Parallel implicit methods of alternating directions

From MaRDI portal





The paper is devoted to the description of parallel algorithms for solving two-dimensional and three-dimensional difference boundary value problems that use computer systems with a rectangular array of processors with shared memory. The implicit alternating irection methods employing the block method of cyclic reduction without backward sweep are used for solving three-diagonal ``one-dimensional systems. The efficiency of parallelizing is investigated with allowance made for communication overheads. To improve the performance, a two-dimensional bicyclic-network architecture is suggested.











This page was built for publication: Parallel implicit methods of alternating directions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1975092)