A parallel processing framework for spectral based computations

Today, great advances have been made; however the tenet of ‘design first, figure out how to program later’ still lingers in the corridors of Silicon Valley. The focus of this study is however not on making a contribution to compilers or software development, nor on determining an efficient generic parallel processing architecture for all classes of computing. Instead, this study adopts a different design approach, where a class of computing is first selected and analyzed, before determining a suitable hardware structure which can be tailored to the class being considered. The class of computing under investigation in this work is Spectral Methods, which by its very nature, has its own processing and data communication requirements. The purpose of this study is to investigate the processing and data handling requirements of the Spectral Methods class, and to design a suitable framework to support this class. The approach is different from past traditions - the hardware framework is based on software requirements, and in a sense is designed for the processing required, rather that the other way around.

Grantor dc:publisher.institution

Department of Electrical Engineering

Year dc:date.issued

2012

Author dc:creator

• Van der Byl, Andrew

Advisors dc:contributor.advisor

• Inggs, Michael
• Wilkinson, Richardt H

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/11522

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/11522