Improvised HPR Multiplier Using Fast Bipartitioned Hybrid Adder by Approximate Computing

View at publisher Summarise with VIRAT
Authors: K.S. Neelu Kumari, P. Agniraj, I. Azarudeen, P. Chandraprakash
Cited by: 0
References: 12

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The triple modular redundancy (TMR) scheme is generally used to build a fault-tolerant arithmetic unit in critical applications that require higher reliability. While TMR is effective, it also incurs a significant area and power/energy overhead. The primary voter in most TMR designs can also be highly susceptible to soft errors and may result in an error for only a small difference between the modules' outputs. There is also a wide spectrum of applications deployed in critical systems that are inherently error-resilient. In particular, they can tolerate some degree of inexactness at their output while still meeting some reliability level. In this paper, propose a high-precision redundancy multiplier (HPR-Mul) architecture, implementing principles of approximate computing, augmented with a Fast Bipartitioned Hybrid Adder (FBHA) to achieve higher performance while still meeting the prescribed reliability level. The HPR-Multiplier consists of one full-precision (FP) and two reduced-precision (RP) multipliers, together with a simple voter to arrive at an output. Unlike most current$\mathbf{R P}$redundancy multipliers (RPR-Multiplier) that require a complicated voter, the voter of the present HPR-Mul architecture relies on mathematical formulas resulting in a much simpler structure. In addition, the intermediate signals of the FP multiplier are employed as inputs to the RP multipliers thus greatly improving HPR-Mul accuracy. The main contribution of this paper is to replace the HPR-Mul's traditional summation block that suffers from delay, with a Fast Binary Hierarchical Adder (FBHA), a high-speed adder and architecture that utilizes parallel carry-lookahead and carryselect logic. This change is to target the multiplier's critical path to reduce delay and improve performance overall.

Fields of Study & Subjects

Low-power high-performance VLSI design Radiation Effects in Electronics Numerical Methods and Algorithms

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
158–164
Publication date
21 January 2026
Language
EN