Measuring the impact of input data on energy consumption of software

TL;DR

This paper investigates how input data influences energy consumption in embedded software and proposes a method to incorporate this effect into energy models, aiding in more accurate energy prediction.

Contribution

It introduces a novel approach to account for input data variability in software energy models for embedded systems.

Findings

Input data significantly affects energy consumption.

Proposed method improves energy modeling accuracy.

Guidelines for energy-aware software design.

Abstract

The amount of energy consumed during the execution of software, and the ability to predict future consumption, is an important factor in the design of embedded electronic systems. In this technical report I examine factors in the execution of software that can affect energy consumption. Taking a simple embedded software benchmark I measure to what extent input data can affect energy consumption, and propose a method for reflecting this in software energy models.