EVALUATING THE EFFECTIVENESS OF THE IMPLEMENTATION OF COMPUTATIONAL ALGORITHMS USING THE ОреnМР STANDARD FOR PARALLELIZING PROGRAMS.

The relevance of the work lies in the need to increase the efficiency of computer modeling through the use of progressive hardware and software parallelization and synchronization of calculations on modern computers with multi-core architecture. The purpose of the work is to increase the speed of computational algorithms of Thomas algorithm by using advanced technologies for programming parallel computing. Serial and parallel algorithms for software implementation of Thomas algorithm have been developed; a comparative evaluation of the implementation efficiency (speed of execution) of these algorithms by means of an open standard for parallelizing OpenMP programs for a significant (up to 5×107) SLAE order was performed. The use of progressive programming technologies in the implementation of the indicated methods for solving SLAEs made it possible to increase the computation speed by 1,9-2,9 times. The obtained results correspond with the known literature data. In this work, the time values of the software implementation of sequential and computational algorithms of Thomas algorithm for SLAEs of significant order were obtained for sequentially and parallelized into two streams using OpenMP tools. The expediency of such parallelization for SLAE of the order of more than 2,5×105 is shown. The practical significance of the work lies in the use of the results obtained in the numerical study of stationary and non-stationary, linear and nonlinear processes in mathematical modeling problems, where the multiple solution of a significant order SLAE is the most resource-intensive stage. [ABSTRACT FROM AUTHOR]