- Journal Home
- Volume 22 - 2025
- Volume 21 - 2024
- Volume 20 - 2023
- Volume 19 - 2022
- Volume 18 - 2021
- Volume 17 - 2020
- Volume 16 - 2019
- Volume 15 - 2018
- Volume 14 - 2017
- Volume 13 - 2016
- Volume 12 - 2015
- Volume 11 - 2014
- Volume 10 - 2013
- Volume 9 - 2012
- Volume 8 - 2011
- Volume 7 - 2010
- Volume 6 - 2009
- Volume 5 - 2008
- Volume 4 - 2007
- Volume 3 - 2006
- Volume 2 - 2005
- Volume 1 - 2004
Cited by
- BibTex
- RIS
- TXT
The multilevel fast multipole algorithm (MLFMA) has shown great efficiency in solving large scale electromagnetic scattering problems. However, when unknowns become up to tens of millions, it is not trivial to keep high performance because of the complicated structure and calculation of MLFMA. In order to get rid of the bottleneck caused by load balancing, a parallel data partitioning strategy is proposed based on the hierarchical structure of an oct-tree of MLFMA. We present our data partitioning strategy in the light of different layers' properties including the processing of three kinds of layers in the tree and a fine-grained decomposition. We also put forward a solution of a coexisting data correlating problem, using a transition layer. Meanwhile, with the purpose of minimizing communication time in distributed memory system, a redundant technique is applied in the distributed layer. Parallel efficiency analysis demonstrates that the computational cost in parallelization of MLFMA can be asymptotically cut, and a high parallel efficiency can be obtained in our implementation.
}, issn = {2617-8710}, doi = {https://doi.org/}, url = {http://global-sci.org/intro/article_detail/ijnam/625.html} }The multilevel fast multipole algorithm (MLFMA) has shown great efficiency in solving large scale electromagnetic scattering problems. However, when unknowns become up to tens of millions, it is not trivial to keep high performance because of the complicated structure and calculation of MLFMA. In order to get rid of the bottleneck caused by load balancing, a parallel data partitioning strategy is proposed based on the hierarchical structure of an oct-tree of MLFMA. We present our data partitioning strategy in the light of different layers' properties including the processing of three kinds of layers in the tree and a fine-grained decomposition. We also put forward a solution of a coexisting data correlating problem, using a transition layer. Meanwhile, with the purpose of minimizing communication time in distributed memory system, a redundant technique is applied in the distributed layer. Parallel efficiency analysis demonstrates that the computational cost in parallelization of MLFMA can be asymptotically cut, and a high parallel efficiency can be obtained in our implementation.