InTheLoop | 06/24/2013
The weekly newsletter for Berkeley Lab Computing Sciences employees
June 24, 2013
Unusual Supernova Is Actually Perfectly Normal
August 2011 saw the dazzling appearance of the closest and brightest Type Ia supernova since Type Ia’s were established as “standard candles” for measuring the expansion of the universe. Labeled SN 2011fe, the supernova was caught by the Palomar Transient Factory less than 12 hours after it exploded in the Pinwheel Galaxy with help from supercomputers and networks at NERSC and ESnet. Now, astronomers are using it to build a benchmark atlas for normal Type Ia’s. Read more.
Shreyas Cholia Named NERSC OSP Deputy Group Lead
Shreyas Cholia has been named Deputy Group Lead for NERSC’S Outreach, Software, and Programming (OSP) Group. OSP Group Lead David Skinner is now spending a significant amount of time on NERSC science engagement and data topics. Shreyas will help with daily operational OSP duties as well as the following topical areas:
- NERSC software services for staff, source code control, software engineering, github services
- Authentication, SSO, and NERSC’s Certification Authority
- Web services for users such as science gateways, Globus Online, science database services, and NERSC user web software
- Development of middleware technologies and APIs to facilitate new modes of user science at NERSC, such as NEWT and other RESTful APIs, Globus, HTC frameworks, AMQP.
NPR Science Friday: A Calculating Win for China's New Supercomputer
China’s “Tianhe-2” (Milky Way 2) supercomputer took first place in one recent speed test, clocking in at 30 quadrillion calculations per second—about twice as fast as the best American machines. The U.S. still has more supercomputers than any other nation, but some experts say computer speed is a measure of a country’s scientific innovation, and worry the U.S. is lagging behind. Last week, NPR’s Science Friday discussed this situation with Berkeley Lab Deputy Director Horst Simon and Argonne’s Associate Lab Director Rick Stevens. Listen here.
Sudip Dosanjh to Give Keynote Address at TERATEC Forum
The TERATEC 2013 Forum, the international meeting for simulation and high performance computing, is being held this week, June 25–26, at Ecole Polytechnique in Palaiseau, France. NERSC Director Sudip Dosanjh will give a keynote address on “Exascale and Big Data at NERSC,” presenting NERSC’s technology roadmap through 2015. NERSC Chief Technology Officer John Shalf will give a presentation on “Active Power Management Technology Challenges and Implications for Programming Models.”
TERATEC is an organization of 86 industrial and technology companies, universities, and research laboratories dedicated to high-performance simulation and computing.
NERSC Training Event: Introduction to HPC using GPUs
NERSC, NVIDIA, and The Portland Group will present a one-day workshop “Introduction to High Performance Computing Using GPUs” on Thursday, July 11, 2013 on the University of California Berkeley campus in Sutardja Dai Hall room 250. This workshop is targeted at users who are familiar with C/C++ and/or Fortran and high performance computing and want to learn about programming for GPUs.
Registration is free and open to all NERSC users; Berkeley Lab researchers; UC students, faculty, and staff; and users of the Oak Ridge Leadership Computing Facility. The workshop will be broadcast to remote viewers. For more information and to register (both local and remote attendees), go here.
Reminder: Retirement Celebration for Bailey, Hules at 3:00 pm Wed., June 26
Associate Lab Director Kathy Yelick invites all Computing Sciences staff to a reception from 3:00–5:00 pm Wednesday, June 26, to recognize the contributions and impending retirements of David Bailey and John Hules, both of whom have helped raise the visibility of CS. The event will be held in the Lab cafeteria and refreshments will be served.
Since joining the Lab in 1998, Bailey has served as chief technologist first for NERSC, then for CRD, and then as leader of the Complex Systems Group. He is well known for his work in HPC benchmarking, exploring the digits of pi, and numerous books and papers on applied mathematics.
Hules joined CS in 1997 and has been responsible for developing and producing the NERSC annual report since then. He has also helped shape and edit numerous other reports, from strategic plans to operational assessments. He has also gained a strong reputation as a diligent proofreader of papers, articles, etc.
This Week’s Computing Sciences Seminars
Memory Access Optimization for Iterative Tomography on Many-Core Architectures
Monday, June 24, 11:00 am–12:00 pm, 50B-4205
Wim van Aarle, University of Antwerp, Belgium
Iterative reconstruction methods for tomography, despite their virtues, are known to be slow compared to analytic reconstruction methods, mainly because of the computationally very intensive forward and backward projection operations. Fortunately, relying on modern many-core architectures with large vector registers, such as the recently introduced Intel Xeon Phi coprocessor, can offer relief. However, to optimally benefit from such systems, the peak performance of the algorithms should not be bound by the memory bandwidth. In this presentation, strategies are discussed that improve the performance of tomographic projection operations by optimizing their memory accesses. Data locality is exploited to hide data access latency, and knowledge of the cache architecture is used to optimally distribute the projection operation over many computing cores. This work has been done in collaboration with Pieter Ghysels and Wim Vanroose.
Scalable and Flexible Bulk Architecture
Wednesday, June 26, 10:00–11:00 am, 380 Soda Hall, UC Berkeley
Xuehai Qian, University of Illinois at Urbana-Champaign
To help parallel computing become mainstream, one of the main design considerations for multicore architectures should be support for programming productivity. This means designs that can deliver high performance and efficiency while relieving programmers and compiler writers from managing low-level tasks, and designs that help minimize the chance of parallel programming errors.
In this talk, I will present an overview of the ideas behind the Scalable and Flexible Bulk Architecture. The architecture has a set of novel techniques for programmability, while retaining scalability and flexibility. In particular, I will present Volition, the first hardware scheme that detects Sequential Consistency Violations (SCVs) in a relaxed-consistency machine precisely, in a scalable manner, and for an arbitrary number of processors in the cycle. Volition enhances programmability, while inducing negligible traffic and execution overhead. Moreover, I will present the scalable cache coherence protocols for the atomic block execution.
CS Summer Student Brownbag: DOE Joint Genome Institute
Thursday, June 27, 1:30–3:00 pm, OSF
Kirsten Fagnan and Shane Canon, LBNL/NERSC
New Features and Directions for Tools and MPI
Friday, June 28, 12:00–1:00 pm, OSF 238
Kathryn Mohror, Lawrence Livermore National Laboratory
Please join us for discussions of the changes and upcoming features in the MPI Standard for performance analysis tools and debuggers. Bring your wish list for new features you’d like to see for performance and debugging in MPI!
Link of the Week: Sizing Up Big Data, Broadening Beyond the Internet
Internet companies were just the start. Virtually every field, from science to sports to public health, is being transformed by data-driven discovery and decision-making, reports Steve Lohr in The New York Times article “Sizing Up Big Data, Broadening Beyond the Internet.” This is the lead article in a special section, Big Data 2013, which discusses many different applications, from politics to aesthetics.
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 6,000 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 DOE 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.