Title :
Impact of Two Factors on Several Domain Decomposition Based Parallel Incomplete Factorizations for the Meso-scale Simulation of Concrete
Author :
Jian-ping, Wu ; Jun, Zhao ; Jun-qiang, Song ; Xiao-mei, Li
Author_Institution :
Sch. of Comput. Sci., Nat. Univ. of Defense Technol., Changsha, China
Abstract :
It is an efficient technique to parallelize incomplete factorizations based on domain decomposition. In this paper, three are considered for solving the derived sparse linear systems from the meso-scale simulation of concrete, including the block Jacobi analog, the classical additive Schwarz, and the so-called factors combination method. For the grid may be of unstructured ones, the graph partitioning method is exploited to perform domain decomposition. The different partitioning schemes are compared and analyzed. The results show that the communication structure is one of the vital factors for the efficiency but not the all. Further, the degree of overlapping is also tested and the results show that for the current simulation, the degree of 1 is enough.
Keywords :
concrete; graph theory; parallel processing; structural engineering computing; additive Schwarz; block Jacobi analog; called factors combination method; domain decomposition; graph partitioning method; meso-scale concrete simulation; parallel incomplete factorizations; sparse linear systems; Additives; Ash; Capacitive sensors; Computational modeling; Concrete; Convergence; Linear systems; Sparse matrices; Symmetric matrices; Testing; additive Schwarz; domain decomposition; graph partitioning; incomplete factorization; parallel preconditioning;
Conference_Titel :
Information and Computing (ICIC), 2010 Third International Conference on
Conference_Location :
Wuxi, Jiang Su
Print_ISBN :
978-1-4244-7081-5
Electronic_ISBN :
978-1-4244-7082-2
DOI :
10.1109/ICIC.2010.146