Energy‐Efficient and Fast IMPLY‐Based Approximate 4:2 Compressor Applying NAND Gates for Image Processing

In modern computing applications, the large amount of data that should be moved around between the main memory and processing unit is an essential bottleneck, resulting in time wastage and significant energy dissipation. Also, transistor down-scaling as a solution is impossible because it causes many problems. The issue of changing technology has been raised to overcome these problems. One of these new technologies is the memristor, which can serve as both a computing and a memory unit at the same time. Memristors are suitable for In-Memory Computing (IMC) applications because they are compatible with memristive crossbar arrays. In addition to changing technology, approximate computing can achieve higher efficiency in errortolerant applications. This paper uses memristors and approximate computing, and a Material Implication (IMPLY)-based serial approximate 4:2 compressor is proposed. The block reduces the computational steps by up to 40.38%, and the maximum energy consumption improvement is 40.74% compared to the serial exact cell. Also, the proposed approximate 4:2 compressor is applied in the 8-bit multiplier structure to analyze its application in processing applications. The acceptable reduction of error analysis metrics shows that the accuracy is acceptable. The appropriate functionality of the proposed cell in the image multiplication application is confirmed.