Optimization of MLHL-SIM and SIM Algorithm Using OpenMP

Abstract

In this research a parallel version of two existing algorithms that implement Maximum Likelihood Scale Invariant Map (MLHL-SIM) and Scale Invariant Map (SIM) is proposed. By using OpenMP to distribute the independent iterations of for-loops among the available threads, a significant reduction in the computation time for all the experiments is achieved. The higher the size of the considered map is, the higher the reduction of the computation time in the parallel algorithm is. So, for two given datasets, measured times are up to a 29.45 % and a 36.21 % of the sequential time for the MLHL-SIM algorithm. For the SIM algorithm it also reduces the computation time being a 42.09 % and a 36.72 % of the sequential version for the two datasets respectively. Results prove the improvement on the speed up of the parallel version.