http://swrc.ontoware.org/ontology#Article
Parallel Hierarchical Matrices with Adaptive Cross Approximation on Symmetric Multiprocessing Clusters
en
[Regular Papers] boundary element method, matrix approximation, hierarchical matrices, adaptive cross approximation, parallel scalability, symmetric multiprocessing clusters
ACCMS, Kyoto University / JST CREST
JST CREST / Information Initiative Center, Hokkaido University
JST CREST / Department of Electrical Engineering, Kyoto University
JST CREST / Department of Electrical Engineering, Doshisha University
Akihiro Ida
Takeshi Iwashita
Takeshi Mifune
Yasuhito Takahashi
We discuss a scheme for hierarchical matrices with adaptive cross approximation on symmetric multiprocessing clusters. We propose a set of parallel algorithms that are applicable to hierarchical matrices. The proposed algorithms are implemented using the flat-MPI and hybrid MPI+OpenMP programming models. The performance of these implementations is evaluated using an electric field analysis computed on two symmetric multiprocessing cluster systems. Although the flat-MPI version gives better parallel scalability when constructing hierarchical matrices, the speed-up reaches a limit in the hierarchical matrix-vector multiplication. We succeeded in developing a hybrid MPI+OpenMP version to improve the parallel scalability. In numerical experiments, the hybrid version exhibits a better parallel speed-up for the hierarchical matrix-vector multiplication up to 256 cores.
We discuss a scheme for hierarchical matrices with adaptive cross approximation on symmetric multiprocessing clusters. We propose a set of parallel algorithms that are applicable to hierarchical matrices. The proposed algorithms are implemented using the flat-MPI and hybrid MPI+OpenMP programming models. The performance of these implementations is evaluated using an electric field analysis computed on two symmetric multiprocessing cluster systems. Although the flat-MPI version gives better parallel scalability when constructing hierarchical matrices, the speed-up reaches a limit in the hierarchical matrix-vector multiplication. We succeeded in developing a hybrid MPI+OpenMP version to improve the parallel scalability. In numerical experiments, the hybrid version exhibits a better parallel speed-up for the hierarchical matrix-vector multiplication up to 256 cores.
AA00700121
Journal of information processing
22
4
642-650
2014-10-15
1882-6652