Constructing a Fa for Hardware Hastening For Dsp

Abstract

CS representation continues to be broadly accustomed to design fast arithmetic circuits because of its natural benefit of getting rid of the big carry-propagation chains. Hardware acceleration continues to be demonstrated a very promising implementation technique for digital signal processing (DSP) domain. However, research activities have proven the arithmetic optimizations at greater abstraction levels compared to structural circuit one considerably effect on the data path performance. Instead of adopting a monolithic application-specific integrated circuit design approach, within this brief, we present a manuscript accelerator architecture composed of flexible computational models that offer the execution of a big group of operation templates present in DSP popcorn kernels. Extensive experimental evaluations reveal that the suggested accelerator architecture provides average gains as high as 61.91% in area-delay product and 54.43% in energy consumption in comparison using the condition-of-art flexible data paths. We differentiate from previous creates flexible accelerators by enabling computations to become strongly carried out with carry-save (CS) formatted data. Advanced arithmetic design concepts, i.e., recoding techniques, are employed enabling CS optimizations to become carried out inside a bigger scope compared to previous approaches.