Random number generation and simulation on parallel computers

Richard P. Brent, University of Oxford

Pseudo-random numbers are often required for simulations performed on parallel computers. The requirements for parallel random number generators are more stringent than those for sequential random number generators. As well as passing the usual sequential tests on each processor, a parallel random number generator must give different, independent sequences on each processor.

We consider the requirements for a good parallel random number generator, and give examples of (probably) good and (definitely) bad uniform generators. We also describe a new class of generators for the normal distribution (based on a recent proposal by Wallace). These generators can give very fast vector or parallel implementations. Implementations of both uniform and normal generators on the vector/parallel Fujitsu VPP300 will be discussed.





Performance Portability, Heterogeneity and Integration

Marco Vanneschi, University of Pisa

I will report on the experience of our national project on the realization, utilization and performance evaluation of the SkIE (Skeleton-based Integrated Environment) programming environment. The goal of SkIE is to favour performance portability across different platforms, to manage the heterogeneity of architectural (standard and proprietary) components and to integrate the variety of standard software tools into a uniform environment in which compile-time optimizations are done according to the skeleton model. A first industrial version of SkIE is now operating on PQE2000 MIMD+SIMD prototypes and on workstation clusters. The talk will report this experience with reference to scientific and industrial applications, and will discuss the issues in the realization and utilization of advanced programming environments for parallel machines.



Quantum Cryptography

Paul Townsend
Networks and Technology Centre, BT Laboratories, Martlesham Heath, Ipswich

A defining feature of today's digital information systems is that their operation can be fully described by the laws of classical physics. However, there has recently been a growing interest in a new class of information systems that are fundamentally quantum-mechanical in nature. These quantum information systems offer a range of new functionalities that cannot be achieved with any conventional classical system. These new features occur because the properties of quantum and classical information are fundamentally different. For example, it is not possible, even in principle, to make a perfect copy of quantum information and, furthermore, any attempt at copying will unavoidably lead to a corruption of the information in question. It is these properties that are directly exploited in quantum cryptography. The technique enables the secrecy of information transmitted over public networks to be tested in a fundamental way, since an eavesdropper will inevitably introduce readily detectable errors in the transmission. Quantum cryptography offers the intriguing prospect of certifiable levels of security that are guaranteed by fundamental physical laws. In this talk I will discuss the background to the technique and describe recent progress at BT Laboratories in the development of practical optical fibre-based quantum cryptography systems.






Very Distributed Media Stories: Presence, Time, Imagination
Glorianna Davenport, MIT Media Laboratory

Throughout the ages, the story-form has evolved to meet social and economic needs. In turn, this transformation -- manifested in forms such as painting, song, theater, the novel, movies, and television -- has driven formal and technological innovation. Today, story is perched on the cusp of the digital revolution. Electronic sensing devices, computer animations, electronic collage, computational story engines, and communication networks promise to immerse the participant user in a fantasy realm which is far more demanding than has previously been imagined.
How will this new medium be articulated? In our participation as "users," we can no longer be doomed to type, point, and click our way through the digital universe, regardless of the task at hand. Complex story environments of the future must actively challenge our ability to act and transform our emotional state. Moving and performing tasks within them must bear some resemblance to the body-motion and sensory experience of real-world encounters; but, the higher-order use of metaphor and synthetic emphasis is also necessary if the experience is to be perceived as story-like. So far, video games are the only form of electronic entertainment which have successfully shaken the general public's notion of television viewing; however, video games rarely make us gasp with fear or tug at our heart-strings.

Over the past twenty-five years, storymakers and engineers have grappled with the implementation as well as with the underlying architecture for interaction. Recent breakthroughs in sensor technology, computer vision,
immersive display devices, and the construction of semi-autonomous software "agents" are rapidly increasing the reactive potential of information media. The next generation of digital media will feature information-rich, dynamically adaptive, highly distributed environments which seamlessly merge the real and virtual worlds. As bits themselves become graspable and manipulable, the separation between maker and consumer shrinks dramatically, while the connection between consumer and story becomes more tangible, personalized, and intimate. In the future, the story designer and content producer will focus on creating environments and presentation systems in which adaptable, mutable, and computationally inexpensive manipulations are applied to a range of content structures.



Computing with Java and CORBA on the Pragmatic Object Web
Geoffrey Fox, University of Syracuse

We discuss computing using "Web technologies" which can link millions of servers, a factor of hundred more clients or even billions of highly mobile smart cards. We base the discussion on the growing power and capability of commodity computing and communication technologies largely driven by commercial distributed information systems. These systems are built from CORBA, Microsoft's COM, Javabeans, and less sophisticated web server and networked approaches. One can abstract these to a three-tier model with largely independent clients connected to a distributed network of servers. The latter host various services including object and relational databases and of course parallel and sequential computing. This network of servers encompassing disparate object and web standards is termed the "Pragmatic Object Web" and we show how to use these concepts even when there is no one universal standard. In particular we describe JWORB -- a Java server built at NPAC supporting CORBA and Web protocols simultaneously. These ideas raise a number of issues -- What is the role of parallel computing? Is the opportunity linking of computers or linking of people using computers? What is the role of academic computer science in a world largely driven by industry -- should research and/or curricula adapt or just wait out a passing fad? What are the architecture and performance issues? What are the appropriate language and operating systems? What are appropriate message passing standards, eg is MPI relevant or built around yesterday's vision?