DocumentCode :
1902398
Title :
Research and Implementation of Jacobi Algorithm Based on MPI with Checkerboard Decomposition
Author :
Li, Peizheng ; Cao, Yizhen
Author_Institution :
Comput. Sch., Commun. Univ. of China, Beijing, China
Volume :
3
fYear :
2012
fDate :
23-25 March 2012
Firstpage :
166
Lastpage :
169
Abstract :
In order to achieve parallel algorithm ported from serial algorithm, this paper proposes a simplified model based on MPI checkerboard decomposition. The process of analysis, design and implementation of parallel program will be speed up, by emphasizing data and calculation decomposition. Jacobi algorithm is used as an example to explain this simplified model. The example results show that simplified model is more suitable for implementation of parallel program than Foster´s task/channel model, and checkerboard decomposition has better speedup than row/column decomposition. Furthermore, this simplified model applies to design not only Jacobi algorithm but also other parallel algorithms based on checkerboard decomposition.
Keywords :
Jacobian matrices; application program interfaces; message passing; parallel algorithms; parallel programming; program diagnostics; Jacobi algorithm; MPI checkerboard decomposition; parallel algorithm; parallel program analysis; parallel program design; parallel program implementation; serial algorithm; Algorithm design and analysis; Arrays; Data models; Equations; Jacobian matrices; Mathematical model; Partitioning algorithms; Checkerboard decomposition; HPC; Jacobi algorithm; MPI; Parallel computing;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer Science and Electronics Engineering (ICCSEE), 2012 International Conference on
Conference_Location :
Hangzhou
Print_ISBN :
978-1-4673-0689-8
Type :
conf
DOI :
10.1109/ICCSEE.2012.312
Filename :
6188190
Link To Document :
بازگشت