Jaswinder Singh - US grants

In this effort research and education at the boundary of parallel systems and applications will be performed. It is motivated by the need for application insights to make fundamental advances in parallel systems, and the recent convergence in multiprocessor architectures to general purpose nodes connected by a communication architecture. On the research side, challenging, irregular, dynamic applications will be parallelized on real systems in close collaboration with domain researchers, particularly in probabilistic protein structure determination and adaptive unstructured meshing, and their implications for parallel architecture and software understood. General techniques for partitioning such complex applications will be developed, particularly for the emerging class of cache- coherent distributed-memory multiprocessors. The applications will be distributed as benchmarks to the parallel computing community. Architectural implications, particularly communication demands, scaling and resource distribution, will be examined, and scaling methodologies developed. Funds permitting, a range of commercial systems will be evaluated using applications, to understand the impact of their communication architectures and tradeoffs, and insights into programming and performance debugging environments studied. On the education side, a new undergraduate course in parallel architecture will be developed, and a textbook on parallel architecture is being written to articulate the convergence and develop parallel architecture as an engineering rather than exotic discipline.

9806751
Singh, Jaswinder Pal
Li, Kai
Princeton University

Experimental Software Systems: Integrated Applications and Systems Research in Software Shared Memory

A coherent shared address space (SAS) is an attractive programming model for parallel computing. It is emerging as commercially successful when implemented in hardware on tightly coupled multiprocessors. With less tightly coupled clusters of workstations or symmetric multiprocessors becoming important platforms, it is important to support this model in software on clusters as well. Otherwise, the need to run applications on both types of platforms may drive the more difficult explicit message passing model to dominate. This research develops and evaluates software SAS systems on clusters of various scales and organizations. The research is distinguished by being highly application-driven, including collaboration with application scientists. It examines protocols/systems and applications simultaneously without treating either as fixed. Protocols and systems are enhanced based on bottlenecks encountered in real applications, the role of hardware support is examined, and different software approaches are compared. The focus is on both programming ease and performance, and includes comparing the SAS and message passing models on tightly-coupled systems and clusters. The research will result in new, scalable software SAS systems, a greater integration of applications and systems research that is now necessary for major advances, and a better understanding of the tradeoffs in programming models across platforms.

9818308
Singh, Jaswinder Pal
Ostriker, Jeremiah
Princeton University

Applications and Programming Environments for Cache-Coherent Distributed Memory Machines

This research instrumentation enables research projects in:
- Parallel Applications and Programming Environments for Computational Cosmology on Cache-Coherent DSM Architectures,
- Parallel Applications and Programming Environments for Computational Protein Structure Determination on Cache-Coherent DSM Architectures, and
- Parallel Workload for System Evaluation and their Implications for Software and Hardware.
To support integrated application-driven research in the aforementioned projects, this award contributes to the purchase of an upgrade to an existing 64-processor SGI Origin2000 multiprocessor by the Departments of Computer Science and Astrophysics together with the Princeton Plasma Physics Laboratory at Princeton University. The equipment will be used to continue and enhance the research projects listed and to support new projects between computer science and the Plasma Physics Laboratory. The first project, a collaboration between Computer Science and Astrophysics, deals with problems particularly challenging for parallel implementation due to the combination of interacting models and a highly irregular dynamic nature. The goals are to develop parallel algorithms and implementations for a range of computations, to develop generalizable and portable techniques preserving locality while providing load balance, to extend existing parallel languages; use runtime systems of these languages, to build libraries of partitioning techniques and
algorithms. The second project, a collaboration between CS and the School of Medicine at Stanford University, with similar goals, simulates inherently slow processes in structural biology. The last project will broaden the set of workloads available. The SPLASH suites of parallel applications in critical emerging application areas serve as benchmarks for architectural studies, as well as testbeds for programming languages and compilers.

This Integrative Graduate Education and Research Training (IGERT) award supports establishment of a multidisciplinary graduate training program of education and research at Princeton to provide the next generation of computer scientists and users of high-performance computing with integrated graduate training across disciplinary boundaries. On one hand, dramatic advances in computer technology are poised to revolutionize computational science as an equal partner of theory and experiment, but their nature is such that harnessing them will require cross-disciplinary fertilization and expertise in both application areas and computer science. At the same time, advances in computer science will increasingly be driven by knowledge of applications, both scientific simulations and others. The PICAS program will address the entire computational and information pipeline in a variety of areas, from models and methods through parallel algorithms and systems to immersive visualization and information management. It will train a new breed of researchers to cross-disciplinary boundaries, develop new areas between disciplines, and exploit synergies. Centered around the computer science department, the program will include many departments throughout the university. Program components include curriculum development, cross-department advising, integrated research across disciplines, cross-cutting annual thematic programs to focus collaboration, and activities like a recent interdisciplinary seminar series that attracts many researchers from various local institutions.

IGERT is an NSF-wide program intended to facilitate the establishment of innovative, research-based graduate programs that will train a diverse group of scientists and engineers to be well-prepared to take advantage of a broad spectrum of career options. IGERT provides doctoral institutions with an opportunity to develop new, well-focussed multidisciplinary graduate programs that transcend organizational boundaries and unite faculty from several departments or institutions to establish a highly interactive, collaborative environment for both training and research. In this second year of the program, awards are being made to twenty-one institutions for programs that collectively span all areas of science and engineering supported by NSF. This specific award is supported by funds from the Directorates for Computer and Information Science and Engineering, for Mathematical and Physical Sciences (Office of Multidisciplinary Activities), for Engineering, and for Education and Human Resources.

EIA-9975011
Princeton University
Kai Li

Next Generation Software: Adaptive, Performance-Portable Software for Next-Generation and Immersive Applications

A new generation of applications is becoming very important for high-performance computing, including collaborative design, interactive walkthroughs and large data visualization, and telepresence. They require tremendous resources including CPU, memory, storage, and audio/visual devices, and they have substantially different characteristics, performance goals and system interactions than traditional scientific applications. For example, they have extremely irregular and unpredictable data access needs and workload distributions, they interact more dynamically and with many more types of input/output sensors and devices, they involve dynamic user interaction and steering, and their goal is to deliver the best possible quality at a fixed output refresh rate rather than a solution of fixed quality in the minimum possible time. As computer architectures become more complex, it becomes increasingly difficult to develop such applications to achieve the desired performance. Three properties are critical: (i) high performance for rich interactive behavior, (ii) adaptability and isolation in all layers (i.e. the complexity, and unpredictability demand that each layer of application or system software must adapt to the layers above and below it-through performance modeling and through runtime feedback and adaptation-and should try to shield the neighboring layers from each other's complexity), and (iii) performance portability across component upgrades and across the different major types of platforms that may be used in such environments. Our goal is to develop the software building blocks, runtime systems and design methodologies to assist such application development.

This MRI award will serve to purchase a 256 Processor/1536 core SGI Altix UV1000 with 9.2 TB of RAM and 14.4 TB of raw scratch disk space. The instrument will provide scientists at Princeton University, the Institute for Advanced Studies, and partner institutions with the computational resources needed to model multi-scale phenomena in the sciences and engineering. The flexibility of the Altix architecture, which supports both shared and distributed memory applications, along with an outstanding bus architecture to support the addition of extra processing units such as GPGPUs is an ideal platform for developing algorithms for multi-scale problems. The setting of the instrument in the University?s High Performance Computing Research Center will facilitate a cross-disciplinary approach combining expertise in applied mathematics, computer science and domain-specific disciplines enabling innovative approaches for memory intensive applications. The new instrument will play an essential role in educating a new generation of scientists and training students across many disciplines in the use of advanced modeling tools on modern computer platforms, contributing to new graduate student certificate programs offered by PACM, the Program in Applied and Computational Mathematics, and PICSciE, the Princeton Institute for Computational Science and Engineering. Finally, the instrument will provide a necessary link between local and national facilities, preparing the Princeton scientific community to the emerging multicore and massively parallel architectures of the future.

The instrument will enhance international scientific cooperation by contributing to projects like the Munich-Princeton collaboration in cosmological computational science, and will contribute to science education of the general public through collaboration with the American Museum of Natural History in New York City, with planned new visualizations for use in the Cosmos series and in conjunction with the Museum?s ongoing public education work on earthquakes and geologic movement. Women?s participation in computational projects enabled by the instrument will set examples to encourage greater access of women to science. Finally, access provided to partners at
California State University-Northridge will contribute to training minority scientists and engineers.