Title :
Investigating efficient parallelization techniques for the characteristic basis function method (CBFM)
Author :
Ludick, D.J. ; Davidson, D.B.
Author_Institution :
Dept. Electr. & Electron. Eng., Univ. Stellenbosch, Matieland, South Africa
Abstract :
Two message-passing parallelization schemes for the Method of Moment (MoM) based Characteristic Basis Function Method (CBFM) are presented. The CBFM algorithm is implemented in the high-level scripting languages, Python and Octave and is parallelized using the Message Passing Interface (MPI) standard. The MPI implementations used for the parallelization, i.e. MPI for Python and MPI Toolbox for Octave and Matlab, are compared in terms of bandwidth, latency, speed-up and efficiency. The results are obtained from simulations conducted on an IBM e1350 Linux Cluster.
Keywords :
application program interfaces; authoring languages; computational electromagnetics; message passing; method of moments; parallel programming; CBFM algorithm; IBM e1350 Linux cluster; MoM; Octave; Python; characteristic basis function method; electromagnetic problem; high-level scripting language; message passing interface; message-passing parallelization scheme; method-of-moment; Bandwidth; Clustering algorithms; Delay; Distributed computing; Electronic mail; Equations; Linux; Mathematical model; Matrix decomposition; Packaging;
Conference_Titel :
Electromagnetics in Advanced Applications, 2009. ICEAA '09. International Conference on
Conference_Location :
Torino
Print_ISBN :
978-1-4244-3385-8
Electronic_ISBN :
978-1-4244-3386-5
DOI :
10.1109/ICEAA.2009.5297410
