| 摘要 |
PURPOSE:To find an inverse with a double accurate value in a short processing time and with high accuracy, by finding the inverse by using an approximate expression by Taylor expansion. CONSTITUTION:Assuming f(x)=1/x, it is possible to obtain the approximate expression of a formula by applying the Taylor expansion on f(A+B.2<-c>) (A, B: value of single precision). The inverse of (A+B.2<-c>) can be obtained by using the formula. To find the inverse, firstly, 1/A is found by referring to an inverse table 100, and it is stored in a register A104. Next, (1/A)<2> is found by using a multiplier 101, and it is stored in a register B105. Then, the multiplication of (1/A)<2> and B is performed by using the multiplier 101, and the integer part of a multiplied result is obtained at a shifter 102, and by subtracting it from the register A104, the high-order eight bits of the inverse can be obtained. Also, similarly by finding a decimal part by a shifter 102, and calculating the two complements of the part by an arithmetic computing element 103, the low- order eight bits of the inverse can be obtained.
|
