Performance Analysis of Alternative Open Source Parallel Computing Approach to OpenMP on Multicore Processors

In this era of evolving computing field, parallel computing is one of the fastest changing fields. There have been numerous researches made and still the research is going on in the areas of data decomposition, parallel algorithms to get more performance through parallelism. In this paper we have achieved parallelism by using an open source made available with the name libdispatch (implementation of Grand Central Dispatch Services). This package has been ported to the FreeBSD, however this can also be used under open source environment. In this paper we have experimented with matrix multiplication in sequential and parallel programs. The categories of experimentation done are: Sequential Programming with GCC and Clang; Parallel programming with OpenMP and clang with dispatcher. Better performance has been observed in Clang for both sequential and parallel computations