Richard Comfort Brower, Ph.D. - US grants

This award is for the development of an interdisciplinary curriculum for undergraduate students in computer science, the natural sciences and engineering. The curriculum will focus on massively parallel computing and will use Boston University's 64-node CM-5. The courses will be project- oriented emphasizing direct experience with computational problem-solving in the sciences and with programming in a massively parallel environement. This award is for the development of an interdisciplinary curriculum for undergraduate students in computer science, the natural sciences and engineering. The six project oriented courses that are proposed: one in fundamental methods and five in advanced computational topics, will emphasize direct experience with computational problem- solving in the sciences and with programming in a massively parallel environment. The results of the curriculum and materials development will be disseminated through summer workshops, publications, and presentations at professional meetings.

We propose an interdisciplinary training program, based on student recruiting in chemistry, physics and engineering, focused on the application of high performance computing to scientific research. The research focus of this program - molecular dynamics simulations which seek to characterize the behavior of complex systems of many interacting "particles" (ranging from atoms in proteins to stars in the galaxy) - is an important unifying theme in applying high performance computing to several critical areas of science and technology. Our program builds on the strengths of the Center for Computational Science at Boston University which has served as a focal point for interdisciplinary research collaboration of the PI's. Since 1988, we have pioneered the development of massively parallel supercomputing in an academic context on our Connection Machine. The GRT program will provide an organized framework of seminars, discussions and facilities which will bring graduate students from different disciplines together to focus on common computational themes which arise in the effective exploitation of high performance computing in science. The training program will include teaching and internship options.

The major component of the proposal is the development of methods designed to substantially increase the range of biomolecular systems to which advanced computational docking and design methodologies can be applied. The most successful methods and applications to date have been limited to proteins whose crystal structures are known, and which remain relatively rigid during complex formation. The main obstacle limiting the analysis of more general systems in which side chains and backbones change conformation is the target function. The proposal therefore focuses on methods for obtaining accurate, rapidly evaluatable semi-empirical free energy functions, with particular emphasis on solvation. These include methods for eliminating the time intensive surface area calculations required by current semi-empirical free energy functions, as well as several new approaches to solvation. Docking algorithms that can take full advantage of free energy- target functions will also be developed and tested in a variety of applications, exploiting the extensive crystallographic and thermodynamic databases available. Attention will be given to the systematic comparison of currently available and emerging semi-empirical free energy evaluation procedures, including the delineation of their range of applicability. J

9633385 Klein, Gould and Brower This is a new award which is funded jointly by the Office of Multidisciplinary Activities/MPS, and the Divisions of Materials Research, Mathematical Sciences and Advanced Scientific Computing. An integrated theoretical and computational investigation will be made of the structural properties of fragile glass-forming liquids and the relation of this structure to the mechanisms of relaxation. Glasses have become an increasingly important class of materials in several new technologies. They possess advantages such as light weight and ease of processing, but suffer from degradation through creep, fatigue, and embrittlement. To understand these mechanisms, it is necessary to understand the structure of glasses and how they relax. At present, experiments can provide information about relaxation processes in glasses, but do not yield much information about their structure or relation of this structure to the relaxation mechanisms. In contrast, computer simulations can provide information about structure, but only on relatively short time and length scales. Moreover, theoretical understanding of glasses is in a relatively crude state. In this research, a sequence of models will be studied beginning with a mean-field model of a glass-forming liquid and proceeding via various approximation methods to include non-mean-field effects. Several parallel computer architectures will be used to extend the size and the time scale of the systems simulated. The simulations will be based on an efficient message passing molecular dynamics program, and by more speculative methods based on Fourier acceleration algorithms and on cellular automata models. The results of the theoretical investigations and computer simulations, in conjunction with ongoing laboratory experiments at Boston University and other locations, will facilitate an understanding of the relation between the relaxation observed experimentally in deeply supercooled liqui ds and glasses and the existence of structures seen in computer simulations. Understanding of this relation will make it possible to use the structure of glasses to predict their temporal evolution. The research will be done in collaboration with colleagues at Boston University in the Departments of Physics and Electrical, Computer and Systems Engineering, and the Center for Computational Science. In addition, colleagues in the Physics Departments at Clark and Brandeis Universities, the Center for Computational Materials at NIST, and at Thinking Machines Corporation will participate. Graduate students will also participate through the NSF-sponsored Graduate Research Training Program at the Center for Computational Science at Boston University. Computer platforms include Boston University's 38-processor SGI Power Challenge array, a work station cluster and several CAM-8 machines. A workstation cluster at Clark University and the NIST Cray YMP will also be used. %%% This is a new award which is funded jointly by the Office of Multidisciplinary Activities/MPS, and the Divisions of Materials Research, Mathematical Sciences and Advanced Scientific Computing. The research combines aspects of NSF programs on Advanced Materials and Processing and on High Performance Computing and Communications. An integrated theoretical and computational investigation will be made of the structural properties of fragile glass-forming liquids and the relation of this structure to the mechanisms of relaxation. Glasses have become an increasingly important class of materials in several new technologies. They possess advantages such as light weight and ease of processing, but suffer from degradation through creep, fatigue, and embrittlement. To understand these mechanisms, it is necessary to understand the structure of glasses and how they relax. At present, experiments can provide information about relaxation processes in glasses, but do not yield much information about t heir structure or relation of this structure to the relaxation mechanisms. In contrast, computer simulations can provide information about structure, but only on relatively short time and length scales. Moreover, theoretical understanding of glasses is in a relatively crude state. The results of the theoretical investigations and computer simulations, in conjunction with ongoing laboratory experiments at Boston University and other locations, will facilitate an understanding of the relation between the relaxation observed experimentally in deeply supercooled liquids and glasses and the existence of structures seen in computer simulations. Understanding of this relation will make it possible to use the structure of glasses to predict their temporal evolution. The research will be done in collaboration with colleagues at Boston University in the Departments of Physics and Electrical, Computer and Systems Engineering, and the Center for Computational Science. In addition, colleagues in the Physics Departments at Clark and Brandeis Universities, the Center for Computational Materials at NIST, and at Thinking Machines Corporation will participate. Graduate students will also participate through the NSF-sponsored Graduate Research Training Program at the Center for Computational Science at Boston University. Computer platforms include Boston University's 38-processor SGI Power Challenge array, a work station cluster and several CAM-8 machines. A workstation cluster at Clark University and the NIST Cray YMP will also be used. ***

Large scale numerical computation for quantum field theory is entering a new phase. For the first time cost effective dedicated Terascale hardware provides the capability to simulate the full non-linearities of the fundamental theory of nuclear forces and to confront experimental data with ab initio theoretical predictions. Indeed there is a range of high precision experimental results coming from the major nuclear and particle physics laboratories that can no longer be fully understood by conventional models. They require accurate full Quantum Chromodynamics (QCD) simulations to interpret the data. However the full potential of the international lattice gauge theory initiative will be missed if new or substantially improved algorithms are not developed and optimized on the multi-Terascale architectures for the next decade. The main problem in getting accurate answers for lattice QCD is an efficient algorithm for determining the effects of quark loops on physical quantities. This effect was ignored in earlier quenched approximations which just treated the effects of the gluons of QCD. To solve this problem one needs to evaluate the eigenvalues of the (Dirac) quark propagator in the presence of external gluon fields. In this proposal we will make a concerted and sustained r effort to find new algorithms for the Dirac inverter for QCD in the context of the evolving scientific objective of the National Lattice Gauge theory initiative and the actual architectures being brought into production. Our design focus is on using the best physical and empirical knowledge of the QCD vacuum to guide the design of appropriate multi-scale Krylov space inverters which are known to dominate the compute intensive kernel of all QCD simulation codes.

0749300
Brower

0749202
Brannick

0749317
McCormick

Numerical solutions to Quantum Chromodynamcs on a lattice are critical to high precision experimental tests of the standard model and an ab-initio understanding of nuclear matter. The core of these calculations involves inverting a Dirac matrix which becomes increasingly ill conditioned as the lattice is refined. Consequently while Terascale computing hardware has exposed this new physics, it is incapable of fully accommodating it. On the other hand, if lattice QCD algorithms are reformulated to exploit and reveal the physics at this finer microsale, Petascale hardware does have the potential for opening up a new era of physics discovery. This award brings together a close collaboration of leading experts in applied mathematics and theoretical physics to meet this challenge by the application of new multi-level algorithms for QCD simulations. The central mission of the proposed Multigrid Quantum Chromodynamics at the Petascale project (MGQCD) is: to develop new and significantly more robust multigrid methods for enabling more complex and higher fidelity physics for lattice QCD calculations; to support their migration into Petascale simulations; and to engage the broader scientific community through collaborative research and educational activities that highlight the multigrid methodology.

0946441
Brower

This proposal will be awarded using funds made available by the American Recovery and Reinvestment Act of 2009 (Public Law 111-5).

This project will operate a small experimental GPU cluster configured for application physicists and applied mathematicians working in Lattice Field Theory (LFT) and Computational Fluid Dynamics (CFD). The research goal is to develop new algorithms and programming strategies to fully realize the potential of emerging many-core technologies,as well as to implement these algorithms in production application codes on Graphical Processing Units.
With Nvidias CUDA framework and the widely-supported OpenCL standard, such architectures appear eminently suited for parallel scientific applications. The project will focus on rethinking the underlying algorithms to properly map them to the hardware, with an emphasis on employing many GPUs in parallel.

We propose to arrange US participation in the workshop entitled "Intelligent Software: The Interface Between Algorithms and Machines", to be held on October 19-21, 2009 at the University of Edinburgh, hosted by the Centre for Numerical Algorithms and Intelligent Software (NAIS)[1], that will support research in the efficient application of algorithms to next generation computer architectures. The goal of this workshop is to bring together leading US and European researchers at the interface between computational science, applied mathematics, and computer science to discuss algorithms pervasive in the scientific applications community, and their implementation and deployment for evolving target architectures. By bringing together researchers working together across this divide, a scheme will be suggested with the goal of advancing computational science infrastructure so that it can live up to its full potential.

This award supports the development of software tools for advanced algorithms on a cluster of high performance graphics processing units(GPUs). The initial goal of this library will focus on a single driver application, the fundamental numerical study of Nuclear Forces (Lattice Quantum Chromodynamics: LQCD), and a second target application, the numerical simulation of the exciting nano-technology of graphene. These are both multi-fermion problems well suited to solution via multi-scale algorithms on many-core architectures. This pilot project draws on experience gained by the small team at Boston University and Harvard in developing Dirac solvers for lattice field theory. Two building blocks from prior research are (1) the construction of an adaptive multigrid (MG) solver for the Wilson Dirac operator of LQCD, which demonstrates a 20x speedup compared to the best Krylov solvers in production code and (2) a highly optimized Krylov solver for the same operator on GPUs (but without multigrid), realizing a 10x improvement in price/performance over traditional clusters. The library will unite these feature and generalize their domain of applicability.

As an example of the broader impact, it is estimated that combining these two technologies (MG algorithms and GPU architectures) will yield a 100-fold improvement in price/performance for the most compute-intensive component of LQCD simulations. Such an advance would be truly transformative, making an immediate impact in nuclear and particle physics. At the same time, it will serve as a prototype of the more generic problem of mapping hierarchical algorithms onto heterogeneous architectures, a challenge of paramount importance on the path to the exascale. The software library will be designed to bring similar benefits to graphene technology and to evolve to accommodate additional target application and additional domain decomposition algorithm to mitigate the communication bottleneck of Exascale designs. The award will provide partial support for two postdoctoral scholars who play essential roles in this project.

Large, complex, multi-scale, multi-physics simulation codes, running on high performance computing (HPC) platforms, are essential to advancing science and engineering research in disciplines such as lattice field theory, astrophysics and cosmology, computational fluid dynamics/fluid structure interaction,and high energy density physics. Progress in computational science together with the adoption of high-level frameworks and modular development have produced widely used community simulation software specific to individual communities. These state-of-the-art codes have been under development and optimization for several years and currently simulate multi-scale, multi-physics phenomena with unprecedented fidelity on petascale platforms. Currently each of these codes have solvers with varied performance characteristics, but all face challenges because of changing hardware architecture. Efforts underway to cope with these challenges, are largely fragmented. While it is true that the scientific codes used in various domains differ significantly from one another, many solutions are likely to be conceptually similar, even if they differ in details. The goal of the proposed conceptualization project, Software Institute for Methodologies and Abstractions for Codes (SIMAC) is to find common abstractions and frameworks applicable across a broad range of applications through cooperation, coordination and interdisciplinary interactions among the participants. The core group of participating codes includes FLASH (astrophysics, cosmology, CFD, HEDP), Cactus (CFD, numerical relativity, and quantum relativity), the code suite used by the Lattice QCD community, and Enzo (cosmology).

The proposed collaborative research will produce benefit beyond the four simulation codes and collaborating institutions by exploring: a common software infrastructure applicable to a broad range of science and engineering application domains; an engagement model between computer science research and application development; a multidisciplinary immersion program for research, education and training of students, postdoctoral fellows and visitors on future platform architectures.