APPLICATIONS

Over the past 40 years, gains in supercomputing performance have tracked progress in integrated circuit scaling, resulting in exponential improvements in system-level performance. However, changes in device physics now seriously threaten further sustained progress toward exaflops HPC systems. At the upcoming SC08 Conference in Austin, a one-day workshop on Sunday, Nov. 16, will look at the challenges to achieving exascale performance. Entitled “Energy Efficiency and the Path to Exascale Computing,” the workshop will feature invited talks by many of the leading experts in computer architecture, programming and software, and the design of data centers. The workshop will be held in Room 16A/B of the Austin Convention Center. The workshop is open to all SC08 attendees with Technical Program registration, as well as to attendees opting for a workshops-only registration. Details about registering for SC08, to be held Nov. 15-21, can be found at its Web site. The workshop, organized by John Shalf of Lawrence Berkeley National Laboratory, Stephen Elbert of Pacific Northwest National Laboratory (PNNL) and Thomas Sterling of Louisiana State University’s Center for Computation and Technology, will begin with a summary of a recent DARPA study highlighting issues associated with exaflops computing architecture. This will be followed by a series of invited talks covering a wide range of approaches to mitigating the exaflops computing roadblocks, including novel computer architectures, energy-aware algorithm design, energy efficiency metrics, and energy efficient facility design. Presentation highlights include: * Thomas Sterling, Erik Debenedictis of Sandia National Laboratories and Dean Klein, vice president of memory system development at Micron Technology, will talk about the ramifications of silicon technology trends on computer and memory architecture. * Alan Gara, chief architect of IBM’s BlueGene, and Steve Scott, chief technology officer of Cray Inc. will talk about exascale computer architecture. * Kathy Yelick, director of DOE’s National Energy Research Scientific Computing Center, Padma Raghavan, director of the Institute of CyberScience at Penn State, and Vivek Sarkar, professor of computer science at Rice University and lead author of DARPA Exscale Software Study, will describe programming models, algorithms, and the entire software stack needed for exascale computing. * Finally, Dan Reed, director of Datacenter Design and Multicore Strategy at Microsoft, and Andres Marqez, technical manager of the Energy Smart Data Center project at PNNL, will talk about the design of datacenters to accommodate future exascale computing platforms. More information about the workshop can be found at: its Web site.