Serial Computing Is Dead; the Future Is
Parallelism
SearchDataCenter.com (06/30/08) Botelho, Bridget
Serial computing is extinct and the future belongs to parallel computing,
argued Dave Patterson, head of the University of California, Berkeley's
Parallel Computing Laboratory, during his keynote speech at the Usenix
conference. Parallel processing can now be executed on a single chip
across multiple cores, thanks to the emergence of multicore chips, while
Patterson contended that serial computing has reached its limits in terms
of memory and power. He maintained that programmers who require greater
performance must write programs capable of leveraging multiple cores via
parallelism, and researchers at his lab have concentrated on applications
that ought to be parallelized. The proper writing and implementation of
parallel programs can address power issues and performance bottlenecks.
But computer scientist Andrew S. Tanenbaum, a recipient of the Usenix
Lifetime Achievement Award, said that writing parallel applications can
give rise to more problematic software rather than less. "Sequential
programming is really hard, and parallel programming is a step beyond
that," he said. "I have a great fear that we will have all of these cores,
and our software programs will be even worse." Patterson noted that the
success of parallel computing depends on its ability to improve efficiency,
accuracy, and productivity, but he cautioned that the majority of
programmers are not ready to write suitable parallel programs.
Click Here to View Full Article
to the top
Study Refutes Niche Theory Spawned By the Web
Wall Street Journal (07/02/08) P. B5; Gomes, Lee
In his 2006 book "The Long Tail," Wired magazine editor Chris Anderson
outlined a theory that society is "increasingly shifting away from a focus
on a relatively small number of 'hits' [mainstream products and markets] at
the head of the demand curve and toward a huge number of niches in the
tail" as a result of the vast number of choices facilitated by the Web, but
a new study published in the Harvard Business Review counters that
assumption, writes Lee Gomes. An analysis of data for online video rentals
and song purchases has led Harvard marketing professor Anita Elberse to
conclude that online and offline shopping patterns are essentially the
same. She says the importance online shoppers ascribe to hits and
blockbusters is growing rather than shrinking thanks to the Web. Elberse
also cites qualitative social research implying that Anderson's theory may
have incorrectly characterized consumers as being eager to escape the
limitations of physical inventories so that they can enjoy a wider variety
of niche products. She notes that there is a major element of social
conformity in cultural consumption, in that consumers tend to want to
experience the same things others are experiencing. Gomes acknowledges
that patterns of cultural consumption are definitely being reshaped by the
Web, but says these changes do not appear to be having the kind of dramatic
leveling effect on demand curves as forecast by the Long Tail. "While
whole new cultural categories--YouTube videos, for example--are indeed
emerging, they seem to quickly settle into the same winner-take-all dynamic
experienced in the pre-Google age," Gomes concludes.
Click Here to View Full Article
to the top
45th Design Automation Conference (DAC) Awards Best
Papers to Authors From Texas A&M University and Stanford University
Business Wire (06/30/08)
Researchers from Texas A&M University and Stanford University received
Best Paper awards during the 45th Design Automation Conference. Texas
A&M's Wei Dong, Peng Li, and Xiaoji Ye were honored for the technical paper
"WavePipe: Parallel Transient Simulation of Analog and Digital Circuits on
Multi-core Shared-memory Machines." The researchers offer a new method for
parallel transient simulation that promises to make computer-aided design
tools much faster. Stanford's Sung-Boem Park and Subhasish Mitra discuss a
new debugging approach for application-specific instruction-set processors
in the paper "IFRA: Instruction Footprint Recording and Analysis for
Post-silicon Bug Localization in Processors." The strategy offers
high-speed information scanning. The researchers will receive a $1,500
prize as part of the award. "Both of these papers highlight topics of high
importance in the EDA industry," says DAC's Patrick Groeneveld. "We are
proud of DAC's continuing role as a focal point for the latest research and
development in our industry."
Click Here to View Full Article
to the top
Milepost Project: Speed Mobile Web App Development
Dr. Dobb's Journal (06/30/08)
IBM and its European Union partners will move forward with the MachIne
Learning for Embedded PrOgramS opTimization (Milepost) project, an
initiative that could make it easier for developers to respond to the
customization and personalization demands of consumers for new mobile
applications. Milepost will include artificial intelligence technology
that automatically learns how to best optimize new programs for embedded
processors in mobile devices. "The Milepost solution uses artificial
intelligence and machine learning to understand what kind of compiler
optimizations are optimal for use with each new hardware design," says Mike
O'Boyle, professor of computer science at the University of Edinburgh and
project coordinator for Milepost. "This will help completely automate
compiler construction and enable more rapid code design of hardware and
software--dramatically reducing time to market in these systems." Milepost
unveiled a prototype version of the software at the GCC Summit, and it was
able to improve the performance of a state-of-the-art compiler by 10
percent within one month, rather than several years. The partners plan to
release a full version of the compiler optimization software into the GCC
main product at the end of the three-year project, and the Milepost GCC
version will be made available to the open source community when it is
released in June 2009.
Click Here to View Full Article
to the top
Adobe Teams With Google, Yahoo for Flash Search
eWeek (07/01/08) Taft, Darryl K.
Adobe, Google, and Yahoo are working together to make Flash content
searchable on search engines. Adobe's Eric Wittman says Adobe is providing
optimized Adobe Flash Player technology to Google and Yahoo to enhance
search engine indexing of the Flash file format so search engines can read
information that is currently unavailable. The collaborative effort will
provide more relevant automatic search rankings for the millions of rich
Internet applications (RIA) and other dynamic content that runs in Adobe
Flash Player. The effort means that RIA developers and rich Web content
producers will no longer need to amend existing and future content to make
it searchable, says Adobe's Justin-Everett Church. "There are millions of
things built in Flash and there have been concerns because of search engine
compatibility," Church says. "We have a piece of technology to remedy this
and we're working with Google and Yahoo. Adobe evangelist Ryan Stewart
says the Yahoo and Google Flash Player will programmatically move through
the stages of a Flash application to obtain data from the server, capturing
text and data inside the Flash-based application. Consultant Bill Hunt
says the agreement is a particularly important breakthrough for companies
with large, interactive sites that are built with Flash and RIAs because
they can now be found, crawled, indexed, and ranked.
Click Here to View Full Article
to the top
Robot Learns to Use Tools
Technology Review (07/01/08) Grifantini, Kristina
University of Massachusetts Amherst roboticists are developing the UMass
Mobile Manipulator (UMan), a robot that can learn to use objects it has
never encountered before. UMan pushes objects around a table to observe
how they move. Once the robot identifies an object's moving parts, it
starts experimenting with the object and attempts to use it for different
tasks. UMan uses a Web cam to find objects, and has software that analyzes
differences between adjacent pixels to guess where objects are located.
The robot then prods the object and revises its estimate of the object's
shape based on how the object moves. UMan will continue to push and prod
the object until it is satisfied that it understands how the object moves.
Wherever movement is restricted the robot concludes that there is a joint.
UMan then uses that information to determine how to manipulate the object
and how different elements, such as multiple joints, interact. UMan is
currently not equipped to pick up objects, instead it just manipulates them
on the surface of the table. The robot has successfully learned how to
manipulate scissors, shears, and several types of wooden toys. UMan has a
single arm about a meter long that has seven degrees of freedom, which
makes the arm very similar to a human arm in flexibility. The researchers
expect UMan will soon be able to use past experiences as a guide to
handling new objects.
Click Here to View Full Article
to the top
Tongue Drive System Lets Persons With Disabilities
Operate Powered Wheelchairs, Computers
Georgia Institute of Technology (06/30/08) Vogel, Abby
Georgia Institute of Technology professor Maysam Ghovanloo and graduate
student Xueliang Huo have developed the Tongue Drive system, a new
assistive technology that enables individuals with disabilities to operate
various devices by moving their tongues. "This device could revolutionize
the field of assistive technologies by helping individuals with severe
disabilities, such as those with high-level spinal cord injuries, return to
rich, active, independent and productive lives," Ghovanloo says. The
Tongue Drive system uses a tiny magnet attached to an individual's tongue.
The user moves their tongue to direct a cursor across a computer screen or
to move a wheelchair. Ghovanloo says the tongue was chosen as the control
point because it connects to the brain by a cranial nerve that generally
avoids damage in severe spinal cord injuries or neuromuscular disease.
Tongue movements are also fast, accurate, and do not require a significant
amount of concentration or effort. An array of magnetic field sensors
mounted on a headset outside the mouth, or on an orthodontic brace inside
the moth, track the magnet's movement. Sensor output signals are
wirelessly transmitted to a portable computer, which can be carried on the
user's clothing or wheelchair, where they are processed to determine the
relative motion of the magnet in respect to the array sensors in real time.
The system can recognize a large number of tongue movements, each of which
can represent a different command, and a unique set of movements can be
tailored for each individual.
Click Here to View Full Article
to the top
Mobile Users Make Same Mistakes as Disabled PC
Users
University of Manchester (07/01/08)
Able-bodied people who use mobile phones make the same errors at a similar
frequency as physically impaired users of desktop computers, according to
new research from the University of Manchester. Researchers in
Manchester's School of Computer Science reviewed an earlier study on
physically disabled users from scientists at the University of Edinburgh,
and put mobile users through the same experiments. They found that mobile
users also press the wrong key and press the same key repeatedly by
mistake, often click the wrong area of the screen and click the screen
multiple times in error, and make mistakes when trying to drag and drop
information. The researchers suggest that small assistive computer
programs could also be used by mobile users. "In recent years solutions
have been built to help disabled users and it is hoped these solutions ...
can now be applied for the benefit of mobile phone users," says researcher
Tianyi Chen. The Engineering and Physical Sciences Research Council funded
the two-year Reciprocal Interoperability between Accessible and Mobile Webs
(RIAM) project.
Click Here to View Full Article
to the top
As Intel Nears 40, Technologist Offers His Look Into
Future
Wall Street Journal (07/01/08) P. B6; Clark, Don
Intel's Patrick Gelsinger, one of the company's top technologists, expects
the number of computing engines packed on each chip to increase
significantly in the future. He says Intel is building a foundation for
the "many core" era, in which products will feature tens to hundreds of
electronic brains. Although the extra processing power will not affect
common computing jobs, Gelsinger says programmers will take advantage of it
to enable many new applications. For example, he says medical images that
take hours to process will become instantly available and interactive,
improving and quickening diagnoses. Accurate speech recognition will
replace typing, and the basic interface software that controls how
computers look and feel will evolve to better represent what users want.
Gelsinger says future computers will become immersive, intuitive, and
interactive. He also predicts that computing and Internet capability will
eventually be available to every person on the planet, 24 hours a day.
Such ubiquitous connectivity will use combinations of microprocessors,
sensors, and other devices to improve and further enhance the capabilities
of products such as ATMs, cars, and door locks, creating a world where
humans interact with computers without even realizing it.
Click Here to View Full Article
to the top
Cisco, IBM, Intel, Juniper and Microsoft Fight Cyber
Terror Together
Network World (06/27/08) Greene, Tim
Cisco, IBM, Intel, Juniper, and Microsoft have formed the Industry
Consortium for Advancement of Security on the Internet (ICASI) in an effort
to respond faster to "global, multivendor cyber threats." Such attacks
have a wider impact because they target multiple products or protocols in
products, and they pose problems for both end users and vendors. The forum
will serve as a platform that encourages members to share critical data
about specific attacks more readily. ICASI plans to improve practices for
addressing multi-product security threats, and set security response
standards that it can share with the rest of the industry. "To date there
has not been a trusted vendor environment that allows companies to
identify, assess, and mitigate multi-product, global security challenges
together on the customers' behalf," the group said in a statement. "ICASI
aims to fill this void."
Click Here to View Full Article
to the top
'Gordon Gekko' Trading Bot Profits From Mood
Swings
New Scientist (07/01/08) Graham-Rowe, Duncan
Trading software is increasingly supplanting flesh-and-blood traders in
foreign exchange and similar markets because of programs' faster reaction
time to market events. University of Southampton researcher Krishnen
Vytelingum has co-developed a new trading program that can adjust how
aggressively it trades to changing market conditions. The program can
change its trading behavior based on how aggressively other traders are
behaving, or based on a projection of future market conditions culled from
past market trends. Vytelingum says trading algorithms could become more
profitable if they could recognize market dynamics. He developed the
program using simulated markets, which offer a more comprehensive test than
real markets. When benchmarked against the best agents, the program could
generate 5 percent more profit on average than more conventional agents,
and boost market efficiency when utilized by all traders, Vytelingum
says.
Click Here to View Full Article
to the top
To Sing Like Shakira, Press 'One' Now
American Friends of Tel Aviv University (06/30/08)
A computer has been successfully trained to assess the quality of
vibrato--the pulsating change of pitch in a singer's voice--by a Tel Aviv
University research group, which has also developed a biofeedback-based
application designed to coach singers. The application can train singers
to emulate the vibrato qualities that are most appealing to the human ear,
and it recently earned first prize at the International Cultural and
Academic Meeting of Engineering Students in Istanbul. TAU researcher Noam
Amir and his team fed a computer numerous recordings by students singing
vibrato who had their vibrato rated by human teachers, and employed
mathematical measurements to correlate vibrato styles to their quality as
judged by the teachers. The computer was then able to evaluate the vibrato
quality of new voices by itself, and generate ratings similar to those
given by the experts. The researchers gave singers the ability to
visualize and enhance their vibrato in real time through the addition of a
biofeedback loop and a monitor. Amir thinks this research could also be
applied to automated call centers, and he hopes to be able to train
computers to recognize a spectrum of emotions so that a live receptionist
can intercede when a caller gets upset.
Click Here to View Full Article
to the top
Swarm Robotics: Debugged Naturally for 120 Million
Years
Control Engineering (06/20/08)
At the recent Sensors Expo, keynote speaker MIT roboticist James McLurkin
noted that software for swam robotics has been debugged naturally for 120
million years and outlined three objectives engineers should strive to
accomplish. McLurkin told attendees that they should work to help young
people understand that engineering is cool by describing what they do and
by getting involved in engineering-related activities for young people.
McLurkin also said the United States is not creating enough engineers.
Engineers and companies should donate time, hardware, or software to
educational institutions. For example, he said the Eclipse open-source
programming toolkit has been very useful at universities. Finally,
McLurkin said that engineers can help find a way out of global warming by
working on it themselves and inspiring the next generation of engineers to
solve the problem. McLurkin said he is working with MIT Computer Science
and Artificial Intelligence Laboratory professor Leslie Kaelbling on
distributed algorithms for swarms of mobile robots in an effort to
understand how to use local interactions among nearby robots to produce
large-scale group behaviors from the entire swarm. Small, swarm robots
could search through rubble after an earthquake or other disasters to find
survivors, relaying information back to medium-sized robots that analyze
the data and direct larger robots to remove the rubble.
Click Here to View Full Article
to the top
Speak Up: Devices and Programs Are Getting Better at
Translating Languages
Economist (06/25/08)
The shortage of translators in Iraq has created a demand for machines that
can translate between Arabic and English. Several two-way translating
devices have been developed as part of the Spoken Language Communication
and Translation System for Tactical Use (TRANSTAC) program, run by the
Defense Advanced Research Projects Agency. IBM, BBN Technologies, and SRI
are the three main TRANSTAC participants. Although IBM and SRI have
developed systems that can reliably translate tens of thousands of words
between Iraqi Arabic and American English, TRANSTAC manager Mari Maeda says
translation devices still need improvement. Hardware needs to become more
reliable and error rates need to be reduced to one word in six from the
current one in 10, and names of people and places are still a significant
hurdle for the machines. Despite these shortcomings, significant progress
has been made that is helping expand the use of translation devices and
systems. One reason is that special hardware is not always necessary to
run translation programs. SRI's Iraqcomm program runs on Windows XP and
can work on the rugged notebook computers carried by soldiers and foreign
civilians. Like many language systems, Iraqcomm uses statistical models to
recognize speech patterns and sounds, to filter out sounds that are not
speech, and to predict the next word in a sentence.
Click Here to View Full Article
to the top
PCs Getting Smarter, But Brains Too Hot
Nikkei Weekly (06/23/08) Vol. 46, No. 2342, P. 16; Kawai, Tomoyuki
As data processing speed and capacity increase, so does the heat generated
by CPUs, and heat dispersion has become a pressing issue for PC makers. If
current trends continue, computer chips will be running at speeds above 10
GHz and producing enough heat to melt themselves by 2010. Chipmakers are
working on ways to reduce the chips' power consumption to lower heat
output, with one strategy being to design multicore CPUs that can more
efficiently process commands. Accelerated processing and power consumption
reduction can be implemented simultaneously via a multicore architecture,
and IBM is developing a system that uses a multicore CPU to rapidly
generate 3D images of the brain from sets of cross-sections captured by a
magnetic resonance imaging device. The power consumption of IT equipment
is expected to increase by a factor of five by 2025 and by a factor of 12
by 2050, according to Japan's Ministry of Economy, Trade, and Industry.
Cooling down CPUs is likely to help alleviate the problem of global
warming, when one considers not just the IT equipment, but all the consumer
electronics that are in use, along with their chips' growing power
consumption.
Click Here to View Full Article
- Web Link to Publication Homepage
to the top
Smart HCCI Cars: They'll Talk to Themselves, and to the
Pump
CITRIS Newsletter (06/08) Slack, Gordy
CITRIS researchers are developing engines that are more fuel efficient and
release fewer emissions than current engines, and one area of research is
the Homogeneous Charge Compression Ignited (HCCI) engine, which blends the
cleanliness of spark-ignition engines with the efficiency and adaptability
of diesel engines. HCCI engines are designed to premix fuel like a spark
engine, but then compression-ignite the fuel the way a diesel engine does.
However, the compression ignition creates a problem in which the engine's
advantages are undercut by temperature variations across the cylinders.
University of California at Berkeley engineer Robert Dibble says each HCCI
cylinder needs to be subjected to constant monitoring and adjustment so
that the engines' practicability can be maintained, and he and his
collaborators are developing sensors and controllers that will keep
temperatures constant throughout the engine or compensate for temperature
differentials by modifying pressure ratios within the cylinders. Wireless
sensors that will report on interior cylinder conditions many times a
second are being developed by UC Berkeley professor Albert Pisano, while
Dibble and UC Berkeley engineering professor Van Carey have been discussing
with oil companies the possibility of using wireless communication
connections between the vehicle and the fuel pump. "The car can pull up to
the station and tell the pump that its efficiency is low, for instance,"
Carey says. "And the pump responds, perhaps by determining that the engine
is having incomplete combustion and needs a higher octane fuel." Carey
says the sensors could also provide data to the station on the vehicle's
oil, fan belts, coolant, tire pressure, and brakes to help the car continue
operating at its highest levels of safety and efficiency.
Click Here to View Full Article
to the top
Who's Got the Time?
Government Computer News (06/23/08) Vol. 27, No. 15, Jackson, Joab
The discontinuity between computerized and human timekeeping complicates
the continued operation of the Internet and computer networks, which depend
on accuracy and precision. "Time is a perfect example of something that
needs to be taken out of the realm of human interaction because we don't do
it well, and machines do it well," says former Homeland Security Department
metadata program manager Michael Daconta. Machines' interaction with time
typically takes one of two forms, says Sita architect Stephen Colebourne:
Marking a period of time or gauging an interval between events. Demand for
more granular levels of accuracy is rising as networks and computers
increasingly depend on time, and Internet Engineering Task Force engineers
are working to divide time measurements in the Network Time Protocol into
even finer chunks. Whereas atomic clocks that the U.S. Naval Observatory
uses to set the official U.S. time count the number of atomic vibrations,
National Institute of Standards and Technology physicist Till Rosenband is
developing an even more precise atomic clock that counts ion vibrations.
Human or sidereal time must be adjusted every few years because the solar
system's precision lags behind that of atomic clocks due to the fact that
that the human day is slowly lengthening. Observers say a problem could
crop up from computer systems' growing dependency on accurate time
measurement, and some researchers have proposed the elimination of the leap
second and strict adherence to atomic time as a solution. The use of
time-zone offsets employed by all operating systems is suggested by Steve
Allen, a researcher at the University of California's Lick Observatory.
Click Here to View Full Article
to the top