Given the exponential increase in the complexity of nowadays parallel systems, parallel applications often fail to exploit the full power of the underlying hardware. At scale, it is not uncommon for applications to run at parallel efficiencies in the low single digits. Moreover, their optimization is extremely difficult due to the inherent complexity of the systems and of the applications themselves. Therefore, a variety of projects aim to develop tools for the measurement, analysis, and visualization of parallel program performance in order to help and guide users in the optimization process.

This meeting was the third in a series of seminars related to the topic “Performance Analysis of Parallel and Distributed Programs”, with previous meetings being the Dagstuhl Seminar 02341 on “Performance Analysis and Distributed Computing” held in August 2002 and Seminar 05501 on “Automatic Performance Analysis” in December 2005. While these seminars concentrated on the “analysis” part of performance analysis, at the most recent seminar the focus was on the building blocks of program instrumentation and modeling that are prerequisites for the analysis phase. As a result, the presentations of the participants concentrated on several fundamental issues related to instrumentation for generating highquality performance data, methodologies for performance modeling leading to accurate predictions for the performance, and on the ways in which these techniques are combined for the performance analysis of applications and systems.

The program consisted of 28 presentations and practical tool demonstrations as well as two “open mic” sessions where time was set aside for spontaneous discussions and ”brain storming”. The seminar brought together a total of 48 researchers and developers working in the area of performance from universities, national research laboratories and, especially important, from three major computer vendors. The goals were to increase the exchange of ideas, knowledge transfer, foster a multidisciplinary approach to attacking this very important research problem with direct impact on the way in which we design and utilize parallel systems to achieve high application performance.

The presentations can be grouped thematically as follows:

• Session “Performance Analysis in General”
• Session “Instrumentation”
• Session “Modeling”
• Session “Scalability”
• Session “Tools”
• “Open Mic” Sessions
• Session “Short Announcements”

Despite the larger than normal number participants, the seminar was very successful due to the dedicated professionalism and discipline of the participants on one side and the very helpful and professional staff of Dagstuhl on the other side. Lively discussions and spontaneous computer demonstrations continued every day well beyond midnight. It is important to note that the group meeting and residential aspects of Dagstuhl and the five-day format provide a continuity of thought and discussion unavailable in other conference, workshop, or meeting settings. At Dagstuhl, we have time for considered (and reconsidered!) dialogs whose impact last well beyond the meeting week.

A half-day excursion including a guided tour of “Burg Eltz” and a wine-tasting and dinner at the lovely down of Beilstein on the Mosel river completed the program.