Title of Presentation: Parallel Applications for the Masses

Primary (Corresponding) Author: Andrew Connolly

Organization of Primary Author: University of Pittsburgh


Abstract:  With the new generation of astrophysical simulations and surveys the community is facing the challenge of how to analyze the ever increasing data volume. Developing algorithms that scale efficiently with size of data set and number of processors requires substantial expertise in parallel programming. The steep learning curve associated with this development limits our ability to extract information for the surveys in which we invest many hundreds of man years of work. In this talk I will discuss the development of algorithms for astronomy ranging from correlation functions to the tracking of moving sources that can be implemented on single and multiprocessor machines. I will describe a framework that we have development to ease the movement of serial algorithms to massively parallel systems and how this might benefit machine learning in astrophys ics.