In a recent work on multiple constant multiplication (MCM) problems, a fully predictable sub-linear runtime heuristic was introduced, called Radix-2 ^r MCM. This method shows competitive results in speed, power and area, comparatively with the leading algorithms. In this study, the authors combine Radix-2 ^r MCM with an exact common subexpression elimination (CSE) algorithm. The resulting algorithm denoted Radix-2 ^r -CSE allows a substantial reduction in the number of addition/subtraction operations in MCM by maximising the sharing of partial terms after an initial recoding in Radix-2 ^r MCM. The savings over Radix-2 ^r MCM ranges from 4.34 to 18.75% (10% on average) when considering a set of 14 benchmark finite impulse response filters of varying complexity.

中文翻译：

---

基数2^[R 使用公共子表达式消除进行重新编码以进行多次常数乘法

在有关多重常数乘法（MCM）问题的最新工作中，引入了一种完全可预测的亚线性运行时启发式方法，称为Radix-2 ^[RMCM。与领先的算法相比，该方法在速度，功率和面积上显示出具有竞争力的结果。在这项研究中，作者结合了Radix-2 ^[R具有精确的通用子表达式消除（CSE）算法的MCM。结果算法表示为Radix-2 ^[R -CSE通过在Radix-2中进行初始重新编码后最大程度地共享部分项，从而大大减少了MCM中的加/减运算次数 ^[RMCM。比Radix-2节省的钱 ^[R 当考虑一组14个复杂度各异的基准有限冲激响应滤波器时，MCM的范围为4.34％至18.75％（平均为10％）。