Despite all the promise of nanoscale computer components and nanoscale devices, there is a downside: Experts say the circuits and chips can become less reliable and more expensive to produce, leading to variability in behavior from device to device over the course of their lifetime.
The National Science Foundation (NSF) recently awarded a $10 million, five-year grant to help address the problem. The funding agency’s Expeditions in Computing program, which funds projects that “promise significant advances in the computing frontier and great benefit to society,” has given the green light to a research project that rethinks and enhances the role software can play in a new class of computing devices that are adaptive and highly energy-efficient. Specifically, the research team is exploring “Variability-Aware Software for Efficient Computing with Nanoscale Devices,” a mission to develop computing systems that sense the nature and extent of variations in their hardware circuits and expose these variations to compilers, operating systems and applications, driving adaptions in the entire software stack.
Such variability-aware computing systems would benefit the entire spectrum of embedded, mobile, desktop and server-class applications by dramatically reducing hardware design and test costs for computing systems, while enhancing their performance and energy efficiency, researchers say. Applications like search engines and medical imaging systems would also benefit, but the project team’s initial focus will be on wireless sensing, software radio and mobile platforms with plans to transfer these advances to additional areas moving forward.
“We envision a world where system components – led by proactive software – routinely monitor, predict and adapt to the variability in manufactured computing systems,” says Rajesh Gupta in prepared remarks about the project grant. Gupta is director of the Variability Expedition and a professor of computer science and engineering at the University of California, San Diego’s Jacobs School of Engineering. “Changing the way software interacts with hardware offers the best hope for perpetuating the fundamental gains in computing performance at lower cost of the past 40 years.”
As transistors and components on chips get smaller, semiconductor makers are experiencing lower yields and more variability, which means more components are being thrown away because they don’t meet the timing-, power- and lifetime-related specifications. Researchers on the program maintain that the trend toward parts that can’t reliability scale in capability or cost will cripple the computing and information technology industries, if not addressed. A fluid software-hardware interface, the researchers maintain, will mitigate the variability of manufactured systems and make them more robust, reliable and responsive to changing operating conditions.