Comparative Analysis of Automatic Parallelization Techniques

Abstract

With multicore design emerging as the chief design pattern for the microprocessor industry in a major way constant efforts are being made to improve its performance. The only way to extract maximum performance from a multicore based system as of now is through the automatic extraction of threads from sequential applications with the assistance of tools including, compilers and runtime optimizers. The increasing popularity of clusters and other similar forms of distributed computing calls for the requirement of automatic parallelization in compilers. This requirement for techniques through which automatic parallelization can be achieved has set of the development of various different methods with varying efficiencies. The objective of this paper is to test the efficiencies of two such techniques by performing a comparative analysis of the techniques based on various criteria like time complexity for instance. The techniques we are going to work on are namely the Scalar and Array Analysis technique and the Smith waterman technique. As mentioned previously these techniques will be analyzed using various criteria to arrive at a conclusion as to which one of them is superior performance wise.