Benchmarking CPUs and GPUs on embedded platforms for software receiver usage

Abstract

Smartphones containing multi-core central processing units (CPUs) and powerful many-core graphics processing units (GPUs) bring supercomputing technology into your pocket (or into our embedded devices). This can be exploited to produce power-efficient, customized receivers with flexible correlation schemes and more advanced positioning techniques. For example, promising techniques such as the Direct Position Estimation paradigm or usage of tracking solutions based on particle filtering, seem to be very appealing in challenging environments but are likewise computationally quite demanding. This article sheds some light onto recent embedded processor developments, benchmarks Fast Fourier Transform (FFT) and correlation algorithms on representative embedded platforms and relates the results to the use in GNSS software radios. The use of embedded CPUs for signal tracking seems to be straight forward, but more research is required to fully achieve the nominal peak performance of an embedded GPU for FFT computation. Also the electrical power consumption is measured in certain load levels.