normalized floating point number of the arithmetic result, a threshold value holding part 11 for holding the threshold value of the reference of judging whether or not the arithmetic result is to be corrected, an exponent comparison part 12 for comparing the compared result of an exponent before the arithmetic operation and the exponent after the arithmetic operation with the threshold value and judging whether or not to correct the arithmetic result and an arithmetic result change part 14 for correcting the arithmetic result based on the compared result of the exponent comparison part 12. When the compared result of the exponent before the arithmetic operation and the exponent after the arithmetic operation is larger than the threshold value, the arithmetic result is corrected to be '0'.;COPYRIGHT: (C)1995,JPO