Responding to the Software Crisis in DOE Scientific Computing

A Minisymposium, held at DOE/Germantown, on Monday September 14, 2015

Jump to the Call for Participation (below)

Agenda

Times Title and Speaker
8:30am 8:37am Welcome
Thomas Ndousse-Fetter, ASCR and David Lesmes, BER
8:38am 8:45am Introduction
David Bernholdt, ORNL
8:45am 9:10am ASC Software Engineering Lessons Learned
Greg Pope, LLNL
9:10am 9:35am High-Impact Modeling and Simulation: Realizing the Potential for the Applied DOE Offices and Industry
John Turner, ORNL
9:35am 10:00am Observations on the SciDAC Software Experience
Bob Lucas, USC
10:00am 10:25am A Facility’s View of User Software Engineering Practices
Richard Gerber, NERSC
10:25am 10:40am BREAK
10:40am 11:05am Preparing Scientific Software for Exascale
Jack Wells, OLCF
11:05am 11:30am Overview of the IDEAS Software Productivity Project
Lois Curfman McInnes, ANL
11:30am 12:30pm Panel: Addressing DOE’s Software Engineering Challenges
Panelists: Jim Davenport, BES; Thuc Hoang, ASC; Dorothy Koch, BER; Mark Pederson, BES;
Moderator: Mike Heroux, SNL
12:30pm ADJOURN
Informal discussions can continue at lunch, in cafeteria

Call for Participation

Background

A confluence of trends is creating challenges of crisis proportions for scientific software developers. The rapid growth in computational capability has fed the demand for increasingly sophisticated and powerful scientific software, which also tends to be larger and more complex. However, trends in computer architecture as we approach and exceed exascale may require re-examination of everything from programming models to algorithms and their implementations. Scientific software developers face increasing difficulties with this environment, which impacts their productivity in developing software and obtaining scientific results with it, and even the quality and credibility of those results. Improving HPC application developer productivity is one of the five themes of the National Strategic Computing Initiative.

Purpose

This half-day mini-symposium will examine the current state of the engineering of scientific software in DOE, coming challenges, and current and possible responses. Audience participation will be strongly encouraged.

Who Should Attend

The mini-symposium is open to all DOE program managers and other interested parties. Program managers with computationally- and software-intensive programs are particularly encouraged to attend.

Topics

DOE researchers and program managers will present their experiences with scientific software development across the DOE complex, how current trends are exacerbating the challenges, and how they are responding. Specific topics include:

  • The increasing size and complexity of scientific software across DOE;
  • Additional challenges posed by architectural trends for exascale computers;
  • Experience and lessons learned regarding software engineering and software quality in the ASC (NNSA) program, CASL (NE), CAEBAT (EERE), and SciDAC (SC) programs;
  • The impact of application teams’ software engineering capabilities on their ability to get the most out of DOE computing facilities, and on the facilities’ ability to support them;
  • How the computing facilities are helping application teams prepare for pre-exascale and exascale class systems;
  • How current projects and programs are addressing software engineering issues; and
  • Ideas for additional ways to respond to the software crisis.

Organizers/Contacts

of the Interoperable Design of Extreme-scale Application Software (IDEAS) project.