优化的可重构多常数乘法器生成算法

Optimized Generation Algorithm for Reconfigurable Multiple-constant Multiplier

针对线性变换中单个定点数输入与多组定点常数相乘的问题,以加/减法器、移位器和多路选择器为基本单元,提出一种可重构多常数乘法器(RMCM)生成算法。该算法分别运用局部优化和全局优化2种策略,通过计算多常数乘法器(MCM)之间的拓扑相似度,对MCM的有向无环图(DAG)进行合并,选取其中估算面积最小的DAG作为输出。实验结果表明,利用该算法生成的乘法器在面积上具有优势,可用于实现视频多标准中多组不同系数的线性变换。

Aiming at the problem of the Multiplier of a fixed-point input by multiple sets of fixed-point constants in linear transformation, this paper proposes Reconfigurable Multiple-constant Multiplier（RMCM） algorithm making use of adders/subtractors, shifters and multiplexers. It generates RMCM in two strategies, which are local optimization and global optimization, and selects the best Directed Acyclic Graph（DAG） with the smallest areas as the output. Experimental result shows that the Multiplier generated by the proposed algorithm has less area costs and can be applied in linear transformation with several different sets of coefficients.