InTheLoop | 04.02.2012
April 2, 2012
Berkeley Lab-Led Institute to Help Solve Data-Intensive Science Challenges
As one of the nation’s leading funders of basic scientific research, the Department of Energy has a vested interested in helping researchers effectively manage and use large datasets. To this end, DOE’s Office of Advanced Scientific Computing Research has announced a five-year project, the Scalable Data Management, Analysis, and Visualization (SDAV) Institute, which will be led by Berkeley Lab’s Arie Shoshani. Read more.
DOE Report Calls for Paradigm Shift in Applied Math Research
The Report of the DOE Workshop on Mathematics for the Analysis, Simulation, and Optimization of Complex Systems, of which CRD’s John Bell and David Brown are co-authors, has just been released by DOE. The report, titled A Multifaceted Mathematical Approach for Complex Systems, identifies several priority research directions for DOE-funded mathematics, and comments that the “type of research activities envisioned for multifaceted mathematics represents a paradigm shift from the traditional single-investigator model for applied mathematics research.... This type of paradigm shift will enable applied mathematicians to make a significant impact on scientific and engineering challenges that face DOE.”
Revealing the Secrets of Clean Coal
As the demand for power increases, and renewable energy sources remain years away from offsetting this need, one fact is clear: Coal is here to stay. So researchers supported by the Department of Energy are investigating relatively “clean” methods for extracting energy from coal—like gasification. Using NERSC systems, a scientist from the University of Utah has developed tools to model and validate the complex processes of coal gasification. Read more.
HPC Must Ramp Up Efficiency to Deal with Extreme-Data Flood
To move toward exascale computing, scientists must focus on saving energy used in computing in every possible way, according to an article in ASCR Discovery. The energy-conservation requirements go beyond the hardware to each fundamental process of computing, including data management, analysis and visualization. As scientists struggle to manage the massive amount of data generated by experiments and simulations, the challenges of data-intensive science and exascale computing begin to merge.
Several Computational Research Division researchers are quoted in the article: Arie Shoshani discusses I/O strategies; Hank Childs comments on data mining; and Wes Bethel points out the impact of data storage methods on I/O. A global storm simulation by Michael Wehner and Prabhat provides the first illustration for the article. Read more.
NUFO Science Exhibition on Capitol Hill Includes NERSC, ESnet
The National User Facility Organization (NUFO) showcased user facility science at an outreach event on Capitol Hill last week (March 28–29). This event highlighted the significant role that scientific user facilities play in science education, economic competitiveness, fundamental knowledge, and scientific achievements. It consisted of back-to-back exhibits in the House and Senate office buildings and included posters from each NUFO facility, including NERSC and ESnet. In addition, there were specialty posters on industrial involvement and small business generation, and scientific accomplishments.
Scientists from each user facility were available to answer questions. David Skinner and Julian Borrill represented NERSC, while Craig Tull and Chris Tracy represented ESnet.
The Berkeley Lab contingent included representatives from the Advanced Light Source, the National Center for Electron Microscopy, the Molecular Foundry, and the Joint Genome Institute. They met with Rep. Nancy Pelosi and Senator Barbara Boxer, as well as member of the House Energy and Water Development Appropriations Subcommittee.
Brainstorming Sessions to Build Better Support for Scientific Software Use, Development
Ever wonder why we have training in Microsoft Word, but not in Python? Do you want to see the Lab support software developer training, user groups, code repositories and build systems? Wish you knew other software developers working on software problems similar to yours? Computing Sciences staff are organizing three brainstorming sessions to discuss ways to build community and support mechanisms for software developers across the Lab. Interested staff are invited to join any of the three sessions:
- 12:00–1:00 pm Wednesday, April 4, in Perseverance Hall (54-130)
- 3:00–4:00 pm Thursday, April 5, at OSF (room 238)
- 10:00–11:00 am Monday, April 9, in Perseverance Hall (54-130)
For more information or to suggest topics for the agenda, contact Deb Agarwal at email@example.com or x7078.
CS Diversity Working Group to Sponsor Summer Students
The CS Diversity Working Group will once again be able to sponsor three or four diversity summer students. If you have selected a student to work with for the summer who is from an under-represented population and would like them sponsored, send a request to have the student sponsored to firstname.lastname@example.org. Include the student’s name, who they will be working with, and the position description. The slots will be awarded on a first-come, first-served basis, and applications are already coming in, so act quickly. This program is made possible by the generous support of CS senior management.
Safety Note: Skunks Have the Right of Way
There have been two sightings of skunks recently, outside the Building 50 complex and near the path to the cafeteria. Skunks are most likely to be seen in the early morning or evening; fortunately, their stripe enhances visibility. If you see one, keep your distance, and the skunk will most likely do the same. If you have any concerns, please contact CS Safety Coordinator Betsy MacGowan.
This Week’s Computing Sciences Seminars
Interactive Data Graphics: Exploration, Explanation, and Storytelling
Tuesday, April 3, 11:00 am–12:00 pm, Soda Hall, Wozniak Lounge, UC Berkeley
How do you tell a story with a chart? How can interactive graphics allow readers to ask questions and discover stories of their own? We will look at examples from the award-winning digital book “Our Choice” and elsewhere, and consider how interactive data design can balance exploration and explanation.
Alternating Direction Methods for Sparse and Low-Rank Optimization: LAPACK Seminar
Wednesday, April 4, 12:10–1:00 pm, 380 Soda Hall, UC Berkeley
Shiqian Ma, IMA
In this talk, we propose several alternating direction type methods for solving sparse and low-rank optimization problems. We propose alternating linearization methods (ALM) that can solve unconstrained composite optimization problems. We show that the iteration complexities of the basic and accelerated ALMs are O(1/eps) and O(1/sqrt(eps)), respectively, for obtaining an eps-optimal solution. We also propose an inexact alternating direction method of multipliers that can solve convex optimization problems with three separable blocks. The global convergence result is established. Numerical results on robust PCA, sparse PCA and latent variable graphical lasso with millions of variables are reported to demonstrate the efficacy of our proposed methods.
Using Cloud to Reduce Operations and Rock Science!
Wednesday, April 4, 2:00–3:00 pm, 50F-1647
Sebastian Stadil, Scalr, Inc.
You got your PhD to do science, right? So how come you spend so much time preparing machines and applying for grants? Well, the cloud can’t help you with the grants, but it sure can help you reduce the time you spend on provisioning, configuring, and operating compute infrastructure. Come listen to Sebastian Stadil, founder of the Scalr open source cloud management project, talk about the tools you can use to get that unfair advantage over your arch-nemesis across the Bay.
Benchmarking and Protecting Adiabatic Quantum Computation: Quantum Information and Computation Seminar
Wednesday, April 4, 2:00–3:00 pm, 410 Hearst Memorial Mining Bldg., UC Berkeley
Daniel Lidar, University of Southern California
USC and Lockheed Martin recently jointly founded a new quantum computing center housing the 128 qubit Rainier chip built by D-Wave. I will report on our efforts to benchmark the chip, by comparing its performance in solving random instances of spin glasses against classical solvers. I will also describe our very recent theoretical and experimental work on designing decoherence-protected adiabatic quantum computation.
New Approaches to Light Transport
Wednesday, April 4, 2:30–3:30 pm, 373 Soda Hall, UC Berkeley
Wenzel Jakob, Cornell University
The central goal of physics-based rendering is generating photorealistic images by simulating the flow of light through increasingly elaborate mathematical models of our world. Research in this area entails finding suitable approximations of the relevant physical phenomena, as well as efficient numerical algorithms to solve the resulting equations. In this talk, I will discuss my work on both of these fronts.
I will begin with a description of light transport through volumetric materials whose composition exhibits an inherent internal structure; good examples of this are cloth, hair, and skin. The standard radiative transfer framework does not apply here, since it assumes a certain rotation invariance that these materials do not satisfy. I will describe how to incorporate structured volumes into physics-based rendering, and what repercussions this has on the underlying equations and solution methods.
I am also interested in obtaining a better understanding of the geometry of light paths. In most scenes light travels only along a subset of the space of possible paths; for specular surfaces these are the Fermat paths, which form a lower-dimensional manifold in path space. This simple observation has far-reaching implications that I will describe in the second part of the talk. The resulting theory cleanly incorporates specular scattering (formerly an awkward corner case) into the standard framework and has algorithmic implications for glossy and volumetric scattering. The end result will be a new rendering technique that creates images by “walking around” on the manifold of paths informed by its high-dimensional differential geometry.
I will conclude the talk with short review of other graphics research I have conducted, followed by an outline of avenues for future work.
EECS Colloquium: Advancing the Impact of Research: The Push for Open Access Publishing
Wednesday, April 4, 4:00–5:00 pm, 306 Soda Hall (HP Auditorium), UC Berkeley
David Hodges and David Wagner, UC Berkeley
Now that most scholarly and professional publication is electronic, costs of distribution are largely unrelated to volume of usage. Thus the old subscription-based publication business model needs revision. UC libraries and many others are demanding action to reduce the cost of library acquisitions. Open Access is an obvious choice.
Many authors prefer that there should be no charge for access to published research. This is possible if the up-front costs of peer review, editorial work, electronic indexing and archiving, etc. are covered via authors fees. (Fees range from $500 to $3000.)
PLoS One is a spectacularly successful non-profit open-access electronic journal based on author fees. It directly challenges the grip of the for-profit subscription-based publishers Nature, Elsevier, and Science in the life sciences, and is expanding into other fields of science and engineering.
Non-profit publishers IEEE, ACM, Usenix, now dominate in EECS publishing. Where are they on the path to open access? Or are they candidates for replacement? That is the focus for our brief presentations, followed by questions and discussion.
Dave Hodges is the current IEEE VP for Publications. David Wagner is an advocate for open access within the ACM.
GCC and C++, Again
Friday, April 6, 11:00 am–12:00 pm, 50F-1647
Paolo Carlini, Oracle
About a year of development later, a new GCC series is out, sporting a good number of smaller and larger improvements (and some regressions!), as usual. For example, the new C++11 Standard, finally released a few months ago, required the implementation of a new set of flexible memory built-ins, exploited by the runtime library; transactional memory support is available for the first time, both for C and C++. The presentation will go through some of the new features, in particular those likely to be of interest to the parallel programming community, without hiding at the same time a few known issues and also some recurring criticisms (e.g., diagnostics quality) hopefully to be in part addressed over the next year.
Understanding Quaternions: Visual Computing Lab Lunch Talk
Friday, April 6, 12:00–1:00 pm, 510 Soda Hall, UC Berkeley
Ron Goldman, Rice University
Ever since their discovery over a century and a half ago, it has been known that quaternions can be employed to rotate vectors in three dimensions. Today contemporary computer graphics uses quaternions to:
- provide compact representations for rotations and reflections of vectors in three dimensions;
- avoid distortions due to floating point computations during rotations;
- enable key frame animation by spherical linear interpolation.
Other applications of quaternions in computer graphics and geometric modeling include tubing and texturing curves and surfaces, visualizing streamlines, generating three-dimensional Pythagorean hodograph curves, and constructing conformal transformations on triangular meshes. Yet while the formal algebra of quaternions is well known in the graphics and modeling communities, the derivations of the formulas for this algebra and the geometric principles underlying this algebra are not well understood. The goals of this talk are:
- to provide a geometric interpretation for quaternions, appropriate for contemporary computer graphics;
- to present better ways to visualize quaternions, and the effect of quaternion multiplication on points and vectors in three dimensions;
- to develop simple, intuitive proofs of the sandwiching formulas for rotation and reflection;
- to show how to apply sandwiching with quaternions to compute perspective projections (NEW).
Automatically Enhancing Locality in Irregular Applications
Friday, April 6, 1:30–3:00 pm, 50F-1647
Milind Kulkarni, Purdue University
Over the past several decades of compiler research, there have been great successes in automatically enhancing locality for regular programs, which operate over dense matrices and arrays. Tackling locality in irregular programs, which operate over pointer-based data structures such as trees and graphs, has been much harder, and has mostly been left to ad hoc, application specific methods. In this talk, I will describe efforts by my group to automatically improve locality in a broad class of irregular applications, those that traverse trees. The key insight behind our approach is an abstraction of data structure traversals as operations on vectors. This abstraction lets us design transformations, predict their behavior and determine their correctness. I will present two specific transformations we are developing, “point blocking” and “traversal splicing,” and show that they can deliver substantial performance improvements when applied to several real-world irregular kernels.
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.