March 12, 2012
Berkeley and Beijing Team Up to Tackle Light Source Data Tsunami
As detector technologies improve, light source scientists believe that limitations in software and computer systems to manage and analyze experimental data are among the biggest barriers to producing science. Now, Berkeley Lab and Beijing Computational Science Research Center (CSRC) are teaming up tackle this problem. Read more.
Scientists Discover a New Kind of Neutrino Transformation
Some unprecedentedly precise measurements from the Daya Bay Neutrino Experiment are revealing how electron neutrinos and their antineutrino counterparts “oscillate” into different flavors as they travel. This finding may eventually solve the riddle of why there is far more ordinary matter than antimatter in the universe today. Read more.
CRD Researchers Participate in Workshop on Extreme-Scale Solvers
Last week (March 8–9) the DOE Office of Advanced Scientific Computing Research sponsored a Workshop on Extreme-Scale Solvers: Transition to Future Architectures in Washington, DC. The objective was to bring together experts in the development of scalable solvers to determine research areas needed for extreme-scale algorithms and software to effectively utilize 100 petaflop systems and prepare for exascale systems.
Osni Marques and Esmond Ng of CRD were on the organizing committee for the workshop. Other CRD participants included John Bell, Sherry Li, Sam Williams, and CRD Division Director David Brown.
This Week’s Computing Sciences Seminars
Reducing the Effective Stiffness in Computational Fluid-Structure Interaction Problems
Monday, March 12, 10:00–11:00 am, 50F-1647
Lauren Fovargue, University of North Carolina
The immersed boundary (IB) framework for biological fluid-structure interaction problems has been widely used to study everything from heart mechanics to insect flight. However, it is well known that the computational modeling techniques used in IB cause the time evolution equations to be very stiff. It has been shown that the time step required for stability is significantly smaller than that imposed by the fluid solver alone. This increased computational cost is significant in most cases, and current efforts to reduce this cost have yet to been seen in computing applications. These efforts, including fully implicit, parallel and adaptive techniques, may not be popular due to the complexity of the algorithms and the need to rewrite the fluid solver. I will describe a semi-implicit, multi-rate strategy for stiff immersed boundary problems based on the blob projection method and spectral deferred corrections. The method utilizes a splitting in both the temporal and spatial domains to evaluate the stiffest terms in the evolution created by the immersed boundary on smaller time scales.
Auto-Tuning Stencil Computations and Multigrid Solvers
Monday, March 12, 10:30–11:30 am, 50B-4205
Cy Chan, Massachusetts Institute of Technology
Auto-tuning has become an essential tool for achieving high performance for scientific codes on parallel hardware platforms while maintaining portability. In this talk, we will discuss two recent projects. The first is a novel auto-tuning framework that can be applied to a generalized class of stencils. While previous auto-tuners have focused on specific stencil kernels, our framework strives to allow the tuning of general stencil kernel inputs without much additional work from the programmer. The second project is a novel tunable multigrid algorithm that focuses on high-level algorithmic flexibility by optimizing cycle shapes using a recent auto-tuning programming language and compiler called PetaBricks.
Application-Aware Strategies for Managing Performance-Energy-Resilience Tradeoffs
Tuesday, March 13, 10:00–11:00 am, 50B-4205
Manu Shantharam, Pennsylvania State University
In current petascale systems and beyond, energy efficiency and resilience are increasingly important in addition to performance. In this talk, I will focus on two recent results that indicate the interplay between performance, energy, and resilience. I will discuss opportunities for leveraging tradeoffs between these attributes.
First, I will illustrate the challenges posed by soft errors on supercomputing systems. I will analyze the propagation of a single soft error during the solution of a sparse linear system using iterative methods and discuss results of an empirical evaluation. Next, I will propose techniques that seek to enhance overall system energy and throughput efficiencies using application speedup-profiles. I will demonstrate their effectiveness at delivering overall system energy improvement by managing the tradeoffs between faster workload completion and slower execution of one or more applications within the workload.
I will conclude my talk with a brief overview of some of the other results from my thesis and future research directions.
Printed Electronic Systems: The Confluence of Printing and Semiconductors
Wednesday, March 14, 12:00–1:00 pm, 310 Sutardja Dai Hall (Banatao Auditorium), UC Berkeley
Live broadcast at mms://media.citris.berkeley.edu/webcast
Vivek Subramanian, UC Berkeley
In recent years, there has been significant interest in the applications of printed electronics for the realization of flexible displays, fully-printed RFID tags and embedded sensors. Printing of active circuitry is expected to enable a dramatic reduction in the overall cost of these systems, allowing for integration of electronic barcodes and product quality detection systems into consumer goods, as well as ushering in an era of low cost flexible displays and content delivery appliances. Printing techniques that have been considered range from high-speed commercial gravure printing through ultra-scaled inkjet printing. While many of the printing techniques under consideration have evolved from techniques already widely deployed in graphic arts applications, the requirements for printed electronics are in many ways dramatically different from those that exist for conventional graphic arts.
In this talk, I will review the tremendous progress that has occurred in printed electronics over the last decade, and will discuss the challenges that remain. I will discuss the challenges associated with utilizing printing to realize printed semiconductor-based circuits. Additionally, I will overview the state of the art in printed electronic materials. I will review our work on developing materials, processes, devices, and circuit architectures for all-printed electronic systems including RFID tags, displays, and sensing systems.
Link of the Week: “Oddball” Dark Matter Baffles Astronomers
Dark matter and galaxies parted ways in the collision of two galaxy clusters 2.4 billion light-years away. The new results from NASA’s Hubble Space Telescope are contrary to predictions, and now astronomers are left trying to explain dark matter’s seemingly oddball behavior in the Abell 520 merging galaxy cluster.
During the collision of galaxy clusters that formed Abell 520, the dark matter collected into a “dark core” containing far fewer galaxies than would be expected if the dark matter and galaxies hung together. Most of the galaxies apparently have sailed far away from the collision. Current theories of dark matter predict that galaxies should be anchored to the invisible substance, even during the shock of a collision. The initial observations, made in 2007, were so unusual that astronomers shrugged them off as the result of poor data. Instead, to their chagrin, the Hubble observations helped confirm the earlier findings. Read more.
About Computing Sciences at Berkeley Lab
The Lawrence Berkeley National Laboratory (Berkeley Lab) Computing Sciences organization provides the computing and networking resources and expertise critical to advancing the Department of Energy's research missions: developing new energy sources, improving energy efficiency, developing new materials and increasing our understanding of ourselves, our world and our universe.
ESnet, the Energy Sciences Network, provides the high-bandwidth, reliable connections that link scientists at 40 DOE research sites to each other and to experimental facilities and supercomputing centers around the country. The National Energy Research Scientific Computing Center (NERSC) powers the discoveries of 7,000-plus scientists at national laboratories and universities, including those at Berkeley Lab's Computational Research Division (CRD). CRD conducts research and development in mathematical modeling and simulation, algorithm design, data storage, management and analysis, computer system architecture and high-performance software implementation. NERSC and ESnet are Department of Energy Office of Science User Facilities.
Lawrence Berkeley National Laboratory addresses the world's most urgent scientific challenges by advancing sustainable energy, protecting human health, creating new materials, and revealing the origin and fate of the universe. Founded in 1931, Berkeley Lab's scientific expertise has been recognized with 13 Nobel prizes. The University of California manages Berkeley Lab for the DOE’s Office of Science.
DOE’s Office of Science is the single largest supporter of basic research in the physical sciences in the United States, and is working to address some of the most pressing challenges of our time. For more information, please visit science.energy.gov.