CTWatch
November 2007
Software Enabling Technologies for Petascale Science
John Mellor-Crummey, Rice University
Peter Beckman, Argonne National Laboratory
Keith Cooper, Rice University
Jack Dongarra, University of Tennessee, Knoxville
William Gropp, Argonne National Laboratory
Ewing Lusk, Argonne National Laboratory
Barton Miller, University of Wisconsin, Madison
Katherine Yelick, University of California, Berkeley

2
1.1 Community Outreach and Vision-Building

Achieving petascale performance with applications will require a close collaboration between scientists developing computational models and computer science teams developing enabling technologies. To engage the community in the challenges and to foster interdisciplinary collaborations, we have established the CScADS Summer Workshops – an annual series of workshops that will focus on topics related to scalable software for the DOE’s leadership-class systems. In July 2007, we held our first series of four workshops in Snowbird, Utah.

  • Automatic Tuning for Petascale Systems. This workshop brought together researchers to discuss some of the code generation challenges for multicore processors that are the building blocks for emerging petascale systems, to identify some of the opportunities afforded by the use of automatic tuning techniques, and to explore opportunities for collaboration among tuning researchers.
  • Performance Tools for Petascale Computing. This workshop brought together researchers to discuss the challenges of measurement, attribution, analysis, and presentation of application performance for leadership computing platforms. An important workshop goal was to explore opportunities for research teams to break their software into components to foster community collaboration and accelerate development of effective performance tools for petascale systems.
  • Petascale Architectures and Performance. This workshop brought together computer scientists and SciDAC application scientists who aim to develop codes for the leadership computing platforms. The goal of this workshop was to introduce the computer scientists to the SciDAC applications and to familiarize the application scientists with the emerging leadership computing platforms, software libraries and tools for parallel computing, and effective strategies for parallel programming.
  • Libraries and Algorithms for Petascale Applications. This workshop brought together computer scientists working on algorithms and libraries with members of the SciDAC application teams. The principal workshop goal was to identify challenges for library and algorithm developers from the needs of the SciDAC applications and to foster collaboration between the communities. Workshop topics included the use of multicore processors and the use of automatic tuning in libraries.

The latter two workshops included “hands-on” sessions in which computer scientists and application scientists collaboratively explored application challenges on leadership computing platforms.

1.2 Research and Development

Several national reports, such as the 2000 report Information Technology Research: Investing in our Future by the President’s Information Technology Advisory Committee, have pointed out that open-source software represents an opportunity to address the shortage of software support for programming high-end systems. The power of this approach has been amply demonstrated by the success of Linux in fostering the development of operating systems for high-performance clusters.

The CScADS research program focuses on strategies for improving the productivity of application developers for developing high-performance codes for leadership-class machines. Rather than attack a narrow range of problems within this space, CScADS will explore a broad spectrum of issues because we believe that there is a high degree of synergy to be exploited.

Research on software support for high-end systems cannot be performed in a vacuum. Direct interaction between application developers and enabling technologies teams can clarify the problems that need to be addressed, yield insight into strategies for overcoming performance bottlenecks, identify how those strategies might be automated, and produce a vision for new tools and programming systems.

1.3 Open-Source Software Infrastructure

To facilitate the research, both within CScADS and in the community at large, we are developing the CScADS Open Software Suite. This suite will include an open-source software infrastructure to support compiler/programming-language research, development, and evaluation based on the Open64 compiler as well as Rice’s D System compiler infrastructure. Other components will include software infrastructure for performance tools, including support for binary analysis, instrumentation, data collection, and measurement interpretation that will draw from Rice’s HPCToolkit2 and Wisconsin’s Paradyn3 and Dyninst tools, and a range of libraries that help harness the power of leadership-class platforms composed of multicore processors.

Pages: 1 2 3 4 5 6

Reference this article
Mellor-Crummey, J., Beckman, P., Cooper, K., Dongarra, J., Gropp, W., Lusk, E., Miller, B., Yelick, K. "Creating Software Tools and Libraries for Leadership Computing," CTWatch Quarterly, Volume 3, Number 4, November 2007. http://www.ctwatch.org/quarterly/articles/2007/11/creating-software-tools-and-libraries-for-leadership-computing/

Any opinions expressed on this site belong to their respective authors and are not necessarily shared by the sponsoring institutions or the National Science Foundation (NSF).

Any trademarks or trade names, registered or otherwise, that appear on this site are the property of their respective owners and, unless noted, do not represent endorsement by the editors, publishers, sponsoring institutions, the National Science Foundation, or any other member of the CTWatch team.

No guarantee is granted by CTWatch that information appearing in articles published by the Quarterly or appearing in the Blog is complete or accurate. Information on this site is not intended for commercial purposes.