Fine-Grained Critical Path Analysis and Optimization for Area-Time Efficient Realization of Multiple Constant Multiplications

摘要

In this paper, critical path of multiple constant multiplication (MCM) block is analyzed precisely and optimized for high-speed and low-complexity implementation. A delay model based on signal propagation path is proposed for more precise estimation of critical path delay of MCM blocks than the conventional adder depth and the number of cascaded full adders. A dual objective configuration optimization (DOCO) algorithm is developed to optimize the shift-add network configuration to derive high-speed and low-complexity implementation of the MCM block for a given fundamental set along with a corresponding additional fundamental set. A genetic algorithm (GA)-based technique is further proposed to search for optimum additional fundamentals. In the evolution process of GA, the DOCO is applied to each searched additional fundamental set to optimize the configuration of the corresponding shift-add network. Experimental results show that the proposed GA-based technique reduces the critical path delay, area, power consumption, area delay product and power delay product by 32.8%, 4.2%, 5.8%, 38.3%, and 41.0%, respectively, over other existing optimization methods.