Power-Area Efficient Computing Technique for Approximate Multiplier with Carry Prediction

Many modern applications which are error resilient in nature including searching on web, processing of images and videos, machine learning and artificial intelligence, and computer vision have a tendency to accept approximate outputs. These applications use approximate computing to trade the accuracy for efficiency and performance. This paper proposes two designs for approximate multipliers which incorporate the carry prediction and truncation technique. The findings demonstrate that the suggested carry prediction technique is effective with notable decreases in power dissipation, latency, and area while having no major impact on the normalized error distance.