With all programming about to become parallel programming, the industry’s famous Wintel twins, Microsoft and Intel, said this afternoon that they will be putting $20 million into university research into the art over the next five years.
Their objective is to make parallel programming easier and mainstream, the stuff of laptops, desktops and mobile devices, ubiquitous from the “laptop to the petaflop.”
They are funneling the money into two new Universal Parallel Computing Research Centers (UPCRCs) – one at Berkeley, the other at the University of Illinois at Urbana-Champaign (UIUC) – that will seek breakthroughs in parallel programming applications, architecture and operating systems software.
An additional $8 million will come from UIUC, and Berkeley has applied for $7 million in funds from a state-supported program that matches industry grants.
Needless to say, this is the first joint industry and university research alliance of this magnitude in the United States focused on mainstream parallel computing, historically the low man on the university research totem pole, and made necessary by the prospect of the many-core processor – like Intel’s 80-core research processor – soon becoming the basis of all computing.
These many-core chips, and eventually chips will have hundreds of cores, can handle multiple instructions and tasks simultaneously, and therein lays the problem: how to program the darn things.
The object of the research into predominantly single-socket parallelism will be to co-design hardware and software in parallel and the most provocative statement made during the hour-long press conference unveiling the initiative today was made by Professor Marc Snir, who will run the center at Urbana-Champaign. He said this investment means “Microsoft and Intel are willing to consider fundamental changes.”
Both Microsoft and Intel are looking for client-focused mass-market applications like cell phones that recognize and tell you who that guy is walking towards you at a cocktail party wanting to shake your hand or the Jewish mother PDA that checks your vital signs and tells you to take your medicine or go to the doctor.
The research will complement and extend existing parallel computing programs at UC Berkeley, UIUC, Microsoft and Intel. The centers’ agenda aligns closely with both Intel’s Tera-scale Computing Research Program and Microsoft’s Technical Computing Initiative.
Microsoft and Intel will have royalty-free access to the intellectual property the centers develop and a right to seek exclusive licenses but the object of the game is to open source any software breakthroughs under a BSD license and let the technology community develop them further.
Apparently discussions at least at Berkeley, which has already left its imprint on Unix, databases and RISC chips, go back three years and Intel and Microsoft evaluated 25 top-tier institutions in the field of parallel computing research before lighting on Berkeley and UIUC.
David Patterson, a professor of computer science and an expert in computer architecture, will run the center at Berkeley, which will include 14 members of the Berkeley faculty as well as 50 doctoral students and post-doctoral researchers.
Patterson said Berkeley’s charter is application-driven, impacting domain experts, with what he called “Ninja programmers” getting the most out of it.
The center at Urbana-Champaign under Snir, a computer science professor, and Wen-Mei Hwu, a professor of electrical and computer engineering, will include another 20 faculty members and 26 graduate students and researchers.
Snir expects Urbana’s work to impact domain experts and the mainstream programmer.
Patterson believes the centers will impact computing for the next 30 or 40 years and “recast the foundations of information technology.”
A “revolution” is afoot that will bring on new platform architectures, operating system architectures, programming methods and tools, and application models, they said, and “affect the entire industry, from consumers to hardware manufacturers and from the entire software development infrastructure to application developers who rely upon it.”
Parallelism is the next great hope of both energy efficiency and performance. “If you want to write a fast program, you are going have to write a parallel program,” Patterson said. No more of this business of developers sitting back on their collective keister waiting for the next-generation chip to speed up their programs.
According to their press release, “We think these new applications will have the ability to efficiently and robustly sense and act in our everyday world with new capabilities: rich digital media and visual interfaces, powerful statistical analyses and search, and mobile applications.
Ultimately, these sensing and human interface capabilities will bridge the physical world with the virtual.”