It takes a great deal of energy for computing systems or data centers to patch up critical errors, but what if we devoted less power to fixing less urgent issues? That is the basic idea behind EnerJ — a brand new power-saving system that can cut a chip’s energy consumption by 90 percent, because of prioritizing critical problems over people who are less threatening. Unlike, say, liquid cooling techniques , the University of Washington’s framework focuses exclusively at the programming side of the equation and revolves around two interlocking pieces of code: one which handles crucial, precision-based tasks (e.g., password encryption), and another designed to handle processes which may continue to operate, even if facing small errors. The system’s software would separate the 2 codes, meaning that energy from one element of the chip would never be used to repair an enormous problem that any other should address, while allowing engineers to more efficiently allocate voltage to every region. The system has already cut energy usage by as much as 50 percent in lab simulations, but researchers think the 90 percent threshold is easily within their reach, with computer engineering professor Luis Ceze (pictured above) predicting that the system will even be capable of increase battery life by an element of ten. The team is hoping to release EnerJ as an open-source tool this summer, but for now, you will discover additional information inside the PR after the break.
Soaring energy consumption by ever more powerful computers, data centers and mobile devices has many experts trying to reduce the energy use of those devices. Most projects to this point talk about more efficient cooling systems or energy-saving power modes.
a school of Washington project sees a task for programmers to attenuate the energy appetite of those and zeroes within the code itself. Researchers have created a system, called EnerJ, that reduces energy consumption in simulations by as much as 50 percent, and has the capability to chop energy by up to 90 percent. They’re going to present the research next week in San Jose on the Programming Language Design and Implementation annual meeting.
“We know that energy consumption is a giant problem,” said author Luis Ceze, a UW assistant professor of computer science and engineering. “With our system, cellular phone users would notice either a smaller phone, or an extended battery life, or both. Computing centers would notice a lower energy bill.”
The elemental idea is to exploit processes that could survive tiny errors that happen when, say, voltage is decreased or correctness checks are relaxed. Some examples of possible applications are streaming audio and video, games and real-time image recognition for augmented-reality applications on mobile devices.
“Image recognition already must be tolerant of little problems, like a speck of dust at the screen,” said co-author Adrian Sampson, a UW doctoral student in computer science and engineering. “If we introduce a number of more dots at the image as a result of errors, the algorithm may still work correctly, and we are able to save energy.”
The UW system is a general framework that creates two interlocking pieces of code. One is the specific part – for example, the encryption in your bank account’s password. Any other portion is for all of the processes which may survive occasional slipups.
The software creates an impenetrable barrier between the 2 pieces.
“We make it impossible to leak data from the approximate part into the ideal part,” Sampson said. “You’re completely guaranteed that cannot happen.”
While computers’ energy use is maddening and costly, there’s also a more fundamental issue at stake. Some experts believe we’re approaching a limit at the variety of transistors which can run on a single microchip. The so-called “dark silicon problem” says that as we boost computer speeds by cramming more transistors onto each chip, there may now not be any option to supply enough power to the chip to run the entire transistors.
The UW team’s approach would work like a dimmer switch, letting some transistors run at a lower voltage. Approximate tasks could run at the dimmer regions of the chip.
“After I started serious about this, it became increasingly more obvious that this can be applied, not less than slightly, to nearly everything,” Sampson said. “It appeared like i used to be always finding new places where it may be applied, not less than in a limited way.”
Researchers would use this system with a brand new kind of hardware where some transistors have a lower voltage, the force on electrons inside the circuit. This slightly increases the danger of random errors; EnerJ shuttles only approximate tasks to those transistors.
“In case you can afford one error every 100,000 operations or so, you may already save a considerable number of energy,” Ceze said.
Alternative ways to exploit hardware to avoid wasting energy are lowering the refresh rate and reducing voltage of the memory chip.
Simulations of such hardware show that running EnerJ would chop energy by about 20 to twenty-five percent, on average, counting on the aggressiveness of the approach. For one program the energy saved was almost 50 percent. Researchers at the moment are designing hardware to check their ends in the lab.
Today’s computers can also use EnerJ with a purely software-based approach. For instance, the pc could round off numbers or skip some extra accuracy checks at the approximate portion of the code to avoid wasting energy – researchers estimate between 30 and 50 percent savings in line with software alone.
Combining the software and hardware methods they think they are able to cut power use by about 90 percent.
“Our long-term goal can be 10 times improvement in battery life,” Ceze said. “i don’t believe it’s totally out of the question to have an order of magnitude reduction if we continue squeezing unnecessary accuracy.”
This system is named EnerJ since it is an extension for the Java programming language. The team hopes to release the code as an open-source tool this summer.
Co-authors of the paper are UW computer science and engineering professor Dan Grossman, postdoctoral researcher Werner Dietl, graduate student Emily Fortuna and undergraduate Danushen Gnanapragasam. Also taken with the research is doctoral student Hadi Esmaeilzadeh.
Everything Everywhere promises ‘small-scale LTE launch’ in UK by the top of 2012
ASUS MWC teaser video hints at possible hi-res tablet display?
