Lower bounds on power-dissipation for DSP algorithms

Abstract

Presented in this paper is a fundamental mathematical basis for determining the lower bounds on power dissipation in digital signal processing (DSP) algorithms. This basis is derived f r om information-theoretic arguments. In particular, a digital signal processing algorithm is viewed a s a p r o c ess of information transfer with an inherent information transfer rate requirement of R bits/sec. Dierent architectures implementing a given algorithm are e quivalent to dierent communication networks each with a certain capacity C (also in bits/sec). The absolute lower bound on the power dissipation for any given architecture is then obtained by minimizing the signal power such that its channel capacity C is equal to the desired information transfer rate R. The proposed framework is employed to determine the lower-bounds for simple digital lters. Furthermore, lower bounds on the power dissipation achievable via adiabatic logic is also presented thus demonstrating the versatility of the proposed approach.

Energy-efficient Machine Learning in Silicon: A Communications-inspired Approach

“…A Shannon-inspired approach to error compensation turns out to be most effective. In the past, we have proposed the notion of treating the stochastic circuit fabric as a noisy communication channel (Shanbhag, 1996) and develop Shannon-inspired statistical error compensation (SEC) techniques (see Fig. 3(a)) (Hegde & Shanbhag, 2001;Shim et al, 2004;Varatkar et al, 2010) to compensate for the resulting errors at the algorithmic and architectural levels.…”

Least-square estimation of average power in digital CMOS circuits

“…The power values obtained through simulation, are the inputs for the RLS average power estimator. The estimator, the correlation matrix, and the cross-correlation vector values are calculated at each iteration using (17), (15), and (16), respectively, with the input data. The absolute difference between consecutive estimator values is computed after each iteration.…”