Evolution of a Parallel Performance System Allen D. Malony, Sameer Shende, Alan Morris, Scott Biersdorff, Wyatt Spear, Kevin Huck, and Aroon Nataraj Abstract The TAU Performance System R  is an integrated suite of tools for instrumentation, measurement, and analysis of parallel programs target- ing large-scale, high-performance computing (HPC) platforms. Representing over fifteen calendar years and fifty person years of research and development effort, TAU’s driving concerns have been portability, flexibility, interoperabil- ity, and scalability. The result is a performance system which has evolved into a leading framework for parallel performance evaluation and problem solv- ing. This paper presents the current state of TAU, overviews the design and function of TAU’s main features, discusses best practices of TAU use, and outlines future development. 1 Introduction Scalable parallel systems have always evolved together with the tools used to observe, understand, and optimize their performance. Next-generation par- allel computing environments are guided to a significant degree by what is known about application performance on current machines and how per- formance factors might be influenced by technological innovations. State-of- the-art performance tools play an important role in helping us understand application performance and allowing us to focus our attention on future per- formance concerns. Therefore, performance technology must keep pace with the growing complexity of next-generation parallel platforms if they are to contribute to the present and future promises of high-end parallel computing (HEC). We will need a robust performance frameworks that can provide both flexible and portable empirical performance observation capabilities at all lev- els of a system. In short, mapping low-level behavior to high-level performance Performance Research Lab, University of Oregon, Eugene, OR, e-mail: {malony,sameer, amorris,scottb,wspear,khuck,anataraj}@cs.uoregon.edu 1