Technologies for the 21st Century HECC R&D Examples

	ACTS PACI ASCI's Academic Strategic Alliance centers High Performance C++ The M-Machine National Compiler Infrastructure project Parallelization of commercial engineering software Parallel system software tools Computational Aerosciences Project Earth and Space Sciences project Biomedical research Biomolecular computing Supercomputing Research Program Scalable Systems and Software Quorum
ACTS	The Department of Energy's (DOE) Grand Challenge and Accelerated Strategic Computing Initiative (ASCI) applications are the focus of the agency's DOE 2000 Program. A primary thrust is the Advanced Computational Testing and Simulation (ACTS) Toolkit, which will provide an integrated set of software tools, algorithms, and environments to accelerate the adoption and use of advanced computing by DOE programs for mission-critical problems. An example is the Scientific Template Library project, which involves numerous DOE laboratory and university developers who are creating software toolkits written in a data-structure-neutral manner to enable easy reuse and flexibility. DOE supports user access to new high performance computing architectures at the National Energy Research Scientific Computing center (NERSC) and four High Performance Computing Resource Providers at DOE's Argonne, Lawrence Berkeley, Los Alamos, and Oak Ridge National Laboratories.   The National Science Foundation (NSF) Supercomputer centers program has been instrumental in advancing science and engineering research and in enabling the U.S. to lead the world in computational science and engineering. The program is now evolving into the new Partnerships for Advanced Computational Infrastructure (PACI) program.
PACI	The PACI program is based on partnerships that will operate two leading-edge sites equipped with high-end computing systems one to two orders of magnitude more capable than those typically available at a major research university. Systems accessible by the Nation's academic research community will scale to teraflop computing capability within the next two to three years. Partners not directly involved in high end R&D will contribute to education, outreach, and training, and will develop software that will facilitate and enhance both the overall infrastructure and access to that infrastructure. After an intense competition and extensive review cycle, two sites were selected: The National Partnership for Advanced Computational Infrastructure (NPACI), led by the University of California at San Diego (UCSD) The National Computational Science Alliance (NCSA) led by the University of Illinois at Urbana-Champaign (UIUC). At NPACI, high performance compute servers will be located at four major partner sites: University of California-Berkeley, Caltech, University of Michigan, and University of Texas. Several partners will focus on data intensive computing, as computational science and engineering evolve to enable the analysis of massive data sets from remote-sensing sources, numerical simulation output, and discipline-specific databases.   Additional NCSA computational resources will be provided by major partners Boston University and DOE's Argonne National Laboratory. The alliance will provide visual supercomputing capabilities in collaboration with several partners and with one of its strategic vendors, Silicon Graphics, Inc. The current supercomputing center at UIUC has pioneered the use of visualization for comprehending the mass of data produced by simulations or experiments and will continue this thrust in the new alliance.
ASCI's Academic Strategic Alliance centers	DOE's ASCI program focuses on creating leading-edge computational modeling and simulation capabilities to maintain the safety, reliability, and performance of the U.S. nuclear stockpile. The goal of the ASCI Academic Strategic Alliance centers is to validate the scientific methodologies of large scale modeling, simulation, and computation. They will form centers of Excellence to foster research in scientific, computational, and computer science areas that support ASCI and Science Based Stockpile Stewardship (SBSS) objectives. The centers will have collaborative relationships with ASCI laboratories and will facilitate technical partnerships with ASCI researchers.
High Performance C++	The Defense Advanced Research Projects Agency-supported (DARPA) High Performance C++ project is defining a standard model for parallel programming using the C++ programming language. The model will support both data parallel and distributed object task parallelism, enable the construction of portable parallel applications, and provide a target for vendors to support with compiler optimization. The end product will be used on multiple platforms such as symmetric, shared memory multiprocessors as well as massively parallel systems with complex memory hierarchy. DARPA's goal is the seamless integration of science, engineering, and commercial applications that require the resources of new generation supercomputers with software that runs on conventional hardware platforms.
The M-Machine	The M-Machine is a fine grained concurrent computer being designed and built by the Massachusetts Institute of Technology (MIT) Concurrent Very Large Scale Integration (VLSI) Architecture group under the DARPA "Mechanisms for Teraflops" project. The aim of this project is to exploit increased circuit density more efficiently by building multi-ALU (arithmetic logic unit) processing nodes. While building arithmetic units, register files, and memories and replicating them on an integrated circuit is straightforward, new architecture technology is required to organize and control many arithmetic units on a single chip. The M-Machine is designed to execute programs efficiently with any or all granularities of parallelism by means of parallel instruction sequences called H (horizontal)-Threads, as well as by exploiting thread-level parallelism, and masking variable pipeline, memory, and communication delays through V (vertical)-threads. The M-Machine will employ silicon fabrication technology that will deliver 1200 MIPS (millions of instructions per second) and 800 Mflops (millions of floating point operations per second) per node.
National Compiler Infrastructure project	The goal of the National Compiler Infrastructure project is to develop a common platform for compiler development that will support open academic collaboration on compilers and facilitate the transfer of technology to industry. The project combines technology developed under the Stanford University Intermediate Format (SUIF) project and the Zephyr project located at the University of Virginia. The SUIF project, co-funded by DARPA and NSF and based on the Stanford University Intermediate Format parallelizing compiler developed at Stanford University, involves collaboration with researchers at Harvard University, Rice University, and the University of California at Santa Barbara. The infrastructure project is intended from the outset to involve one or more industrial partners, and its objective is to develop a system that is modular, easy to extend and maintain, supports current research, and supports software reuse.   The Zephyr project, supported by DARPA and NSF, is a collaboration between Princeton University and University of Virginia. Zephyr's organizational structure separates the form of its intermediate languages from their content. Zephyr intermediate representations have tree-like forms describable by abstract syntax specifications. The language- or machine-specific content of a Zephyr intermediate representation is described by other compact specifications. Decoupling form from content permits researchers to choose the intermediate language most appropriate for their needs, to extend an existing intermediate language, or to develop a new high level intermediate language, while still using the Zephyr infrastructure.
Parallelization of commercial engineering software	The DARPA-supported Parallelization of Commercial Engineering Software project is a cooperative program between the Federal government and industry. Its objective is to spur the introduction of scalable parallel applications software to U.S. industrial markets by directly providing leading-edge independent software vendors (ISVs) with small parallel development platforms in order to minimize their investment risk. The immediate goal of the program is to produce first-generation parallel implementations of existing commercial engineering software using system software and enablers -- software tools that allow programmers to implement applications effectively -- that are currently emerging as industry standards.   Each of the ISVs involved in the program will gain access to scalable parallel computing hardware, state-of-the-art systems software and enablers, and techniques that will ensure compatibility with MPI (message passing interface) software, a standard programming interface. Although the proprietary nature of the individual application development projects will be protected, IBM's common platform support will ensure that the resulting parallel applications are portable and consistent with industry standards.
Parallel system software tools	The National Aeronautics and Space Administration's (NASA) HECC R&D centers on development of parallel system software tools as well as development of mission-driven applications. NASA's high performance computing research facilities were established to accelerate the transition to new generations of high performance computing technology and include access to the NASA Research and Education Network (NREN), early systems or prototype storage subsystems, and state-of-the-art visualization applications. NASA plans to install a 100-250 gigaflops sustained, scalable teraflops testbed in FY 1998 to support its mission-critical efforts.
Computational Aerosciences Project	NASA's Computational Aerosciences Project (CAS) focuses on using the high performance computing environment to solve a range of problems in aerospace engineering at a cost that reflects the value, flexibility, and short cycle time required by the aerospace community. Aerospace design problems include high speed technology, advanced subsonics technology, and rotorcraft. Research products are made available to system vendors as quickly as possible.   CAS researchers work with computational fluid dynamics simulations of aircraft and energy-efficient jet engines in order to save fuel and reduce design time. NASA has shown that using parallel computing-based simulations can reduce overall design and development time for high pressure compressors in turbofan engines from 18 months to 14 months, resulting in several millions of dollars in savings.   The aim of NASA's Flow Analysis of a Gas Turbine Low Pressure Subsystem project is to simulate the aerodynamic flow in the complete low pressure subsystem (LPS) of a gas turbine engine. The detailed LPS model will be integrated into the Numerical Propulsion System Simulator (NPSS) that will serve as a numerical "test cell." The goal of the LPS project and NPSS is to provide a tool that can significantly reduce risk, uncertainty, and the cost associated with designing advanced gas turbine engines.
Earth and Space Science project	Earth and Space Sciences (ESS) research covers two areas: coupling advanced discipline models into scalable global simulations to better understand global change; integrating models and analysis algorithms for processing, analyzing, and understanding the enormous volumes of data expected from scientific missions. ESS researchers are simulating the earth's climate and are using large parallel computers to model dynamics within the sun, the solar wind, and the formation of galaxies.
Biomedical research	The mission of the National center for Research Resources (NCRR) is to serve as a "catalyst for discovery" for National Institutes of Health-supported (NIH) research by developing resources to enable biomedical research advances. NCRR promotes collaborations within and across scientific disciplines and provides quick, flexible approaches to new and emerging research needs. Its greatest challenge is to keep abreast of rapidly evolving trends in basic and clinical research.
Biomolecular computing	HECC research in biomolecular computing with high performance computers involves extensive and often complex calculations to determine or predict protein structures and their structural and functional changes due to interactions with other molecules or drugs. NIH in partnership with NSF supports research that demonstrates the utility of novel protein functions for the biotechnology industry in areas such as synthesizing models of protein receptors for structure-based drug design.
Supercomputing Research Program	The National Security Agency's (NSA) Supercomputing Research Program focuses on research and applications that will gain order of magnitude improvement in computer support for intelligence processing. NSA's research activities range from the invention and prototyping of new concepts to improvements in the use of leading-edge commercial projects, with a current focus on parallelism. Current projects include support for flight testing MARQUISE (an embedded high performance computer) on Air Force aircraft; in partnership with DARPA, the 3-D Diamond MCM Cube computer project that will build a test vehicle to demonstrate a 3-D computer architecture with a nanosecond system clock; and funding of several fundamental studies to explore the feasibility of quantum computing.
Scalable Systems and Software	DARPA's Scalable Systems and Software program is developing technologies for secure, scalable, distributed computing. The program focuses on creating the architecture and software components needed to reduce risks and accelerate commercial development of new generations of the scalable, high-performance computing systems required by DOD, with a goal of enabling procurement of very high-performance systems that run unique applications without requiring continual hardware, operating system, and applications redevelopment. Advanced technology for teraflop-class systems is an emerging development from DARPA investments in this area.
Quorum	DARPA's Quorum research -- part of its Information Sciences project -- focuses on developing technologies that will allow end users to achieve predictable and controllable end-to-end quality of service for critical DOD computing needs in a global heterogeneous distributed computing environment. These technologies and architectures will be integrated into a global operating system in such a way that this resource pool appears to end users as a single computing platform. They will be demonstrated on key defense problems in shipboard computing and command and control, focusing on human-computer interaction and design methods and technology enabling more natural interaction between people and computers.