Design and Analysis of Approximate 4–2 Compressors for High-Accuracy Multipliers
Tianqi Kong, Shuguo Li
Abstract
Approximate multipliers are applicable in error-resilient applications with relaxed precision constraints, including image processing, multimedia, and data recognition. Such multipliers that sacrifice some accuracy can gain a corresponding increase in electrical performance. This article presents an analysis of the architectures of previously proposed compressors to investigate their performance and accuracy. In this article, we propose five high-accuracy approximate 4–2 compressors with better delay, area, power, and better performance–accuracy tradeoff. Pro1–Pro4 rely on the critical path optimization, while Pro5 derives from the modified sorting technique. This article implements <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$8 \times 8$ </tex-math></inline-formula> and <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$16 \times 16$ </tex-math></inline-formula> multipliers by employing the proposed approximate compressors in TSMC 28 nm. The experimental results indicate that our designs have about 18% delay, 43%–52% area-delay product (ADP) reduction compared to the exact multiplier, and 20%–55% ADP optimization compared to compressors with the same accuracy. This article further verifies the efficacy of the proposed compressors through image blending and matrix multiplication applications.