Centers for Enabling Technologies | Institutes | Applications and Scientific Application Partnerships

Overview of the SciDAC-2 Program

The goal of SciDAC-2 is to advance the state-of-the-art in scientific simulation by creating multi-disciplinary teams comprised of mathematicians, computer scientists and application science domain experts. The program was initiated in 2001 and has already resulted in significant new advances in the areas of astrophysics, particle accelerator design, climate, combustion, fusion, and quantum chromodynamics (QCD). The SciDAC-2 program will continue to focus on creating the next generation of scientific computing software infrastructure and has been expanded to include development of new data management and knowledge discovery tools for large data sets (from both experiments and simulation).

The overall software infrastructure vision is to create a comprehensive, portable, and fully integrated suite of tools that SciDAC application teams can use to effectively manage and utilize petascale computational resources. For example, a critical component of the future success of SciDAC is the development of new high-performance scalable algorithms for core numerical components of scientific simulation and the dissemination of those algorithms via portable high-performance libraries. To realize its vision, the SciDAC program has three tightly-coupled components:

  • Centers for Enabling Technologies (CET): large teams of researchers from multiple institutions centered around developing mathematical and computer science software with a single focus,
  • SciDAC Institutes: university-led centers of excellence that focus on long term research and collaborative software development in a given focus area, and
  • Scientific Applications: teams of application scientists, mathematicians, and computer scientists that are focused on using simulation to achieve breakthrough scientific advances in a variety of DOE mission areas that are impossible using theoretical or laboratory studies alone. Many of the mathematicians and computer scientists funded to perform research on scientific applications are funded through the associated Scientific Application Partnerships (SAP) program.

More information on the SciDAC program and the winning proposals can be found at http://www.scidac.gov.

LLNL's Participation in SciDAC

Lawrence Livermore National Laboratory scientists are participating in 7 Centers for Enabling Technologies, 1 Institute and 6 Scientific Application Partnerships. CET project areas include visualization (VACET), scalable data management (SDM), interoperable meshing/geometry tools (ITAPS), scalable linear and nonlinear solvers (TOPS), structured AMR algorithms (APDEC), component technologies (TASCS), and the earth system grid (ESG). Two of these CETs (ITAPS and ESG) are led by LLNL researchers and collectively they involve a broad cross section of Computation directorate personnel including researchers from the CASC, BACE, CADSE, and DEPCom divisions. The institute project is focused on performance engineering research and involves Computation personnel in CASC. The biggest area of expansion for LLNL personnel was in the area of applications and scientific application partnerships. Topic areas for the winning application projects include quantum simulations, astrophysics, climate, fusion, stress corrosion, and turbulence and involve personnel from Computation, DNT, E&E, and PAT.

Centers for Enabling Technologies

TASCS: Center for Technology for Advanced Scientific Component

PI: David E. Bernholdt (ORNL)

LLNL PI: Gary Kumfert

Collaborating Institutions: Argonne National Laboratory, Binghamton University, Indiana University, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Oak Ridge National Laboratory, Pacific Northwest National Laboratory, Sandia National Laboratories, Tech-X Corporation , University of Maryland, University of Utah

Description: This project will develop and extend high performance computing software component methodology through activities that focus on coupling parallel simulations, support emerging hardware and software paradigms for petascale computing, enhance software quality and robustness, and dynamically adapt applications.

LLNL’s role: LLNL will address the following four areas of component technology research, building upon technology developed by the CASC Components project. First, we will extend our Babel language interoperability technology to support Fortran 90 as needed by SciDAC application collaborators. Second, we will develop component schema and communication protocols between our component repository and component tools developed by our collaborators. Third, we will continue to investigate parallel data redistribution approaches and integrate that support into our Babel language interoperability framework. Finally, we will work with SciDAC application groups to use component technology in large, sophisticated simulation codes. This research effort will help us understand both the benefits and limitations of software component technology in a scientific computing environment.

The Scientific Data Management Center for Enabling Technologies

PI:Arie Shoshani , Lawrence Berkeley National Laboratory

LLNL PIs:Terence Critchlow

Collaborating Institutions: Argonne National Laboratory, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, Pacific Northwest National Laboratory, North Carolina State University, Northwestern University, University of California at Davis, University of California at San Diego, University of Utah

Description: This center will enhance and extend existing scientific data management tools to allow for more interactivity and fault tolerance when managing scientists’ workflows, for better parallelism and feature extraction capabilities in their data analytics operations, and for greater efficiency and functionality in users’ interactions with local parallel file systems and remote storage.

LLNL’s role: LLNL is participating in all PERC focus areas, but is primarily contributing to the modeling and tools efforts. PERC is funding portions of the development of ROSE, a tool for building customizable preprocessors. Generally ROSE-generated preprocessors optimize application use of libraries although PERC is also producing performance analysis tools based on ROSE. LLNL is developing tools based on data-dependent memory-tracing models that will capture the keys to understanding the architecture-independent aspects of an application's memory performance. LLNL is also developing performance assertions, a mechanism that allows a user to identify a performance expectation for a given computation or communication operation and to have some action, such as gathering of detailed performance data, taken when the expectation is not met at run-time.

LLNL has already made significant contributions to PERC. We have designed and developed a tree traversal mechanism in ROSE that is based on attributed grammars in order to support complex transformations dependent upon the application context. We have developed tools that dynamically measure the regularity of an application's memory references; our results demonstrate that these statistics provide simple, intuitive optimization guidance. We have used a prototype of performance assertions to perform initial experiments that demonstrate functionality such as 'raising a performance exception' and 'local performance-based adaptation.'

ESG-CET: Earth System Grid Center for Enabling Technologies

PI: Dean Williams, Lawrence Livermore National Laboratory

Collaborating Institutions: Argonne National Laboratory, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, National Center for Atmospheric Research, Oak Ridge National Laboratory, Pacific Marine Environmental Laboratory, University of Southern California's Information Sciences Institute.

Description: The goals of this proposed five-year project are to (a) sustain the successful existing ESG system, (b) address projected scientific needs for data management and analysis, (c) extend ESG to support the major Intergovernmental Panel on Climate Change assessment in 2010, (d) support the Climate Science Computational End Station at the DOE Leadership Computing Facility at ORNL, and (e) support climate model evaluation activities under the proposed SciDAC2 climate application. To do this, we will broaden ESG to support multiple types of model and observational data, provide more powerful (client-side) ESG access and analysis services, enhance interoperability between common climate analysis tools and ESG, and enable end-to-end simulation and analysis workflow.

LLNL’s role: LLNL’s tasks include: project management, outreach, developments, data management, operations, system architecture, analysis and visualization framework, portal, and liaison science support.  

The goals of this proposed five-year project are to (a) sustain the successful existing ESG system, (b) address projected scientific needs for data management and analysis, (c) extend ESG to support the major Intergovernmental Panel on Climate Change assessment in 2010, (d) support the Climate Science Computational End Station at the DOE Leadership Computing Facility at ORNL, and (e) support climate model evaluation activities under the proposed SciDAC2 climate application. To do this, we will broaden ESG to support multiple types of model and observational data, provide more powerful (client-side) ESG access and analysis services, enhance interoperability between common climate analysis tools and ESG, and enable end-to-end simulation and analysis workflow. For more information contact: Dean N. Williams, Don Middleton, Ian Foster - esg-manage@earthsystemgrid.org.

APDEC: The Applied Partial Differential Equations Center for Enabling Technologies

PI: Philip Colella, Lawrence Berkeley National Laboratory

LLNL PI: David Trebotich

Collaborating Institutions: Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Princeton Plasma Physics Laboratory , University of California at Davis

Description:This project will develop simulation tools for solving multi-scale and multi-physics problems based on finite difference and finite-volume methods on logically-rectangular structured grids combined with block structured adaptive mesh refinement to represent multi-scale behavior.

LLNL’s role:

ITAPS: Center for Interoperable Technologies for Advanced Petascale Simulations

PI: Lori Diachin, Lawrence Livermore National Laboratory

Collaborating Institutions: Brookhaven National Laboratory, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, Pacific Northwest National Laboratory, Renssealer Polytechnic Institute, Sandia National Laboratories, State University of New York at Stony Brook, University of British Columbia

Description: This project will deliver interoperable and interchangeable mesh, geometry, and field manipulation services that are of direct use to SciDAC applications with minimal intrusion into application codes.

LLNL’s role:

TOPS:Towards Optimal Petascale Simulations

PI: David Keyes, Columbia University

LLNL PI: Rob Falgout

Collaborating Institutions: Argonne National Laboratory, Columbia University, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Sandia National Laboratories, University of California at Berkeley, University of California at San Diego, University of Colorado at Boulder, University of Texas at Austin

Description: The primary goals of TOPS are the development, testing, and dissemination of linear and nonlinear solver software, especially for systems governed by partial differential equations with attention to high performance as well as interoperability among solver components.

LLNL’s role:

VACET: Visualization and Analytics Center for Enabling Technologies

PI: Wes Bethel, Lawrence Berkeley National Laboratory

LLNL PI: Rob Falgout

Collaborating Institutions: Argonne National Laboratory, Columbia University, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Sandia National Laboratories, University of California at Berkeley, University of California at San Diego, University of Colorado at Boulder, University of Texas at Austin.

Description: The primary goals of TOPS are the development, testing, and dissemination of linear and nonlinear solver software, especially for systems governed by partial differential equations with attention to high performance as well as interoperability among solver components.

LLNL’s role:

Institutes

Performance Engineering Research Institutes

PI: Robert Lucas, Information Science Insitute, University of Southern California

LLNL PI: Bronis de Supinski, Daniel Quinlan

Collaborating Institutions: Argonne National Laboratory, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, Rice University, San Diego Supercomputing Center, University of Maryland, University of North Carolina, University of Southern California, University of Tennessee - Knoxville.

Description: As we look to the future, achieving good performance on high-end computing (HEC) systems is growing ever more challenging due to enormous scale, increasing architectural complexity, and increasing application complexity. To address these challenges, the SciDAC-3 the Performance Engineering Research Institute for Enabling Technology (PERI) is pursuing a unified, tripartite research plan encompassing:

      Performance modeling and prediction
      Automatic performance optimization
      Pcerformance engineering of high profile applications

The PERI performance modeling and prediction activity will develop and refine our performance models, significantly reducing the cost of collecting the data upon which the models are based and increasing model fidelity, speed and spurred by the strong user preference for automatic tools. This work is building on previous successful activities such as ATLAS, which has automatically tuned components of the LAPACK linear algebra library, the highly successful FFTW library, and other recent work. In our third major component, application engagement, we directly interact with SciDAC applications, including "tiger teams" that focus on particular codes. A summary of PERI on the DOE SciDAC web pages is available from http://www.scidac.gov/compsci/PERI.html

LLNL’s role: LLNL is an active participant in all three aspects of the tripartite PERI research plan. Within the automatic tuning activity, LLNL is leading the development of whole program analysis and optimization work which will introduce optimizations spanning multiple kernels and across files and procedure boundaries. This work, within the ROSE framework, will initiate the using empirical techniques on significantly larger scales than individual kernels. Within the performance modeling activity, LLNL is exploring a performance modeling methodology based on artificial neural networks and piecewise polynomial regression that predicts application run-times accurately and with error bounds as the application's input parameters are varied. We expect this approach will facilitate searching large parameter spaces that arise in the automatic performance tuning activity. Finally, LLNL is leading the PERI tiger team component of the application engagement activity and is working directly with several application teams

Applications and Scientific Application Partnerships

A Scalable and Extensible Earth System Model for Climate Change Science

PI: John Drake, ORNL

LLNL PI: Art Mirin , Philip Smith

Collaborating Institutions: Argonne National Laboratory, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Pacific Northwest National Laboratory, National Center for Atmospheric Research, NASA/Goddard Data Assimilation Office

Description: This SciDAC project will work toward transforming an existing, state-of-the-science, third-generation global climate model, the Community Climate System Model(CCSM3), to a first-generation Earth system model that fully simulates the coupling between the physical, chemical, and biogeochemical processes in the climate system.

LLNL’s role:

FACETS: Framework Application for Core-Edge Transport Simulations

PI: John Cary, Tech-X

LLNL PI: Valerio Pascucci

Collaborating Institutions: Lawrence Berkeley National Laboratory , Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, University of California-Davis, University of Utah

Description: This project focuses on the creation and deployment of scientific visualization and analytics software technology to increase scientific productivity and create new opportunities for scientific insight.

LLNL’s role:

TOPS:Towards Optimal Petascale Simulations

PI: David Keyes, Columbia University

LLNL PI: Ron Cohen,
Tom Epperly

Collaborating Institutions: Argonne National Laboratory, Colorado State University, Columbia University, General Atomics, Indiana University Lawrence Livermore National Laboratory, Oak Ridge National Laboratory, Para Tools, Inc., Princeton Plasma Physics Laboratory, Tech-X Corporation,University of California, San Diego

Description: This project will provide a multi-physics, parallel framework application, FACETS, which will enable whole-device modeling for the U.S. fusion program, and provide the modeling infrastructure needed for ITER, the next step fusion confinement device.

LLNL’s role:

Q-SIMAN: Quantum Simulations of Materials and Nanostructures

PI: Giulia Galli, UC-Davis

LLNL PI: Eric Schwegler

Collaborating Institutions: Lawrence Livermore National Laboratory, Massachusetts Institute of Technology, Stanford University, University of California, Davis, University of California, Santa Barbara, University of Illinois at Urbana Champaign

Description: This project addresses a grand challenge in materials science and chemistry: predict and design molecular and materials properties with controllable accuracy from first principles (i.e., from the fundamental laws of quantum mechanics).

LLNL’s role:

Hierarchical Petascale Simulation Framework for Stress Corrosion Cracking

PI: Priya Vashishta USC

LLNL PI:Randy Hood

Collaborating Institutions: California State University-Northridge, Harvard University, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Purdue University, University of Southern California

Description: This multidisciplinary team will develop a stress corrosion cracking computational framework consisting of modeling techniques, algorithms, analytical underpinnings, and release-quality software for petascale simulations.

LLNL’s role:

Simulations of Turbulent Flows with Strong Shocks and Density Variations

PI: Sanjiva Lele, Stanford

LLNL PI: Andy Cook

Collaborating Institutions: Lawrence Livermore National Laboratory, NASA Ames Research Center, Stanford University, University of California at Los Angeles

Description: Turbulent flows involving interactions with strong shocks and density discontinuities arise in many diverse areas of science and technology. For example: supernova explosions, volcanic eruptions, detonations of natural gas leaks, shock wave lithotripsy to break up kidney stones, as well as the implosion of a cryogenic fuel capsule for inertial confinement fusion all involve dramatic compression and expansion of multi-phase materials, their turbulent mixing and chemical reactions. Strong shock waves, strong acceleration and deceleration of heterogeneous materials and associated turbulent mixing play a critical part in these phenomena. Besides the multi-scale hydrodynamic processes, these phenomena also involve other physics and chemistry rich in its complexity and nonlinearity, such as plasma physics, radiation transport, and complex chemical kinetics. The current ability to predict these flow phenomena is strongly limited by the models of turbulence used, and by the computational algorithms employed.

This project will consider turbulent flow configurations involving shock-turbulence interaction and multi-material mixing for fundamental scientific study, and for systematic model development, for example for use in large-eddy simulations in the context of applications to accelerated multi-material flows. The research will also systematically evaluate different novel numerical approaches for nonlinear, multi-scale shock-turbulence interaction flow problems to establish the best practices and rigorous benchmarks in large-eddy simulations.

LLNL’s role: Flows involving the interaction of strong shocks with turbulence and density interfaces are central to laser-driven implosion of inertial confinement fusion plasmas, as well as in the broader Stockpile Stewardship mission of DOE. LLNL's role will be to develop and test new algorithms capable of capturing strong flow discontinuities while preserving the spectral characteristics of vorticity and turbulence over a broad range of wavenumbers.

Computational Astrophysics Consortium: Supernovae, Gamma Ray Bursts, and Nucleosynthesis

PI: Stan Woosley UC at Santa Cruz

LLNL PI: Rob Hoffman, Louis Howell

Collaborating Institutions: Johns Hopkins University, Lawrence Berkeley National Laboratory, Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Stanford University, University of California at Berkeley, University of Arizona, University of California at Santa Cruz

Description: The Computational Astrophysics Consortium will explosive phenomena, especially supernovae of all types, gamma-ray bursts, and x-ray bursts, and advance our understanding of not only the evolution of stars but also of nucleosynthesis and the mysterious “dark energy” that makes up the majority of our universe

LLNL’s role:

CASC SciDAC Archive
Computation Directorate     Computing Applications & Research Department     Advanced Simulation & Computing     Institute for Terascale Simulation     Scientific Discovery through Advanced Computing
  UCRL-WEB-225269 | Privacy & Legal Notice webmaster-casc@llnl.gov September 29, 2007  
Lawrence Livermore National Laboratory
7000 East Avenue
Livermore, CA 94550
Sponsors
Operated by Lawrence Livermore National Security, LLC
for the Department of Energy's
National Nuclear Security Administration
NNSA Department of Energy LLNL Home LLNL Home LLNL People Search Search LLNL LLNL Home Visiting LLNL Contact CASC