National Aeronautics and Space Administration Ames Research Center HPCC Program Office Moffett Field, Calif. 94035

For Release:
November 15, 1996

Jarrett Cohen Goddard Space Flight Center, Greenbelt, Md. (301/286-2744, jarrett.cohen@gsfc.nasa.gov)

During Supercomputing '96: Pittsburgh Hilton & Towers (412/391-4600)

NASA HIGHLIGHTS COMPUTATIONAL ADVANCES AT SUPERCOMPUTING '96

PITTSBURGH -- The diverse ways NASA advances computational science and information technologies are being featured at Supercomputing '96, November 17­22.

Through research exhibits and a range of technical program activities, NASA's High Performance Computing and Communications (HPCC) Program will highlight its contributions to the aerosciences, Earth and space sciences, space-borne computing, information infrastructure -- and ultimately national competitiveness.

"From hardware, system software and networks to the many and varied Grand Challenge applications that push their capabilities, our staff and funded investigator teams are expanding the frontiers of high-performance computing," said William Feiereisen, NASA HPCC program manager. "The benefits reach across academia, NASA and other government agencies, industry, the K­12 education community, as well as the general public."

Research and Industry Exhibits
NASA HPCC's first-ever research exhibit (#R4) will use a video theater, the World Wide Web (wwW) and colorful imagery to demonstrate leading-edge R&D by its four projects:

• Computational Aerosciences (CAS), which focuses on accelerating the availability of high-performance computing hardware and software to U.S. aerospace manufacturers for use in their design processes.
• Earth and Space Sciences (ESS), which employs scalable parallel computers to further our understanding and ability to predict the dynamically interacting physical, chemical and biological processes that drive the Earth, solar system and universe.
• Remote Exploration and Experimentation (REE), which seeks to develop, in conjunction with industry, low-power, scalable, fault-tolerant, high-performance computing for use in space, enabling greater science return at lower cost.
• Information Infrastructure Technology and Applications (IITA), which strives to hasten implementation of a National Information Infrastructure through new technologies facilitating broader access to NASA data via computer networks.

The exhibit also will include an illustration of the NASA MetaCenter project exploring automatic execution of jobs on remote systems using IBM SP2 systems at NASA Ames Research Center and NASA Langley Research Center.

NASA HPCC's research exhibit will be open Monday, November 18 from 6 to 7 p.m. for the press tour and from 7 to 9 p.m. for the Open House Reception; Tuesday, November 19 from 10 a.m. to 5:30 p.m.; Wednesday, November 20 from 10 a.m. to 5:30 p.m.; and Thursday, November 21 from 10 a.m. to 4 p.m.

Additional technology demonstrations will take place in other research exhibits, including that of Ames' Numerical Aerospace Simulation (NAS) Systems Division (#R40). CAS parallel computing tools on display include p2d2, a portable parallel/distributed debugger, and PMPIO, which enables high-speed disk I/O on applications using MPI (Message Passing Interface). The MetaCenter application will be running here as well.

Features of the California Institute of Technology (#R62) and Los Alamos National Laboratory (#R86) booths are two Beowulf parallel workstations, one built with the NASA Jet Propulsion Laboratory (JPL), that exceed sustained gigaFLOPS performance for a cost of approximately $50,000 (Editor's Note: see details in a separate news release). Other JPL demonstrations in the Caltech booth include a new supercomputer volumetric renderer called ParVox and a space flight mission design tool based on the TRIP parallel trajectory integration package.

NASA Lewis Research Center's research exhibit (#R23) will show HDTV visualizations of a variety of aeroscience fluid dynamics codes, including an axisymmetric full engine analysis package. Virtual reality will simulate use of the Advanced Communications Technology Satellite through real-time display of geographically distributed data. The LSF (Load Sharing Facility) interactive load monitoring system for Lewis' cluster of 60 IBM RS/6000 workstations also will be demonstrated.

NASA visualization imagery also will be on display in the National Coordination Office for HPCC booth (#R54).

The Silicon Graphics, Inc./Cray Research industry exhibit (#707) will be the setting for a JPL talk on using HPCC to process and visualize satellite-collected Synthetic Aperture Radar (SAR) data. An ESS Grand Challenge investigation team, led by JPL's David Curkendall, will demonstrate real-time SAR processing on the new Origin2000 system.

Technical Program and Awards Sessions
NASA HPCC staff and investigators are delivering presentations in a variety of sessions. The schedule is as follows:

Monday, November 18

8:30 a.m. to 5 p.m.

Tutorial: M4 "Introduction to Effective Parallel Computing," Quentin Stout, University of Michigan; Marilynn Livingston, Southern Illinois University

This class emphasizes the conversion of a serial program into an increasingly efficient parallel program. "Some of the tutorial will make reference to lessons learned in working on material that led to our successful NASA proposal" for studying the solar wind, ESS co-investigator Stout said.

Tutorial: M5 "Parallel I/O on Highly Parallel Systems," Samuel Fineberg, Bill Nitzberg, Ames

The instructors will survey programming techniques for obtaining maximum I/O performance, ranging from basic concepts to recent advances in the research community. Examples on commercial supercomputers and workstation clusters as well as academic and research projects will be used to illustrate varying requirements, interfaces and performance.

Tutorial: M6 "Hot Chips for High Performance Computing," Subhash Saini, David Bailey, Ames

Discussion will center on current microprocessors, the supercomputers based on them and the performance of supported hardware/programming models (including combinations such as HPF and MPI). NAS Parallel Benchmark results will provide a cross-machine and cross-model comparison.

Tuesday, November 19

1:30 to 3 p.m. - Technical Papers: 2A Visualization & Education

"STREN: A Highly Scalable Parallel Stereo Terrain Renderer for Planetary Mission Simulations," Ansel Teng, Meemong Lee, JPL; Scott Whitman, Cray Research, Inc.

STREN was developed to simulate the imaging system of the NASA Mars Pathfinder, scheduled for launch in December. With Pathfinder's two cameras only 1.5 meters above the surface, rendering a high-resolution scene will be challenging. STREN allows real-time rendering of stereo images by organizing terrain data in a novel way and using parallel processing to speed computation.

3:30 to 5 p.m. - Technical Papers: 3A Geophysical Applications

"Performance Analysis and Optimization on the UCLA Parallel Atmospheric General Circulation Model Code," John Farrara, UCLA; John Lou, JPL

The code is part of the UCLA Earth System Model partially supported by ESS. Other components under development include parallel oceanic general circulation, atmospheric chemistry and oceanic chemistry models.

"Climate Data Assimilation on a Massively Parallel Supercomputer," Hong Ding and Robert Ferraro, JPL

They describe a parallel version of the Physical-space Statistical Analysis System (PSAS), the core of NASA's atmospheric data assimilation system. A PSAS routine encompassing 80,000 observations achieves 12.6 gigaFLOPS on a 512-processor CRAY T3D and runs 230 times faster than on a CRAY C90 processor.

Wednesday, November 20

3:30 to 5 p.m. - Technical Papers: 6A Algorithms II

"Global Load Balancing with Parallel Mesh Adaption on Distributed-Memory Systems," Rupak Biswas, Ames; Leonid Oliker, Research Institute for Advanced Computer Science; Andrew Sohn, New Jersey Institute of Technology

The authors' tetrahedral mesh adaption scheme coupled with a new global load balancing method resolves simulation features on an unstructured grid and minimizes redistribution cost among the processors. "For large-scale scientific computations, our load balancing strategy gives almost sixfold reduction in solver execution times over non-balanced loads," they state.

Thursday, November 21

10 a.m. to 12 p.m. and 1:30 to 3 p.m. - 1996 Sidney Fernbach Award

ESS co-investigator Gary Glatzmaier, Los Alamos, is this year's recipient for his use of "innovative computational numerical methods to perform the first realistic computer simulation of the Earth's geodynamo and its resultant time-dependent magnetic field." Glatzmaier will deliver a lecture beginning at 10 a.m. and receive the award at 1:30 p.m. Since 1992 the award has honored Fernbach as a pioneer in the development and application of high-performance computers for the solution of large problems.

10 a.m. to 12 p.m. - Panel: "Opportunities and Barriers to PetaFLOPS Computing"

Paul Messina, Caltech and JPL, will moderate an exploration of alternative architecture paths for building supercomputers capable of one thousand trillion operations per second. Machines employing future-generation semiconductors, processor-in-memory chips and superconducting logic are among the options for reaching such performance in 20 years. Thomas Sterling, also of Caltech and JPL, is a panelist.

3:30 to 5:00 p.m. - Roundtable: "Scalable I/O Initiative on Portable Programming Interfaces for Parallel Filesystems"

Discussion will center on a filesystem suite "to define a common set of powerful interfaces for application, compiler and toolkit programmers to use for high-performance I/O on a variety of parallel machine architectures." Messina serves as moderator.

For More Information
Additional information about the NASA HPCC Program is available on the wwW at the following URL:

http://www.aero.hq.nasa.gov/hpcc/hpcc.nasa.htm

- end -