浮点倒数查找表的构造  

The Structure of Floating-point Reciprocal Look-up Tables

作　　者：李蓉[1] 于伦正[1] 时晨[1] 

机构地区：[1]西安微电子技术研究所,陕西西安710075

出　　处：《微电子学与计算机》2007年第7期23-26,共4页Microelectronics & Computer

摘　　要：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。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.

关 键 词：浮点除法 函数迭代 查找表 初始倒数近似值 

分 类 号：TP332[自动化与计算机技术—计算机系统结构]