US National Academy of Sciences report, future of computing

[Phil Bunch]

I thought this new report might be of interest to those interested in computer trends, etc.

http://www.nap.edu/catalog.php?record_id=12980

"The end of dramatic exponential growth in single-processor performance marks the end of the dominance of the single microprocessor in computing. The era of sequential computing must give way to a new era in which parallelism is at the forefront. ..."

"There is no guarantee that we can make parallel computing as common and easy to use as yesterday's sequential single-processor computer systems, but unless we aggressively pursue efforts suggested by the recommendations in this book, it will be "game over" for growth in computing performance. If parallel programming and related software efforts fail to become widespread, the development of exciting new applications that drive the computer industry will stall; if such innovation stalls, many other parts of the economy will follow suit."

[Will]

I've read that this contributed to the demise of MSFS... they released FSX, which was optimized for single core machines, at the exact instant that people migrated to dual- and quad-core computers, meaning that Microsoft was stuck with a product that was going to be behind the technological capabilities of users' hardware for much longer than anticipated.  I wonder if that's true? (Never had FSX myself, can't comment on it from first hand experience.)

[Richard McDonald Woods]

Phil,
I believe that the end of significant improvements in client uni-processors is inevitable, and is not so important. Attempts at introducing programming languages for multi-processor machines is still some way off, and perhaps will never be used by the average Joe.

The need for improvements in the performance of uni-processing languages are not great because most of what a person needs of a client machine (PC, mobile device, etc.) is already available. As the place of processing is fast moving to servers (the cloud), there is a lessening need for local computation.

The complexities of producing easy-to-use multi-threaded programming languages means that their usability and availabilkity is still many years away. I certainly find it extremely difficuilt to envision how I would design the logic for a multi-processor machine, let alone set about using such a language. I, like most people, am firmly stuck in the 'one thing at a time' world.

[Hardy Heinlin]

Java!

[Jeroen Hoppenbrouwers]

The trick is that tasks (broad tasks, such as a complete flight sim or extensive multimedia application) which have a natural parallelism can easily be scattered over multiple threads or processes. This has been this way for decades. Only if you have a task which is difficult to break up in relatively independent, parallel subtasks, you encounter the typical hard problems that computer science was/is fond of and pesters students with ad infinitum.

Personally I don't think "the end of single-thread CPUs as we know it" has much relevance, and I suspect the report to be an attempt to secure research funding to continue the "eternally interesting and promising" research branches that have driven CS for as long as it exists.

Jeroen
ex-research industry fugitive

[Phil Bunch]

I found this wikipedia article on parallel computing to be helpful as I reflected on the development of parallel computers:

https://secure.wikimedia.org/wikipedia/en/wiki/Parallel_computing
-------------------------------------------------------
Yet another chance to reminisce...

About 100 lifetimes ago (in the 1980s), my large corporation R&D workgroup acquired a Floating Point Systems FPS-164 array processor.  It was attached to our DEC VAX 8650 "supermini" computer, and was accessed through it. My job was to help manage the whole system, as well as to make use of the system on behalf of my company's medical radiography R&D.   Remarkably, I and two other scientists, plus one technician/programmer, managed and used the system, defying the company's mainframe enthusiasts who wanted us to stay captives of their IT-based empire.

This was back in the days when 1 MIPS (million instructions per second) was both expensive and usually both hard to access and use.

The FPS-164 array processor came with a lot of parallelized software for things like matrix algebra and other common scientific/engineering computational needs.  Although the two theoretical chemists who were using most of the CPU-hours of the system for molecular orbital calculations soon adapted to the beast, I found it stressful to use for a given task and only used it if I really needed the speed badly.

There is a brief summary of the FPS company here, and it's of some historical interest to see who merged with whom, etc:

http://publishing.cdlib.org/ucpressebooks/view?docId=ft0f59n73z;chunk.id=d0e15843;doc.view=print

This 1980s computer system occupied a 30x30 foot room and required a large dedicated chilled-water air conditioner as well as 25,000 watts of electrical power.   Our 16GB of hard drive space, phenomenal for 3 users at the time, occupied several refrigerator-sized cabinets.  I believe the whole system cost about $1M in 1982 dollars.   My distant memory is that it cost about $300,000 per year just to renew the hardware and software maintenance contracts.

Now, my iPod Touch would run rings around the whole 1980s system, of course.  It's a bit unreal to me to carry 64 GB of solid state storage in my shirt pocket.  "Future Shock" was a prophetic book, and sometimes describes a very real psychological effect, too.

Those were the good old days...and in spite of Moore's Law and computer speedups, there still are plenty of people who want even more computer power, mostly for specialized R&D projects.  The nuclear weapons labs in particular have traditionally been major consumers of the fastest computers and supercomputers.
-------------------------------------------
On a somewhat related note, I recently read that the leading character in "The Soul of a New Machine" died:

http://news.cnet.com/8301-13556_3-20064511-61.html

(Excerpts below)

"Tracy Kidder's 1982 Pulitzer-Prize-winning work of book-length reportage, "The Soul of a New Machine," is perhaps the best narrative of a technology-development project ever written. It's up there with "The Mythical Man Month" and "Showstopper." And the hero of that book was Tom West. The pages open with Tom at the helm of a sailboat in a storm. "In the glow of the running lights, most of the crew looked like refugees, huddled, wearing blank faces. Among them, Tom West appeared as a thin figure under a watch cap, in nearly constant motion."

"Tom West, 71, died Thursday at his home in Westport, Mass., yet another mark of the passing of the minicomputer era, so important to the way the computer era has evolved--and, indeed, how Massachusetts evolved as part of the computer industry."

-----------------
This book and the Data General competitor for my VAX supermini computer became available at the time of the above segment of my work life..  The book captured so much of what it's like to work on a fast-paced high-tech project...and I personally believe that spirit is common to all high-tech R&D efforts.  I believe I recall correctly that Jeroen H also enjoyed and was influenced by this book.

[Jeroen Hoppenbrouwers]

Yup. Sure did. Still trying to live up to the spirit, and finally I think I have found a place to work that will allow me to "sign up" to something like it.

[martin]

+1

Martin