李天怡 博士 (河南师范大学数学与统计学院)

白中治 研究员

Parallel Decomposition Gauss-Seidel Method for Solving Discretized Subdiffusion Equation With Time Nonuniform Mesh

10月15日（周四）11:00-12:00，南楼733

Efficient parallel solution of large-scale all-at-once linear systems arising from nonuniform temporal discretizations of time-fractional diffusion equations is considered. A parallel iterative solver based on the parallel decomposition Gauss--Seidel (PDGS) method is developed for the resulting systems. By exploiting the block lower-triangular and Toeplitz structures of the coefficient matrix, the PDGS iteration can be implemented efficiently in parallel. The convergence of the proposed iteration method is theoretically analyzed. Numerical experiments demonstrate that the method preserves the accuracy of the underlying all-at-once discretization and achieves good computational efficiency for large-scale problems.