|Exhibition and Industrial Session: Silicon Graphics International|
Drs. R.J. van der Pas, |
Silicon Graphics International, Belgium
In this presentation we will discuss the architecture of the MIPS R8000 microprocessor and the Silicon Graphics PowerChallenge system. The latter is a multiprocessor system that is ideally suited for large scale floating point intensive applications. Multiple PowerChallenge systems can be connected via a HiPPI switch to build the PowerChallenge Array, with up to 128 processors.
As software plays a crucial role in every parallel system, we will pay considerable attention to this component as well. Not only automatically parallelizing compilers are available, but also a suite of integrated program development tools and system management tools are offered. Together with very efficient implementations of PVM and MPI, this provides for an environment that allows both the novice and experienced user in parallel programming to be productive from the start.
We will conclude our talk with a discussion on the programming model which supports two popular parallel computing paradigms: shared memory and message passing. A combination of these two on the PowerChallenge (Array) provides for a parallel environment, with possibly non-uniform communication characteristics between tasks.
Silicon Graphics has always been a technical leader.Since our founding over a decade ago, we have applied our visual computing expertise to serve the needs of a highly technical marketplace.Traditionally, customers have turned to Silicon Graphics products to solve visual computing problems in technical, scientific and entertainment applications. None of that has changed. What is different, however, is the sheer number of people now using Silicon Graphics technology. While we remain entrenched as the visual computing leader with our historical customer base, many more people are now beginning to appreciate the benefits our products offer.
Since 1985 Ruud van der Pas has been involved in various aspects of high performance computing. His major focus is on porting and tuning programs for supercomputers. As such he has experiences on a broad variety of architectures, ranging from vector systems to large scale parallel systems. At Silicon Graphics he works in the Multi-Country Area Supercomputing Team.