ECE ILLINOIS team wins Best Paper Award for processor energy efficiency research
5/31/2017 1:19:17 PM
Many emerging technology applications—such as the internet of things (IoT), wearables, implantables, and sensor networks—rely on the same ultra-low-power processor to function. But there’s a problem with this one-size-fits-all approach to using the same processor: their stated energy and power requirements are not application-specific and often are much larger than an application requires, driving up expense, weight, and size.
A team of ECE ILLINOIS researchers working at the Coordinated Science Lab have developed a novel way to measure the maximum amount of power and energy each application needs, providing more accurate estimates of the requirements for the processors. They found, on average, applications overbuilt their energy source by 40 percent.
“Current design techniques are overestimating the worst-case power needed—it’s never as bad as what design tools project,” said Weidong Ye, PhD student in ECE ILLINOIS. “When you have peak power and energy estimates that’re closer to what is needed, the power source or battery can be smaller, making the whole device smaller.”
The team, including Ye, fellow ECE ILLINOIS graduate student Henry John Duwe, III, ECE ILLINOIS Associate Professor Rakesh Kumar, and collaborators from the University of Minnesota, won Best Paper Award for this work at the 22nd ACM International Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS) in April 2017.
The paper, selected out of more than 320 submissions, details a novel approach that will impact a large segment of the processor market, as these ultra-low-power processors have rapidly become the most abundant type of processor in production today, exceeding that of personal computers and mobile processors.
“These processors are already important and are going to become an even more important segment of the market, though they sometimes don’t get attention because they are viewed as simple, embedded devices,” said Duwe. “This new methodology, though, has the potential to be very impactful to the community with broad interest and many opportunities for future application.”
Their technique provides an automated hardware-software analysis of each application, resulting in a more accurate peak power estimate than conventional approaches. Tighter measurements of peak power and energy can be exploited to reduce system size, weight, and cost.
“Emerging IoT computing applications have unique requirements—I am glad that some of our initial approaches to address these requirements are being recognized broadly,” said Prof. Kumar.
For original article visit the Coordinated Science Laboratory site