Study of approaches to parallel computing problems

Abstract

Many applications today require more computing power than traditional uniprocessors can offer Parallel processing provides a cost effective solution to this problem by increasing the number of CPUs in a computer and by adding an efficient communication system between them The workload can be shared between different processors This results in much higher computing power and performance than could be achieved with newlinetraditional single processor systems This work studies how standard problems can be run on multiprocessor machines to improve performance Initially limited parallelism is newlineexamined with the help of the IBM Cell Broadband Engine processor simulator Cell Software Development kit version 3 1 which runs on Fedora Operating system The problems that are implemented on the Cell BE processor are search sort and lexical analyzer All the problems considered cater to data parallelism wherein large data is split into many chunks operated on by different processors running the same code and the individual results are combined The parameter measured for comparison is the speed of execution newline newline