Technology Review: Blogs: TR Editors' blog: Intel, Microsoft Push Parallel Computing

About Us

Insights, opinions, and our editors' analysis of the latest in emerging technologies.

Add RSS Feed

Intel, Microsoft Push Parallel Computing

The companies are providing millions of dollars to bolster practical parallel-computing research.

Wednesday, March 19, 2008

By Kate Greene

Yesterday, in a teleconference, Intel and Microsoft announced that they will start to fund serious efforts to make parallel computing mainstream. The companies are injecting $20 million into computer-science research at the University of California, Berkeley, and at the University of Illinois, Urbana-Champaign, dubbing them the Universal Parallel Computing Research Centers (UPCRC).

For decades, parallel computing--the tricky job of dividing programming tasks among a number of processors--has been the province of academics who write programs for supercomputers. But with the advent of multicore chips--chips with more than one processing center--personal computers are on track to become supercomputers as well.

The main problem outlined at yesterday's teleconference is that there is still no consensus on the best way to easily program general-purpose chips, like the ones in consumer computers, which have a large number of cores. Historically, programmers have been able to wait for the next generation of chips to come out to make their programs run faster. Now that the next generation of chips includes multiple cores, it's not as obvious how to divvy up the tasks among the cores to achieve those same gains.

"Programmers have no choice: if they want fast programs, they're going to have to write parallel programs," says Dave Patterson, the director of the UPCRC and a computer-science professor at Berkeley.

Although no details were given on specific research projects, Tony Hey of Microsoft Research mentioned some areas of interest, which include developing parallel code libraries with chunks of code that are ready to use, testing different memory approaches, and exploring different types of parallel languages.

Below is a roundup of the most interesting quotes from the teleconference.

"The shift in hardware technology from a single core to multicore will have a profound effect in the way we do programming in the future. We're really in the midst of a revolution in the computing industry." --Tony Hey, Microsoft Research

"Parallelism is coming not just to high-end systems, but across a very broad slice of computer science, and we expect to permeate every corner of computer science going forward." --Andrew Chien, director of Intel Research

"The technology revolution that Andrew [Chien] outlined means that every computer will be a parallel computer. Every problem must be a parallel problem ... We want to democratize parallelism." --Marc Snir, codirector of the UPCRC and a professor of computer science at the University of Illinois

"My perspective here is, there's not that much that universities haven't invented [in parallel computing], but there was no strong pull from industry because parallelism wasn't pervasive; it was a small subset of the industry ... Now that the pipeline is unclogged, now that there's strong interest from Microsoft and Intel ... work that's been going on for many years will become much more fruitful." --Marc Snir

Tags: microsoft, Intel, multicore, parallel computing, supercomputing

Comments

	Nightmare on Core Street Mapou on 03/21/2008 at 2:35 PM
Posts: 18 Avg Rating:	One day soon, the computer industry will realize that, 150 years after Charles Babbage came up with his idea of a general purpose sequential computer, it is time to move on and change to a new computing model. The industry will be dragged kicking and screaming into the 21st century. For over 20 years, researchers in parallel and high-performance computing have tried to come up with an easy way to use threads for parallel programming. They have failed and they have failed miserably. Amazingly, they are still continuing to pursue the multithreading approach. None other than Dan Reed, Director of scalable and multicore computing at Microsoft Research, believes that multithreading over time will become part of the skill set of every professional software developer (source: cio.com). What is wrong with this picture? To find out why multithreading is not part of the future of parallel programming, read, Nightmare on Core Street Rate this comment: (Reply)

	Re: Nightmare on Core Street darkstar57 on 03/22/2008 at 11:31 AM
Posts: 2	the brain is a multicore system, so the the way Intel and Microsoft implement it is quite significant for the emulation gang. Rate this comment: (Reply)

	Re: Nightmare on Core Street Mapou on 03/22/2008 at 12:32 PM
Posts: 18 Avg Rating:	Yes, the brain is the perfect analogy. It is amazing that the solution to the parallel computing problem has been pretty much in our faces all along. We have been emulating deterministic, reactive parallel systems like neural networks and cellular automata for decades. And we can do it without using multithreading, mind you. To solve the parallel programming problem, we need to start thinking of operations (additions, subtraction, etc...) as neurons (or cells) in a neural network. But emulating such a low-level network in software would be too slow. This is why we should let the hardware (i.e., the multicore processor) do it. Software should emulate hardware. This, in a nutshell, is what the COSA software model is all about. It's not rocket science but it will require a reinvention of the computer. Check it out. Rate this comment: (Reply)

	Re: Nightmare on Core Street jjwalker on 03/26/2008 at 9:02 AM
Posts: 1 Avg Rating:	Moore's law at work; but for a different reason. Yes, computational capacity is increasing but will it mimic the human mind and reach the holy grail of computer engineering efforts- "Singularity" which will occur when the differences between human and machine intelligence will merge. A milestone has been reached recently when it was determined that the best computers are now smarter than the dumbest humans. WIRED magazine has made in-roads with their efforts to report this phenomenon, but to the disappointment of the scientific community it has not been widely reported. This milestone has been termed "Peculiarity" since it has been a perplexing task to equate human intelligence with computers at the lowest level versus the highest level. Hopefully progress made by Intel will provide more insight into why humans are not as intelligent as they think they is. Rate this comment: (Reply)

	Parallel for the mainstream wmyoungs on 07/10/2008 at 3:06 PM
Posts: 1	Professional software developers are confronted with some difficult choices as we enter the era of multi-core. They realize that in order for their applications to remain competitive they are going to have to adapt them to the new systems. They read in the trade press about the difficulties of parallel programming and they worry about development and maintenance cost, time to market, and the ease with which their applications can be adapted to widely different configurations and new generations of systems. Intel® Concurrent Collections attempts to address some of these concerns by providing a programming model that lets the application domain expert focus on what they know best, a precise description of their program logic, while leaving parallel concerns and system adaptation to the runtime system. The programming model supports all styles of parallelism so there is no need to re-write or re-compile application in order to change style of parallelism (task, data, or pipeline parallelism) or to target new system configurations. To learn more about Intel(r) Concurrent Collections go to: http://softwarecommunity.intel.com/articles/eng/3862.htm This is FREE experimental software. We would love your feedback. Rate this comment: (Reply)