SGI/CRI Software Report
Visual Supercomputing for High-Performance Virtual Environments
The term "visual supercomputing" suggests, from the users' perspective, an ideal for realizing applications: the integration of computing and displays through shared memory. Silicon Graphics/Cray Research's new Origin2000 technology potentially simplifies the real-time visualization of computationally intensive and large memory simulations by extending the distributed shared memory (DSM) interprocessor connectivity bandwidth to the graphics. The graphics engines themselves are multi-gigaflop computers with significant internal raster managers and memories. As supercomputing capabilities are ramped up to multiple teraflops during the next 5 years, we will make similarly aggressive advances in visualization, multimedia and virtual reality for the computational science and engineering community.
HyperTrace: From Simulation to Emulation of the Real World
Computational fluid dynamics (CFD) has been a major application since the early days of high-performance computing. Yet even to this day, CFD remains a highly specialized application in which only few people can benefit directly from the insight it has to offer. The HyperTrace project uses well-established visualization and post-processing techniques, but only through the use of HPC environments we bring this technique to the scale of physical emulation. We create a computational experimental laboratory such that now non-simulation experts can unlock the benefits of high-performance computing and visualization. We draw examples from such diverse manufacturing areas as the automotive, aerospace, pharmaceutical and process industries.
Simulation of Unsteady 3-D Viscous Flow
Robert L. Meakin
The Cray supercomputers have enabled researchers at the Numerical Aerospace Simulation Facility at NASA Ames Research Center to develop computational tools suitable for solving problems with unsteady three-dimensional viscous flow for which accurate methods of prediction are frequently required. Such applications are usually complicated geometrically, may involve relative motion between component parts, and exist in virtually all engineering disciplines. Experimental methods of analysis are often not possible due to excessive cost, model limitations, human safety factors, and time constraints associated with a commercially competitive environment. Unsteady viscous flowfields involving vortical wakes, interference effects, moving shocks, and body motion demand the most advanced computational means available.
The only viable high-order method of prediction for these problems, the Chimera overset grid approach, is discussed. This approach involves the decomposition of problem geometry into a number of geometrically simple overlapping grid components. Components of a particular configuration can be altered or changed completely, without affecting the rest of the grid system. Videos of actual flight tests and Navier-Stokes simulations, together with still images, are used to demonstrate the utility of the approach.
A review of the current status of the Chimera-style overset grid method as it applies to unsteady three-dimensional viscous flow is given. The state of maturity of the various pieces of support software required to use the approach, including grid, flow-solver, and post-process analysis-related software issues are discussed. A variety of recent applications of the method, carried out on Cray supercomputer-based systems, are presented. Current limitations of the approach are identified and used to suggest needs for future developmental efforts.
Challenging Applications on a CRAY Supercomputer Complex: User Support, Optimization, Experience
Wolfgang E. Nagel
The Research Centre Juelich is running one of the most powerful Cray computer systems worldwide (T90/12, T3E/512, M94). As we have to provide resources to the German national supercomputer center HLRZ, users from all over Germany have accounts on this heterogeneous supercomputer complex.
The talk will describe the current status of the systems and the procedures established to make the system highly available for the user community. This will include educational aspects, user support activities, as well as tool development to provide the layer for a successful optimization process. Finally, examples will show the benefit from all this work.
Table of Contents | Author Index | CUG Home Page | Home (Title Page)