Wu Feng, associate professor of computer science and electrical and computer engineering at Virginia Tech’s College of Engineering, received an AMD Research Faculty Fellowship by Advanced Micro Devices Inc. (NYSE:AMD), based in Sunnyvale, Calif.
The award further supports Feng's research in heterogeneous computing, particularly with graphics processing units – commonly referred to as GPUs – as well as the emerging accelerated processing units – known as APUs – from Advanced Micro Devices. Accelerated processing units fuse a traditional processor or central processing unit, dubbed CPU, with a graphics processing unit.
These heterogeneous computing environments are expected to significantly speed-up many general-purpose computation tasks normally executed by a computer’s central processing unit. This research addresses the gamut of computing from large-scale supercomputing systems all the way down to handheld mobile devices, and further work of compacting supercomputers into a small space, a focus of Feng’s research.
“The AMD Research Faculty Fellowship has already catalyzed critical research and development in heterogeneous parallel computing which encompasses using multiple types of processors such as the central processing unit, and emerging graphics processing units,” said Feng. “Due in large part to this fellowship, we now routinely re-purpose the graphics processing unit, which traditionally has been used to drive video displays, to perform massively accelerated general-purpose computation.”
Feng added that he also is at work on creating a toolbox of automated computer language translators that enable an end-user to move his or her application code written in one language into a different language that can run on any type of processor.
“We are pleased to make this award, in recognition of Wu Feng’s accomplishments in power-aware computation and heterogeneous computing and as an acknowledgement of our past experience in working with him,” said Jay Owen, director of Advanced Micro Devices’ External Research Office. “We look forward to his illuminating research findings of benefit to the industry at large.”
The central processing unit has served as the brain of a computer since the 1960s, Feng said. The computing speeds of a central processing unit previously doubled every 24 months by increasing its clock speed, and therefore software got a “free ride” to better performance. However, with clock speed increases slowing, such automatic performance gains can no longer be taken for granted, he added.
“Vendors now increase computational horsepower of a computer by increasing the number of brains, or cores in a processor, such as an AMD OpteronTM processor, as well as by increasing the types of brains – for example, an Intel or AMD processor coupled with a 1,600-core AMD RadeonTM HD 5870 graphics processing unit, resulting in the notion of heterogeneous parallel computing,” Feng said. “However, because of a lack of trained workforce in the art of parallel programming, much of the computational potential of these powerful, multiple-core, heterogeneous computing systems is not yet realized.”
The fellowship, Feng said, will help him conduct research to train such a workforce to realize this potential, thus accelerating the way that computers process information. Feng has worked closely with AMD in the past. In 2004, when energy efficiency in supercomputing was not yet considered an industry priority, Feng co-invented intelligent, self-adapting, energy-efficient software called EnergyFit as part of his Supercomputing in Small Spaces project. This software was initially developed on an AMD Opteron processor.
Feng conducts related research in heterogeneous computing as part of the National Science Foundation’s Center for High-Performance Reconfigurable Computing. He also directs the Synergy Laboratory which conducts basic and applied research in high-performance computing at the synergistic intersection of systems software, middleware, tools and applications software. He also holds an adjunct professorship appointment with the Virginia Bioinformatics Institute at Virginia Tech and with the School of Medicine and its Translational Science Institute at Wake Forest University.
AMD Research is comprised of senior and associate researchers working from major AMD offices, provides technology direction and conducts research critical to the company’s future as a premier semiconductor design firm. Through its External Research Office, the group provides leadership for collaboration with strategic research partners.