| 摘要 |
PURPOSE: A floating point data multiplication method is provided to enlarge a dynamic calculation area in a multiplicator adopting a floating point calculation scheme, and reduce capacity of a used memory while enhancing a noise/signal ratio. CONSTITUTION: The method comprises several steps. Two mantissas of two input data are multiplied and two exponents of the two input data are added(S1). The added result of the two exponents is compared with a maximum limit of an exponent range, and a bit number of an MSB(Most Significant Bit), corresponding to the same bit as a sign bit of the two mantissa operation result is calculated(S2-S4). The added result of the two exponents is compared with a minimum limit of an exponent range, and the two mantissa operation result is right-shifted or left-shifted according to the comparison result(S5-S7).
|
