Energy-efficiency of software and hardware algorithms

Maja H. Kirkeby1, Thomas Krabben1, Mathias Larsen1, Maria B. Mikkelsen1, Mads Rosendahl1, Martin Sundman1, Tjark Petersen2 and Martin Schoeberl2

  1. Department of People and Technology, Roskilde University
    Roskilde, Denmark
    majaht@ruc.dk, krabben@ruc.dk, mamaar@ruc.dk, mariabm@ruc.dk, madsr@ruc.dk, sundman@ruc.dk
  2. DTU Compute, Technical University of Denmark
    Lyngby, Denmark
    s186083@student.dtu.dk, masca@dtu.dk

Abstract

In this article, we compare the energy efficiency of hardware and software implementations of Heapsort and Dijkstra’s algorithm for route finding. The software implementations are written in C for Raspberry Pi, and the hardware implementations are crafted in Chisel for an FPGA. Our objective is to examine how we can fairly compare energy efficiency between hardware and software. This study seeks to identify circumstances where time and energy efficiency diverge, providing preliminary insights that inform hardware selection. Our findings serve as a step towards understanding the complex trade-offs in algorithm performance across different computational platforms.

Key words

energy efficiency, performance, FPGA, CPU, algorithms

Full text

DownloadAvailable in PDF
Portable Document Format

How to cite

Kirkeby, M. H., Krabben, T., Larsen, M., Mikkelsen, M. B., Rosendahl, M., Sundman, M., Petersen, T., Schoeberl, M.: Energy-efficiency of software and hardware algorithms. Computer Science and Information Systems