Design of efficient arithmetic datapath units for digital image and signal processing

Abstract

Arithmetic operations employed in Very Large Scale Integration (VLSI) newlinecircuits majorly involve both exact and inexact computing. Exact computing newlineis used in high accuracy applications. There are applications that demands newlinelow power, high speed, less area relaxing the high accuracy works well with newlineacceptable results. Approximate or in-exact computing returns possibly newlineinaccurate result rather than a guaranteed accurate result where the system newlineintentionally exposes incorrectness to the application layer in return newlineconserving hardware resource. Currently, approximate computing technology newlineincreases efficiency with accuracy in terms of operating time, area and power. newlineApproximate computation is intended to add approximation levels from the newlinelevel of software, algorithms, architecture, logic or transistors to various newlinedesign levels. newlineThe main concern in approximate computing is the identification of the newlinesection of the application that can be approximated. Mostly it is introduced on newlinenon-critical data. Researchers are working on different layers like device newlinelevel, circuit level and high level abstraction levels to reduce power, area and newlinedelay at the cost of tolerable error. Significant attention to decrease power newlineconsumption, as it determines the system cost, packaging, battery life, thermal newlineand electrical limits of the devices is need of the day. They create space and newlineenergy-efficient structures for battery-powered portable electronics. newlineApproximate circuits are used in image and video processing applications and newlineapplication circuits that handle human sensing signals like hearing, touch, and newlinesmell because a little error in the output does not damage the visual quality or newlinehearing sense.Most of the electronic designs are focused on novel power newlinereduction techniques resulting in saving IC power by a large percentage newline newline