Comparative Study of Parallel Programming Models to Compute Complex Algorithm

Abstract

The main goal of this research is to use OpenMP, Posix Threads and Microsoft Parallel Patterns libraries to design an algorithm to compute Matrix Multiplication effectively. By using the libraries of OpenMP, Posix Threads and Microsoft Parallel Patterns Libraries, one can optimize the speedup of the algorithm. First step is to write simple program which calculates a predetermined Matrix and gives the results, after compilation and execution of the code. In this stage only single core processor is used to calculate the Matrix multiplication. Later on, in this research OpenMP, Posix Threads and Microsoft Parallel Patterns libraries are added separately and use some functions in the code to parallelize the computation, by using those functions multi-cores of a processor are allowed. Then execute the program and check its run time, then a timer function is added to the code which periodically checks the time it took for the computer to do the parallelization. First the program is run without the Parallel libraries, and then with the OpenMP, Posix Threads and with Microsoft Parallel Patterns libraries code. Then program is executed for each input Matrix size and result is collected. Maximum 5 trials for each input size are conducted and record the time it took for the computer to parallelize the Matrix multiplication. Finally comparison of the performance in terms of execution time and speed up for OpenMP, Posix Threads and Microsoft Parallel Patterns libraries is done using different Matrix Dimensions and different number of processors.