Infrastructure For Performance Tuning MPI Applications

Abstract

Clusters of workstations are becoming increasingly popular as a low-budget alternative for supercomputing power. In these systems, message-passing is often used to allow the separate nodes to act as a single computing machine. Programmers of such systems face a daunting challenge in understanding the performance bottlenecks of their applications. This is largely due to the vast amount of performance data that is collected, and the time and expertise necessary to use traditional parallel performance tools to analyze that data.The goal of this project is to increase the level of performance tool support for message-passing application programmers on clusters of workstations. We added support for LAM/MPI into the existing parallel performance tool, Paradyn. LAM/MPI is a commonly used, freely-available implementation of the Message Passing Interface (MPI), and also includes several newer MPI features, such as dynamic process creation.In addition, we added support for non-shared filesystems into Paradyn and enhanced the existing support for the MPICH implementation of MPI. We verified that Paradyn correctly measures the performance of the majority of LAM/MPI programs on Linux clusters and show the results of those tests. In addition, we discuss MPI-2 features that are of interest to parallel performance tool developers and design support for these features for Paradyn.