With the advent of fast multipliers, the latency of a floatingpoint division is greatly reduced using a multiplicative method ratherthan a subtractive one. For most multiplicative division algorithms, aninitial reciprocal approximation of the divisor is largely obtained by alook-up table method. Once the initial approximation is made, it isrefined by the use of a functional iteration method until the accuracyof the reciprocal approximation of the divisor is accurate enough toproduce the final quotient. This paper introduces a look-up tableconstruction method using the concept of a bipartite reciprocal tablethat is applied to a piecewise linear approximation. The error of thelinear approximation proposed in this paper is less than a quarter of anulp (unit in the last place). The latency of this reciprocal unit is 3cycles for the IEEE-754 double precision operation
展开▼
