SUPERCOMPUTERS NEXT COMPETITIVE CHALLENGE
TRAVERSE CITY, Mich., Aug. 5 /PRNewswire/ -- Chrysler (NYSE: C), Ford (NYSE: F) and General Motors (NYSE: GM) plan to cooperate on the development of high-performance computer systems to leapfrog technology currently in use by domestic and foreign automakers.
The Big Three's new Supercomputer Automotive Applications Partnership (SCAAP) is the 12th major research-and-development consortium formed under the auspices of the United States Council for Automotive Research (USCAR). SCAAP plans to tap both public and private resources to address supercomputing issues of interest to U.S. auto manufacturers, according to Don Walkowicz, USCAR's executive director.
A key challenge facing the new computer group will be the development of portable software programs that can be easily set up and run on high-performance parallel computers (HPCs) by automotive design engineers.
Increased use of parallel supercomputing will help shorten the time needed to bring new automotive designs to market, speed design of new and more efficient engines to meet environmental standards and boost use of advanced materials, such as carbon fiber, in automotive body parts.
Parallel supercomputers break apart complex mathematical problems and perform simultaneous calculations -- a much faster process compared to the one-step-at-at-time approach followed by traditional computer systems.
Mr. Walkowicz said the new consortium has asked five Department of Energy (DOE) laboratories and the University of Michigan to take part in the initial phases of the program.
"DOE laboratories have a great deal of experience in the setup and programming of HPCs," he said. "Their participation will continue collaborative efforts started in the 1970s."
The five DOE National Laboratories are: Argonne National Laboratory; Lawrence Livermore National Laboratory; Los Alamos National Laboratory; Oak Ridge National Laboratory, and Sandia National Laboratories.
A three-member executive committee will manage the new partnership. Dr. George G. Dodd, head of the Analytic Process Department at the General Motors NAO R&D Center, will chair the committee for the first year. Other committee members include Ron Bienkowski, Chrysler's executive engineer for technical computing, and Wayne Hamann, Ford's director of product and manufacturing systems.
"By teaming up with our national laboratories and other companies and universities like the University of Michigan, we hope to create a clear competitive advantage for the United States in supercomputers," Dr. Dodd said.
"The availability of user-friendly, parallel-supercomputer software for U.S. industry is critical to the success of the nation's High Performance Computing and Communications Program," Dr. Dodd added. "DOE laboratories have unique expertise in HPCs, as well as in the modeling of complex physical phenomenon."
The consortium's initial research will focus on the development of four automotive applications for new parallel supercomputers:
-- Fluid dynamics, including fuel flow and aerodynamics;
-- Structural mechanics, focusing on composites;
-- Computer grid-generation technology, and
-- Visualization of computer simulations.
"It can take months to develop a mathematical description of a vehicle and weeks to process it," said Mr. Hamann. "While this is much faster than building a prototype vehicle and testing it, we want to increase total speed to the point where a virtual software description of a vehicle is used to impact a design early in the development process."
Total speed includes problem formulation, entry, computing and display -- the total cycle time of analysis. This is a broader description of speed than usually applied to computing, where ratings often refer just to the machine itself.
Issues and auto-industry needs facing the new Big Three consortium include:
-- Code Design. Standards need to be set for user acceptability. This may require observation of design engineers and how they interact with HPCs.
-- Portability. Software is required to run on different computers without major modification. Today, HPC languages are often customized for specific problems and computers.
-- Off-the-Shelf Components. Standardized chip designs, favored over custom hardware, need to be selected.
-- Building-Block Programs. Underlying many computer analyses is a common mathematical language or grid-generation technology. The adaptation of this language to specific problems is often a time- consuming process because irregular surfaces require complex mathematical descriptions. SCAAP hopes to develop a common HPC language with a library of sub-routines to address the problem.
-- Three-Dimensional Calculations. Calculating the flow of air over a car's hood, around fenders or past spinning tires is much more difficult than calculating airflow over one point on an airplane because of an airplane's generally more simple, aerodynamic design. SCAAP will look for ways to develop 3-D routines that automotive design engineers can use without resorting to direct help from programmers.
"The consortium wants to make HPC a tool that design engineers can use every day without the need for elaborate support systems," Dr. Dodd said.
He added that several commercial software development-and- maintenance companies will be selected to participate in SCAAP-generated research-and-development work once the consortium has reached agreement on a detailed work plan.
CONTACT: For interviews or further information:
Media Contacts Interview Candidates
Mary Roznowski George Dodd
General Motors Corporation Department Head
313-986-5717 Analytic Process Department
General Motors Corporation
Chris Preuss Ron Bienkowski
Chrysler Corporation Executive Engineer
313-576-8095 Technical Computer Center
Vehicle Engineering Operations
Mark Miller Wayne Hamann
Ford Motor Company Director
313-845-5745 Product & Manufacturing Systems
Car Product Development
Ford Motor Company
Larry Weis Don Walkowicz
USCAR Executive Director
(C F GM) -- DD002 -- X324 08/05/93