Jeldakov

This article describes a parallel algorithm for solving systems of linear algebraic equations with a multi-diagonal (band) coefficient matrix. A scheme of the algorithm for multi-processor shared-memory computing systems is also presented in the article. Data structures for a compact storage of sparse band matrices were implemented. Theoretical and experimental investigations of efficiency of the algorithm were conducted using various problems. Dependences of the program’s speedup on various parameters of the problem and the number of processors were given in this work. The conclusion presents pros and cons of the algorithm.