TELKOMNIKA Indonesian Journal of Electrical Engineering
Vol. 12, No. 11, November 2014, pp. 7869 ~ 7875
DOI: 10.11591/telkomnika.v12i11.6640 7869
Received July 10, 2014; Revised September 4, 2014; Accepted September 25, 2014
An Adaptive Genetic Algorithm for Mesh-Based NoC
Application Mapping
Yizhuo Wang, Zhibiao Zhang*
School of Compute Science and Technology, Beijing Institute of Technology,
Beijing, 10081, China
*Corresponding author, e-mail: carldotz@163.com
Abstract
Application mapping is one of the key problems of Network-on-Chip (NoC) design. It maps the
cores of application to the processing elements of the NoC topology. This paper presents a novel
approach for NoC application mapping, which uses adaptive genetic algorithm (AGA) in the mapping. The
proposed approach adaptively varies the probabilities of crossover and mutation operators in genetic
algorithm, aiming to reduce the overall communication cost of NoC. Experimental results show that the
proposed approach decreases the communication cost by 3% to 7% on average, compared to the existing
approach using Standard Genetic Algorithm (SGA).
Keywords: network-on-chip, application mapping, adaptive genetic algorithm, genetic algorithm
Copyright © 2014 Institute of Advanced Engineering and Science. All rights reserved.
1. Introduction
Network-on-Chip (NoC) is the design of modular and scalable communication
architectures where various processing elements are connected to a router-based network
using appropriate network interface [1-3]. Since application Mapping has direct effect on delay,
power consumption and other performance of NoC, it is a very significant aspect of NoC design
[4]. In its general form, the problem of NoC application mapping is an NP-hard problem [4].
Genetic Algorithm (GA), a stochastic search algorithm based on natural genetic
operations provides a solution to the problem of NoC mapping. Sometimes the performance of
Standard Genetic Algorithm (SGA) based NoC mapping cannot meet the need of practical work
because the parameters of genetic operation are fixed values. Therefore, several adaptive
genetic algorithms based on SGA are proposed. The Adaptive Genetic Algorithms (AGA)
proposed by M. Srinivas and L. M. Patnak [5] is considered to be the most representative, and it
realizes the twin goals of maintaining the diversity of population and sustaining the capacity of
convergence with the use of adaptive probabilities of crossover and mutation.
In this paper, we use AGA in the NoC application mapping. First, we extend the
representation for chromosomes proposed by W. Zhou, et al [6] to a general situation in which
the number of application cores is less than or equal to the number of processing elements
(PEs) of the NoC topology. Second, our mapping approach reduces the communication cost
with an improved adaptive genetic algorithm, F-AGA, which is proposed by X. Cheng [7], and it
amends the expressions for crossover probability pc and mutation probability pm to solve the
problem that pc and pm are zero for the solution with the max fitness. Experimental results
show that our approach decreases the communication cost by 3% to 7% on average, compared
to the approach using Standard Genetic Algorithm (SGA).
The rest of this paper is organized as follows. Section 2 introduces the related work in
NoC mapping and GA briefly. Section 3 presents the definitions used in this paper and
describes our AGA based approach. Section 4 shows the experimental results, and Section 5
summarizes our main contribution.
2. Related Work
The NoC application mapping techniques can be classified as dynamic mapping [8-10]
and static mapping, depending on the time at which the cores of application are assigned to the