Stanford, Chipmakers Team On Simplifying Multi-Core Apps

Stanford University announced this week it is working with the biggest names in computer chips to launch a research lab designed to make parallel programming accessible to the average programmer.

Advanced Micro Devices, IBM, Intel, Hewlett-Packard, Nvidia, and Sun Microsystems have all signed up to participate in the Stanford Pervasive Parallelism Lab. The lab, which officially opens May 2, is a combination of Stanford researchers and chip companies -- all of which are experts in applications, languages, systems software, and computer architecture.

The center is a three-year, $6 million project that will basically focus on what game programmers already know: make several threads work for you at the same time to handle complex processes. In a game environment, processors (CPU, chipset, and graphics) juggle artificial intelligence, graphics rendering, and game physics and implement their algorithms in accessible "domain-specific" languages.

With the commodity server market moving quickly toward increasingly more powerful multicore processors, new tools are needed to help programmers develop business software that can take full advantage of the platforms. What is needed, say the companies, are parallel algorithms, development environments, and runtime systems that scale to thousands of hardware threads.

"Parallel programming is perhaps the largest problem in computer science today and is the major obstacle to the continued scaling of computing performance that has fueled the computing industry, and several related industries, for the last 40 years," said Bill Dally, chairman of the computer science department at Stanford.

The lab's research will be directed by Kunle Olukotun, a professor of electrical engineering and computer science who has worked for more than a decade on multicore computer architecture, in which many processors inhabit the same silicon chip.

To enable the research, the team's hardware experts will develop a test bed called Farm (flexible architecture research machine). The system, which Olukotun said will be ready by the end of the summer, is expected to blend the best of reprogrammable chips with conventional processors.

Stanford is not alone in its parallel programming endeavors. The announcement comes on the heels of news out of the University of California at Berkeley and the University of Illinois at Urbana-Champaign, who are working with multimillion-dollar grants from Microsoft and Intel to tackle the same problem.

"Clearly this problem is big enough and important enough that we need more people looking at it," Olukotun said. "By having more and different approaches to the problem, we're more likely to find a solution."