OPTIMIZATION OF THE PARALLEL ALGORITHM BY REDUCING THE TIME FOR INTERPROCESSOR DATA EXCHANGE
In the article results of researches in the field of construction of parallel algorithms effective for time and volume of computing resources for dis-tributed memory computing systems are presented. A method that allows to improve the schedule of parallel algorithm execution time at the ex-pense of redistribution of operations between processes. As a result, the amount of transmitted messages between processes decreases and the time spent on interprocessor communication is reduced. In implementing the method of volume of computing resources does not change. The method is based on the adjacency list, corresponding to the information graph of the algorithm. The method can be applied in combination with other methods for constructing a schedule for the execution of algorithms, for example, oriented to optimizing the number of computational nodes, to achieve an optimal schedule for a number of parameters.
Authors: Yu. A. Shichkina, M. H. A. Al-Mardi, M. S. Kupriyanov
Direction: Automation and Control
Keywords: Parallel algorithm, algorithm optimization, information graph, operation execution time, algorithm execution schedule, pro-cess, processor, interprocessor data transmission
View full article