Publication Date: 2026
Integration (01679260)108
In computers with Von-Neumann architecture, significant performance degradation and high power dissipation in data-intensive applications are critical concerns and have attracted considerable attention. This is due to the frequent transfer of information between the processor and memory units. The shift toward emerging technologies is considered a solution to overcome these challenges. Memristors are one of these emerging technologies, possessing non-volatile and crossbar array compatibility features, making them a suitable choice for in-memory computing (IMC). As approximate computing is capable of reducing power consumption, hardware complexity, and latency in fault-tolerant applications, it can be integrated with emerging technologies to improve performance. Therefore, in this article, an approximate 4:2 compressor based on the IMPLY method is proposed within a serial architecture. The proposed approximate 4:2 compressor consists of only six NAND gates at the circuit level, and its IMPLY-based memristor implementation utilizes five memristors and takes thirteen computational steps. The results indicate that the proposed approximate compressor (PAC) outperforms its predecessors in terms of the number of memristors, computational steps, energy consumption, and energy-delay product (EDP). It improves these parameters by an average of 37.5%, 58.06%, 56.5%, and 81.74% compared to the IMPLY-based approximate 4:2 compressor, and by a maximum of 28.57%, 75%, 74.22%, and 93.56% compared to the exact one, respectively. Furthermore, the proposed compressor has been applied in an 8-bit approximate multiplier structure for image multiplication and image smoothing applications. MATLAB simulations indicate that it offers a well-balanced trade-off between accuracy and design efficiency for fault-tolerant applications. Copyright © 2026. Published by Elsevier B.V.
