Over the last decades an incredible amount of resources has been devoted to building ever more powerful supercomputers. However, exploiting the full capabilities of these machines is becoming exponentially more difficult with each new generation of hardware. To help understand and optimize the behavior of massively parallel simulations the performance analysis community has created a wide range of tools and APIs to collect performance data, such as flop counts, network traffic or cache behavior at the largest scale. However, this success has created a new challenge, as the resulting data is far too large and too complex to be analyzed in a straightforward manner. Therefore, new automatic analysis approaches must be developed to allow application developers to intuitively understand the multiple, interdependent effects that their algorithmic choices have on the final performance.

The natural first step towards automatic analysis is to visualize the collected data to provide some insight into general trends. This helps both application developers and performance experts to form new hypotheses on potential causes of and solutions for performance problems. Furthermore, intuitive visualizations are highly effective in conveying the results of any analysis and thus are a valuable tool throughout the entire process. Unfortunately, visualizing performance data has proven challenging as the information is highly abstract, non-spatial, and often categorical. While some early attempts of visualizations in performance tools have been proposed, these are rudimentary at best and have not found widespread adoption.

At the same time the information visualization and visual analytics community is developing techniques to visualize, illustrate, and analyze complex, non-spatial data. This has led to new general design principles of visualization tools, color spaces, and user interfaces as well as a wide array of common techniques to tackle a broad range of applications. Unfortunately, so far the overlap between these communities is limited even though both areas could gain significantly from a closer collaboration.

Performance analysis is quickly reaching a stage where highly advanced analysis and visualization techniques will be mandatory rather than extravagant optional components. At the same time, visualization researchers are continuously looking for new application areas that challenge the expressive power of their techniques. Together these fields have the potential to establish the field of performance visualization at the intersection of performance analysis and classical visualization and thereby to significantly impact the future of high performance computing.

This Dagstuhl Perspectives Workshop will gather leading experts from both performance analysis and information visualization to:

1. Introduce experts in each area to the relevant state of the art of the other;
2. Provide a forum to form new collaborations; and
3. Discuss medium term strategies to bring the joined work to the attention of funding agencies.

We intend to start the Perspectives Workshop with several overview talks in each fields to create a common ground on the type of data available, the biggest open challenges, and potentially relevant existing work. Each of these will be followed by an open discussion aimed at ensuring that sufficient background knowledge is provided to all participants. These will be followed by breakout sessions focused on particular topic areas such as multi-dimensional analysis, graph drawing, or correlation analysis. The last portion of the workshop will focus on concrete steps to bring joined work to the attention of the various funding agencies in both fields. Given the potential impact of performance visualization on high performance computing the goal is to shape a new research area at the intersection of performance analysis and visualization.

• Abhinav Bhatele (LLNL - Livermore, US) [dblp]
• Peer-Timo Bremer (LLNL - Livermore, US) [dblp]
• Holger Brunst (TU Dresden, DE) [dblp]
• Hans-Joachim Bungartz (TU München, DE) [dblp]
• Remco Chang (Tufts University - Medford, US) [dblp]
• Hank Childs (University of Oregon - Eugene, US) [dblp]
• Todd Gamblin (LLNL - Livermore, US) [dblp]
• Markus Geimer (Jülich Supercomputing Centre, DE) [dblp]
• Judit Gimenez (Barcelona Supercomputing Center, ES) [dblp]
• Hans Hagen (TU Kaiserslautern, DE) [dblp]
• Daniel A. Keim (Universität Konstanz, DE) [dblp]
• Joshua A. Levine (Clemson University, US) [dblp]
• Naoya Maruyama (RIKEN - Kobe, JP) [dblp]
• Bernd Mohr (Jülich Supercomputing Centre, DE) [dblp]
• Christopher Muelder (University of California - Davis, US) [dblp]
• Klaus Mueller (Stony Brook University, US) [dblp]
• Matthias S. Müller (RWTH Aachen, DE) [dblp]
• Wolfgang E. Nagel (TU Dresden, DE) [dblp]
• Valerio Pascucci (University of Utah, US) [dblp]
• Ulrich Rüde (Universität Erlangen-Nürnberg, DE) [dblp]
• Carlos E. Scheidegger (AT&T Labs Research - New York, US) [dblp]
• Tobias Schreck (Universität Konstanz, DE) [dblp]
• Martin Schulz (LLNL - Livermore, US) [dblp]
• Derek Xiaoyu Wang (University of North Carolina - Charlotte, US) [dblp]
• Felix Wolf (GRS for Simulation Sciences - Aachen, DE) [dblp]