Title: A general-purpose framework for FPGA-accelerated genetic algorithms

Authors: Liucheng Guo; Andreea Ingrid Funie; Zhongliu Xie; David Thomas; Wayne Luk

Addresses: Department of Electrical and Electronic Engineering, Imperial College, London, SW7 2AZ, UK ' Department of Computing, Imperial College, London, SW7 2AZ, UK ' Department of Computing, Imperial College, London, SW7 2AZ, UK ' Department of Electrical and Electronic Engineering, Imperial College, London, SW7 2AZ, UK ' Department of Computing, Imperial College, London, SW7 2AZ, UK

Abstract: FPGA-based genetic algorithms (GAs) can effectively optimise complex applications, but require extensive hardware architecture customisation. To promote these accelerated GAs to potential users without hardware design experience, this study proposes a general-purpose automated framework for creating and executing a GA system on FPGAs. This framework contains scalable and customisable hardware architectures while providing a unified platform for different chromosomes. At compile-time, only a high-level input of the target application needs to be provided, without any hardware-specific code being necessary. At run-time, application inputs and GA parameters can be tuned, without time-consuming recompilation, for finding further good configurations of GA execution. The framework was tested on a high performance FPGA platform using nine problems and benchmarks, including the travelling salesman problem, a locating problem and the NP-hard set covering problem. Experiments show the system's flexibility and an average speedup of 29 times over a multi-core CPU.

Keywords: genetic algorithms; field programmable gate arrays; FPGA; automated framework; general-purpose framework; scalable hardware architectures; customisable hardware architectures.

DOI: 10.1504/IJBIC.2015.073183

International Journal of Bio-Inspired Computation, 2015 Vol.7 No.6, pp.361 - 375

Received: 12 Jun 2015
Accepted: 11 Jul 2015

Published online: 26 Nov 2015 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article