期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2018
卷号：96
期号：19
出版社：Journal of Theoretical and Applied
摘要：Investigating the multi-core architecture is an essential issue to get superior in parallel reenactments. However, the simulation highlights must fit on parallel programming model to build the execution. The main goal of this research is to choose and evaluate parallelism using OpenMP over sequential program. For this purpose, there is a portrayal of two searching algorithms. The calculation is to discover the next edge of Prim's algorithm and single source shortest way of Dijkstra's algorithm. These two algorithm actualized in sequential formulation. Parallel searching algorithms are then implemented in view of multicore processor. The speed-up ratio and efficiency of parallel searching algorithms are tested and investigated in SGEMM GPU Kernel performance dataset with 241600 records and 18 attributes. Results show the dataset with different data sizes achieved super linear speed-up ratio and efficiency on OpenMP by running on 4 cores processor and reduction of the running time over sequential program. More importantly, the new implementation drastically decreases the time of execution for thread 8 for Prims algorithm from 5.16ms to 1.48 ms for Dijkstra algorithm. Parallel calculation is impressively powerful for huge graph size. General outcome shows that multi-threaded parallelism is exceptionally successful to accomplish better performance for dataset based on searching algorithms by separating the primary dataset into sub-datasets to increase diversity on arrangement investigation.
