Abstract of Talk
Performance Tuning of an Embedded Template Analysis Tool
Abstract: As multi core processors become ubiquitous, programs must be designed to utilize the additional power at their disposal. Extracting maximum performance out of these processors requires utilizing latest parallel programming techniques as well as using the latest tools available to increase performance. This thesis delves into enhancing performance of existing programs by applying parallel techniques, utilizing performance analyzers and various other tools developed specially for multi-core system programming. We discuss these techniques and tools and also demonstrate how a Template Analysis program was modified to take advantage of the additional power a multi-core machine has over single core machines to enable its use in an embedded application. The overall efficiency of the application was increased using these techniques, and then tested using several data sets as well as various core machines. More information was gathered using this data and the program was fine tuned to extract maximum performance. Additional considerations were made towards developing a real time Template Analysis program.