Leveraging Error Resilience of Iterative Algorithms for Energy Efficiency: from Concept to Implementation

TL;DR

This paper demonstrates how leveraging the inherent error resilience of iterative algorithms can significantly improve energy efficiency in digital signal processing, through a novel heterogeneous accelerator design.

Contribution

It introduces an adaptive approximation model and a heterogeneous architecture that processes initial iterations approximately and later iterations accurately, achieving energy savings without affecting convergence.

Findings

23% reduction in energy consumption with the proposed accelerator

Adaptive approximation enables energy-efficient iterative processing

No increase in iteration count compared to accurate methods

Abstract

Iterative algorithms are widely used in digital signal processing applications. With the case study of radio astronomy calibration processing, this work contributes towards revealing and exploiting the intrinsic error resilience of iterative algorithms for energy efficiency benefits. We consider iterative methods that use a convergence criterion as a quality metric to terminate the iterative computations. We propose an adaptive statistical approximation model for high-level resilience analysis that provides an opportunity to divide an iterative algorithm into exact and approximate iterations. We realize an energy-efficient accelerator based on a heterogeneous architecture, where the heterogeneity is introduced using accurate and approximate processing cores. Our proposed methodology exploits the error-resilience of the algorithm, where initial iterations are processed on approximate modules while the later ones on accurate modules. The proposed accelerator design does not increase the number of iterations as compared to that of an accurate counterpart and provides sufficient precision to converge to an acceptable solution. Our implementation using TSMC 40nm Low Power (TCBN40LP) technology shows 23% savings in electrical energy consumption.