作者:
DOI: 10.3745/KIPSTA.2003.10A.2.137
关键词:
摘要: Recently, CG (Conjugate Gradient) scheme for the optimization of Rayleigh quotient has been proven a very attractive and promising technique interior eigenvalues following eigenvalue problem, Ax=λx (1) The given matrix A is assummed to be large sparse, symmetric. Also, method amenable parallel computations. proper choice preconditioner significantly improves convergence scheme. We compare preconditioners computation symmetric by CG-type method. considered are Point-SSOR, ILU (0) in multi-coloring order, Multi-Color Block SSOR (Symmetric Succesive OverRelaxation). conducted our experiments on CRAYT3E with 128 nodes. MPI (Message Passing Interface) library was adopted interprocessor communications. test matrices up dimensions were created from discretizations elliptic PDE. All things MC-BSSOR seems most robust preconditioner.