浮点倒数查找表的构造

在浮点除法器的设计中，若用函数迭代或高基数算法进行除法运算，计算开始时，通过浮点倒数查找表获得一个较精确的初始除数倒数近似值，可以减少除法的迭代次数，缩短运算的延迟。即将除数的前几位作为表的一个入口地址，该地址指示的位置存放着满足一定精度的初值倒数近似值。文中详述了几种获得除数倒数近似值的方法，其中包括算法、误差限和精度等。

The Structure of Floating-point Reciprocal Look-up Tables

In design of floating-point division, we can use functional iteration arithmetic or very high radix division arithmetic. In the beginning, we use a look-up table to obtain an accurate initial reciprocal approximation, and reduce the time of delay. The truncated bits of dividend are used to index into a table providing the reciprocal approximation. This paper introduces several means which contain arithmetics, error bounds and presions to obtain the initial reciprocal approximations.