| 摘要 |
<P>PROBLEM TO BE SOLVED: To provide a storing method of a floating point numeric value for easily processing the floating point numeric value by a processor of a fixed point. <P>SOLUTION: A mantissa part is stored in an U bit at a higher order side of a bit field of N bit (N≥(U+L)) with a fixed point numeric value, and an exponent part is stored in an L bit at a lower order side with an integer, when a is the mantissa part and n is the exponent part in an actual number x represented by a*(2^n). In a case of the multiplication of two actual numbers in this kind of format, both actual numbers are multiplied as the fixed point numeric values, only an effective bit of higher order of the obtained result shall be the exponent part, on the other hand, both actual numbers are added as integers, and only an effective bit of the lower order of the obtained result shall be the exponent part, whereby the result of multiplication is obtained with the floating point format. <P>COPYRIGHT: (C)2004,JPO |
