Parallel FDTD Computations Optimized by Genetic Algorithms

Author

Smyk, Adam ; Tudruj, Marek

Author_Institution

Polish-Japanese Inst. of Inf. Technol., Warsaw, Poland

fYear

2011

fDate

3-7 April 2011

Firstpage

99

Lastpage

104

Abstract

The Finite Difference Time Domain method (FDTD) enables simulating electromagnetic wave propagation. To speed up FDTD computations in a multiprocessor system we apply an optimized distribution of the computational load between available processors. The optimization algorithm is based on the analysis of macro data flow graphs of the FDTD programs combined a with a genetic algorithm (GA. We have tested two different GAs with different chromosome structures. We have also analyzed the influence of the kind of genetic operators on the load distribution quality for a given FDTD simulation area. Simultaneously, we have introduced and tested several improvements like a mixed fitness function, a re-selection and compensation operations.

Keywords

finite difference time-domain analysis; genetic algorithms; mathematics computing; parallel algorithms; FDTD programs; chromosome structures; electromagnetic wave propagation; finite difference time domain method; genetic algorithms; genetic operators; load distribution quality; macro data flow graphs; mixed fitness function; multiprocessor system; optimization algorithm; optimized distribution; parallel FDTD computations; Biological cells; Finite difference methods; Flow graphs; Genetic algorithms; Merging; Optimization; Time domain analysis; FDTD; genetic algorithms; graph partitioning; optimization;

fLanguage

English

Publisher

ieee

Conference_Titel

Parallel Computing in Electrical Engineering (PARELEC), 2011 6th International Symposium on

Conference_Location

Luton

Print_ISBN

978-1-4577-0078-1

Electronic_ISBN

978-0-7695-4397-0

Type

conf

DOI

10.1109/PARELEC.2011.28

Filename

5770409