A-Z Index | Phone Book | Careers

InTheLoop | 02.24.2014

February 24, 2014

NERSC Featured on CBS Radio's 'Science Today'

NERSC Director Sudip Dosanjh will be featured on "Science Today," starting this week. The one-minute radio program is produced by the University of California for the CBS Radio Network. In the spot, Dosanjh will explain NERSC's reasons for choosing its newest flagship supercomputer, Edison. »Check the Science Today site for more.

ESnet, Berkeley Lab Host Workshop on Operating Innovative Networks

The next installment of the Operating Innovative Networks (OIN) workshop series will be held Feb. 27-28 at Lawrence Berkeley National Laboratory, drawing 50 university and laboratory network engineers from as far away as New Jersey, Texas, Illinois and Colorado. The workshop is co-hosted by ESnet, Berkeley Lab and CENIC, the Corporation for Education Network Initiatives in California. The program is designed to give attendees the knowledge and training needed to build next-generation campus networks that are optimized for data-intensive science.

Presented by experts from the Department of Energy’s ESnet, Indiana University and Internet2, the workshop series will focus on Science DMZ network architectures, perfSONAR performance measurement software, data transfer nodes and emerging software defined networking techniques. Combined, these technologies are proven to support high-performance, big data science applications, including high-volume bulk data transfer, remote experiment control, and data visualization. The workshops will consist of two days of presentation material, along with hands-on sessions to encourage immediate familiarity with these technologies.

The Berkeley workshop is the fourth in the series and future sessions are planned in Atlanta (March), Boston (May) and Oregon (July). »Learn more. »Download a detailed agenda (PDF | 91KB).

Disordered Materials Hold Promise for Batteries

Because lithium batteries are among the lightest and most energetic rechargeable batteries available, a lot of research is being done to facilitate their use in electronic devices. Using supercomputers at the National Energy Research Scientific Computing Center (NERSC) and other facilities, researchers at the Massachusetts Institute of Technology (MIT) and Brookhaven National Laboratory (BNL) have found a new avenue for such research: the use of disordered materials, which had generally been considered unsuitable for batteries. »Read more.

Seeing the Great Lights of Europe

Beamline scientists at user facilities such as the Lab's Advanced Light Source (ALS) are faced with unprecedented amounts of data, but not always with the hardware or software needed to effectively manage, analyze and share that data. David Brown and Craig Tull of the Computational Research Division and Alex Hexemer of the ALS toured Europe's light sources and computing facilities to get a better picture of how this situation is being addressed elsewhere in the world. »Read more.

March 8: Memorial Service for CRD's Gail Jackson-Maeda

A memorial service will be held Saturday, March 8, for Gail Jackson-Maeda, a Berkeley Lab administrative assistant who died Feb. 5 after a long battle with cancer. The service will be held at 1:30 p.m. at the Unity Center, 1871 Geary Road, in Walnut Creek. Jackson-Maeda joined Berkeley Lab in 2008 and was the administrative assistant for the Advanced Computing for Science Department (ACS) in the Computational Research Division. Memorial donations may be made to any breast cancer research program. She is survived by her husband and son.

Feb. 27 & 28: Art in Science Exhibit, Lectures

This free, two-day event to be held from 5:30 – 9 p.m., Thursday and Friday, Feb. 27 – 28, explores the intersection of image and research with an exhibit, lectures, hands-on demonstrations and music. View original and intriguing images of scientific investigation and meet the scientists and artists who created them, including the Lab's own photographer, Roy Kaltschmidt. Browse multimedia exhibits, get hands-on with origami, and stay for cool presentations on the cutting edge of art-in-science. The event will take place at the Energy Biosciences Building on the UC Berkeley campus, 2151 Berkeley Way. »Learn more.

Reminder: Video Conference Room Training Tomorrow

At 2 p.m. on Tuesday, Feb. 25, there will be a drop-in training session for the recently upgraded video conferencing system in 50B-2222. Anyone may attend. No registration necessary.

This Week's CS Seminars

»CS Seminars Calendar

Big Impact of an Ultra-scale Eigenvalue Computation on the K Computer

Monday, February 24, 1:00pm – 2:00pm, Bldg. 50F, Room 1647

Toshiyuki Imamura RIKEN Advanced Institute for Computational Science, Japan

We introduce the newly developed eigensolver EigenExa. EigenExa is developed for a petascale parallel computer, for example the "K computer." Last summer, we examined its performance by using all the resources of the K computer (82,944 processors). We have successfully solved the world's largest-scale dense eigenvalue problem, one million dimensions, in 3,464 seconds. The basic algorithm and its implementation will be presented.

Solving Generalized Eigenvalue Problems by a Simultaneous Band Reduction and a Divide-and-Conquer Method

Tuesday, February 25, 2014, 1:00pm - 2:00pm; Bldg. 50F, Room 1647

Yusuke Hirota, RIKEN Advanced Institute for Computational Science Japan

We introduce a new solution method for a generalized eigenvalue problem A \bm{x} = λ B \bm{x}, where A and B are dense real symmetric and B is positive definite. The method simultaneously reduces A and B to symmetric banded matrices and directly computes the eigenvalues and the eigenvectors of the banded generalized eigenvalue problem by a divide-and-conquer method without solving standard eigenvalue problems. In this talk, we will address the the algorithms of the band reduction and the divide-and-conquer method and discuss the performance.

A Performance Analysis of the TSQR Algorithm

Wed, February 26, 1:00pm – 2:00pm, Bldg. 50F, Room 1647

Takeshi Fukaya RIKEN Advanced Institute for Computational Science Japan

The TSQR algorithm for computing the QR factorization of a tall and skinny matrix is now one of the well-known communication-avoiding algorithms. In this talk, we will first report the performance of the TSQR algorithm on recent computers in Japan, such as the K computer. Then we will discuss the performance of the TSQR algorithm referring to the obtained performance data and its performance model. In particular, we will mention the possibility that there is a situation in which the traditional parallelized Householder QR algorithm is faster than the TSQR algorithm. In addition to the above discussion, we will also present an approach to estimate the performance gap between those two algorithms.

Link of the Week: More Women in Computer Science Classes

UC Berkeley, Stanford and a handful of other universities have experienced a marked uptick in the numbers of female computer science students, reports the San Francisco Chronicle. Those increases have also coincided with a reimagining of computer science classes, especially introductory ones. At the University of California, Berkeley, a computer science class called the "Beauty and Joy of Computing" has enrolled more women than men since last spring. It was the first time since at least 1993 — as far back as university enrollment records are digitized — that a majority of women enrolled in an introductory computer science course. »Read more.