| Abstract [eng] |
The goal of this work was to examine existing shared memory parallel programming models, figure out their properties and provide recommendations, which would let us determinate the most fitting model for a given problem. To achieve the intended goal, existing shared memory parallel programming models were analyzed. After that, four problems were solved with each of the parallel programming models. Finally, problems’ speedups were experimentally evaluated and the efficiency of models were identified for each problem. After completing these tasks, it was found that different parallel programming models have varying speedups for different types of problems. Problems with one type of properties could be solved more efficiently with one model, while problems with other properties could be solved more efficiently with another model. Because of that, in the work's conclusions recommendations are given to choose parallel programming model based on the type of the problem and its solution’s code structure. |
