Locating the impossible with 'lightening fast' speed

December 20, 2010 By Jason Kornwitz

Northeastern University PhD students Perhaad Mistry (left) and Dana Schaa (right) work in their campus office. Both students recently presented GPU-computing research at Supercomputing'10 in New Orleans, the largest computer show in the world. Credit: Mary Knox Merrill

A terrorist plants a time bomb along a gas line in a residential neighborhood. He e-mails a photo of the death trap to law enforcement officials, but no one can tell exactly where the bomb is located.

The solution may lie in work being done in Northeastern University's Computer Architecture Research Laboratory, where electrical and computer engineering professor David Kaeli and his team have developed supercomputing hardware/software technology to pinpoint the location of people, buildings—or even bombs—10 to 15 times faster than traditional computing methods.

Northeastern researchers are collaborating on the project with colleagues at the University of Virginia and Advanced Micro Devices (AMD), a Sunnyvale, California-based company that develops computer processors for the commercial and consumer sectors.

The innovative technology, which aligns with Northeastern’s commitment to research that solves global challenges in health, security and sustainability, showcases the value of using Graphics Processing Units (GPUs) to help protect the nation’s critical infrastructure, said Kaeli.

“GPU technology will have huge implications for the intelligence community,” he said, noting the plan to open-source the application in the near future.

Here’s how it works: A user uploads a photo to an image database such as Google Earth. The GPU-based application enables the user to run a lightning-fast search, comparing the photo against images in the database until it finds a match.

“There’s no application like it,” said Perhaad Mistry a doctoral student in electrical and computer engineering who helped develop the idea while on co-op at AMD. “Pinpointing certain people or places needs to be done quickly.”

Industry leaders have recently recognized Northeastern’s cutting-edge research in this area of supercomputing.

• NVIDIA, a multinational company that develops graphics processing units, designated Northeastern’s Department of Electrical and Computer Engineering, in conjunction with Massachusetts General Hospital, as a NVIDIA CUDA Research Center for its ongoing development of GPU-accelerated medical image analysis tools. (CUDA stands for Compute Unified Device Architecture, a parallel computing architecture developed by NVIDIA.)

• AMD named Northeastern a Strategic Academic Research Partner for conducting research with the company’s GPUs. AMD also awarded Northeastern a $65,000 grant to conduct GPU-research and invited the Northeastern team to present at Supercomputing 2010, the largest international conference on high performance computing, networking, storage and analysis.

Kaeli is proud of the partnership. “AMD is a major employer of Northeastern computer engineering students and a leader in the field of GPU computing,” he said.