[http://www.openmp.org/ "An industry standard for shared-memory
programming" has much more. News hook was Supercomputing'97. ]
As a result scalable software for such systems will exist, at some
level, only in a message passing model. Message passing is the native
model for these architectures, and higher level models can only be
built on top of message passing.
Unfortunately, this situation has existed long enough that there is
now an implicit assumption in the high performance computing world
that the only way to achieve scalability in parallel software is with
a message passing programming model. This is not necessarily true.
There is now an emerging class of multiprocessor architecture with
scalable hardware support for cache coherence. These are generally
referred to as Scalable Shared Memory Multiprocessor, or SSMP,
architectures [1]. For SSMP systems the native programming model is
shared memory, and message passing is built on top of the shared
memory model. On such systems, software scalability is straightforward
to achieve with a shared memory programming model.
[I don't buy it. Shared-memory is a sham in the long-term. Truly
distributed -- *decentralized* distributed computing -- will require
the message-passing model in spades. Shmem is not how antcolonies
work. Sure, today, this is for supercomputing only, but in the
long-term cheap, slow, and ubiqutous trumps expensive, fast, and
centralized. Good PR though... -RK]
========
October 28, 1997
A New Standard to Govern PCs With Multiple Chips
By JOHN MARKOFF
SAN FRANCISCO -- A group of developers of the fastest
hardware and software for scientific and engineering
computing will announce a new technical standard,
perhaps as early as Tuesday, that is intended to make it
easier for complex programs to operate on different
types of computers.
The new standard, Open MP, is designed for computers
that use either the Unix or Windows NT software
operating systems and employ multiple microprocessors
for parallel computing.
In parallel computing, complex tasks are broken into
simpler pieces that multiple microprocessing chips can
tackle simultaneously. Such capabilities were once the
sole province of supercomputers costing millions of
dollars but more recently are being designed into
top-of-the-market work stations and personal computers
that sell for thousands of dollars.
The new standard is designed for a type of parallel
computer that permits all processors to share a single
pool of computer memory. Open MP will permit programmers
to write a single version of their software rather than
having to tailor a version for each different
manufacturer's machine.
One effect of the Open MP standard will be to increase
the shift of complex scientific and engineering software
development from the supercomputer world to high-end
desktop work stations.
"Developers are finding that desktop parallel systems
are a breakthrough in allowing real-time use of complex
software on low-cost platforms," said David Kuck, a
computer scientist and an expert in parallel computing.
The computer makers that have agreed to support Open MP
are Compaq Computer, Digital Equipment, Intel, IBM and
Silicon Graphics.
The new standard, developed by researchers at Silicon
Graphics and Kuck & Associates, a consulting firm in
Illinois, is significant because it should make possible
the rapid creation of a broader market for the type of
commercial computer programs that previously had to be
custom designed for each machine.
"The software developers have been wanting this for the
last 15 years," said Steven Wallach, a computer industry
pioneer and consultant.
The software breakthrough comes as the computing
industry awaits the arrival next year of Intel's Merced
microprocessor family, designed for high-performance
computers. In recent days and weeks, computer makers
including Digital Equipment, Silicon Graphics and Sun
Microsystems have said they intend to offer Merced-based
computers or to adapt their machines' software to work
with Merced-based computers.
Such increasingly powerful hardware and software can be
expected to create new export-control problems for the
government. Congress moved Monday to re-establish export
controls over high-performance computers shipped to
countries that have nuclear development programs.
The legislation, which was reported out of an Armed
Services conference committee Monday, would force U.S.
computer makers to notify the government if they plan to
sell machines with supercomputer performance to any of
these countries.
But experts say the new standard will make it
increasingly simple for overseas entities to build
machines with supercomputer performance from
off-the-shelf components.
[Markoff investigated a great, lengthy piece yesterday on how the
Russians have been subverting supercomputer export rules, possibly
with help from rogue IBM reps - RK]