In Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

Shweta Jha

will defend her dissertation proposal

Performance Tuning of Parallel Applications

Abstract

The goal of high performance computing is executing very large problems in least amount of time, typically by deploying parallelization techniques.  However, introducing parallelization to an application also introduces synchronization and communication overhead, which in turn creates a performance bottleneck. Performance tuning can help to ease this bottleneck and improve the overall performance of the application.

There are two aspects of an application which can be improved using performance tuning, namely, computational section and the communication section. The time spent in communication operations is a major factor in determining the scalability of parallel applications. Tuning the parameters of a communication library can be used to adapt its characteristics to a particular platform, minimizing the communication time of an application.

The goal of this dissertation is to improve the performance a parallel application by adjusting its communication parameter to a given application. Specifically, we introduce the notion of  a personalized MPI library, highlighting the necessity  and the methodology according to which each application needs to have a communication library tuned for the particular platform. Secondly, this dissertation contributes towards the theoretical understanding of impact and limitations of point-to-point communication performance on collective communication and the overall application. Work presented here is based on the Open M PI communication library, which has a large number of runtime parameters that can be modified without having to recompile the MPI library or the application.

Date: Friday, December 11, 2015
Time: 11:00 AM
Place: PGH 501D
Advisor: Dr. Edgar Gabriel

Faculty, students, and the general public are invited.