Using Local Iterations to Increase the Efficiency of the Parallel Jacobi Algorithm

📄 Project Abstract

The aim of this project is to determine whether the efficiency of an MPI based parallel algorithm can be improved by reducing communication. The main focus on reducing communication is achieved by implenting local iterations in the chosen Jacobi algorithm. Generally, a standard parallel algorithm is a more efficient implementation of the same serial algorithm but functionally identical, it performs the exact algorithm. This project attempts to introduce a more efficient parallel Jacobi algorithm, by relaxing the constraint that it is formally identical to the serial version, though it still solves the same problem. This project extends the taught Modules[2] and[3] 1.2 Objectives * Investigate how performing local iterations affects the efficiency of the parallel Jacobi algorithm. * Investigate how the computed solution is altered by the implementation of local iterations. * Form a conclusion regarding the advantages and disadvantages of a locally iterative approach. 1.3 Minimum Require...