Parallel Programming Recipes

Parallel programming has become vital for the success of commercial applications since Moore’s Law will now be used to double the processors (or cores) per chip every technology generation. The performance of applications depends on how software executions can be mapped on the multi-core chip, and how efficiently they run the cores. Currently, the increase of parallelism in software development is necessary, not only for taking advantage of multi-core capability, but also for adapting and surviving in the new silicon implementation. This project will provide the performance characteristics of parallelism for some common algorithms or computations using different parallel languages. Based on concrete experiments, where each algorithm is implemented on different languages and the program’s performance is measured, the project provides the recipes for the problem computations. The following are the central problems and algorithms of the project: Arithmetic Algebra: Maclaurin Series Calculation for ex, Dot-Product of Two Vectors: each vector has size n; Sort Algorithms: Bubble sort, Odd-Event sort; Graphics: Graphics rendering. The languages are chosen based on commonality in the current market and ease of use; i.e., OpenMP, MPI, and OpenCL. The purpose of this study is to provide reader a broad knowledge about parallel programming, the comparisons, in terms of performance and implementation cost, across languages and application types. It is hoped to be very useful for programmers/computer-architects to decide which language to use for a certain applications/problems and cost estimations for the projects. Also, it is hoped that the project can be expanded in the future so that more languages/technologies as well as applications can be analyze