Intel opens platform for EECS researcher’s groundbreaking work
Intel Corporation recently made approximately 30 cutting-edge heterogenous hardware platforms available for use to researchers selected for the second round of its Hardware Accelerator Research Program. And an MU College of Engineering computer scientist was one of the select few chosen from a worldwide pool to work on these next-generation platforms.
Bill Harrison, an associate professor in the Electrical Engineering and Computer Science Department, will have the opportunity to do groundbreaking research on Intel’s most advanced hardware, beginning in February. He’ll be working on a platform that combines an Intel microprocessor and an Altera Arria-10 Field Programmable Gate Array, or FPGA. An FPGA is computer hardware designed to be configurable even after manufacturing, and when combined with more standard hardware such as a microprocessor, it forms what’s called a heterogeneous hardware platform — two different types of hardware melded into one platform.
“People have been trying to come up with ways to program applications on these architectures, and it’s very challenging,” Harrison said.
Intel’s goal is to spark new research in areas including programming tools, operating systems and applications for accelerator-based computing systems. These architectures have received a lot of research interest, but the difficulty in programming them has limited their impact to-date. The goal is to maximize impact in order to provide a low-resource, high-performing system for future use.
“There are lots of potential applications,” Harrison said. “We’re going into more and more embedded kinds of systems — the buzzwords are ‘cyber-physical systems’ and ‘the Internet of Things’ — so they need ways of producing low resource-intensive applications. … People have been working on reconfiguring these systems for a long time because they offer so many benefits in terms of power and resources.”
Harrison said his proposal was based largely on the premise of building a method to allow the two parts of the architecture — the microprocessor and FPGA — to better understand each other.
“What we’re offering — what we’re bringing to the mix — is a programming languages-based approach where you have a single language. Part will execute on the microprocessor, and part of the program will be turned into hardware,” he said. “Then the two parts of that program talk to each other.”
Researchers will have access to the hardware through a network connection, and details of the projects and their parameters will be discussed at an upcoming meeting of the researchers in February at Intel’s office in Portland, Ore.