A comparative study of novel variable set multiplier scheme and other column compression multipliers

The extensive use of multipliers in high-performance processors has been driving the demand for low-power, high-speed multipliers. This report presents a novel fast multiplier scheme that performs reduction based on non-uniform grouping in the partial product matrix. In every stage, the partial products are grouped into sets of variable heights before reduction. The algorithm is presented and compared with Wallace, Dadda, and Reduced Area Multipliers in terms of gate complexity, delay, interconnects and power utilization for 4-bit, 7-bit, 8-bit and 12-bit versions. The proposed design is observed to have a gate count of 3% lower than Wallace multipliers, comparable to Dadda in most of the cases but slightly higher than Reduced Area multipliers. To fulfill the study, a 12-bit version of each multiplier was implemented in structural Verilog with basic gates and synthesized using Xilinx Vivado. The net cell usage statistics and the power utilization values were extracted and compared. It was observed that the proposed design has better power efficiency than the other multipliers for the examined cases.