On Parallel Application of Triangular Iterative Methods in a Special-Purpose Multi-Processor System

Options of parallel run of triangular iterative methods in a special-purpose matrix multi-processor system are considered. Various decompositions of the grid domain into subdomains and storing the grid information in node processors lead to an efficiency of over 0,95. For elliptical problems and large grids ($N\geq 500$) the most economical is a modified alternating-triangular method.