BEGIN:VCALENDAR
PRODID:-//Microsoft Corporation//Outlook MIMEDIR//EN
VERSION:1.0
BEGIN:VEVENT
DTSTART:20111115T220000Z
DTEND:20111115T221500Z
LOCATION:TCC LL1
DESCRIPTION;ENCODING=QUOTED-PRINTABLE:ABSTRACT: Recent trends involving multicore processors, Graphical Processing Units (GPU) focus on exploiting task- and thread-level parallelism. In this research, we analyze various aspects of performances of these leading architectures including NVIDIA and AMD/ATI Graphical Processing Units, multicore processors such as Intel multicore, AMD multicore, IBM’s Cell Broadband Engine, and SUN’s T2+ UltraSPARC. A maximum of 32-core multicore (Intel and AMD) and 1600-core AMD GPGPU were utilized in this research. The case study used in this thesis is a biological Spiking Neural Network (SNN), implemented with the Izhikevich, Wilson, Morris-Lecar, and Hodgkin-Huxley neuron models. We report and analyze the variations of performance with network scaling, available optimization techniques and execution configuration. Based on the performance analysis of various architectures, a Fitness performance model is proposed and verified with the SNN implementation results. The Fitness performance model predicts the suitability of architecture for accelerating an application.
SUMMARY:Performance Analysis and Fitness of GPGPU and Multicore Architectures for Scientific Applications
PRIORITY:3
END:VEVENT
END:VCALENDAR