Cray Supercomputer Research Articles

Efficient direct solvers for large-scale computational fluid dynamics problems

This paper presents some new results concerning the direct solution of large systems of equations resulting from the fully implicit, coupled solution of the Euler and Navier-Stokes equations in three dimensions. Two methods for such direct solution have been investigated: the first is based on the application of sparse matrix algorithms and the second on the implementation of a Gauss elimination algorithm optimized for vector-parallel processing on Cray supercomputers. Results indicate that sparse matrix algorithms consistently deteriorate for rectangular, i.e. N × N × M grids in three dimensions, as M gets larger than N. On the other hand, Gauss elimination seems to be ideally suited for processing on today's supercomputers, achieving execution rates up to 83% of the Cray-YMP peak efficiency.

Supercomputer modeling of delphinid sonar beam formation

It has been established that some dolphins possess well‐developed acoustic orientation (echolocation) and information gathering abilities, though substantially less is known about the system of sound generation and beam formation. Dolphins use a narrowly focused sound beam that emanates from the forehead and rostrum during echolocation. The primary objectives of this study were to simulate the effects of modeled tissues on beam formation, and to test the viability of various hypothetical sound source locations. 2‐D simulations of sound propagation using parasagittal outlines from reconstructions of CT scans through the forehead tissues of two delphinids were conducted. Finite difference wave propagation programs were run on a Cray supercomputer. Preliminary simulations suggest that beam formation results primarily from reflection off of the skull and air sacs. These results do not depend strongly upon the precise values of velocity and density assumed for the bone. Beam angles closely approximate those measured by experimental methods for a source located in a region of the model corresponding to the monkey lip/dorsal bursae complex (MLDB), and not elsewhere. These results suggest that: (1) the skull and air sacs play the central role in beam formation; (2) the geometry of reflective tissue is more important than the exact acoustical properties assumed; and (3) experimentally observed beam patterns are best reproduced in the simulations when the sound source is placed in the region of the dolphin's head known as the MLDB (Cranford, 1988).

FFTs in external or hierarchical memory

Conventional algorithms for computing large one-dimensional fast Fourier transforms (FFTs), even those algorithms recently developed for vector and parallel computers, are largely unsuitable for systems with external or hierarchical memory. The principal reason for this is the fact that most FFT algorithms require at least m complete passes through the data set to compute a 2 m -point FFT. This paper describes some advanced techniques for computing an ordered FFT on a computer with external or hierarchical memory. These algorithms (1) require as few as two passes through the external data set, (2) employ strictly unit stride, long vector transfers between main memory and external storage, (3) require only a modest amount of scratch space in main memory, and (4) are well suited for vector and parallel computation. Performance figures are included for implementations of some of these algorithms on Cray supercomputers. Of interest is the fact that a main memory version outperforms the current Cray library FFT routines on the CRAY-2, the CRAY X-MP, and the CRAY Y-MP systems. Using all eight processors on the CRAY Y-MP, this main memory routine runs at nearly two gigaflops.

Network traffic flow theory: Microscopic simulation experiments on supercomputers

Microscopic simulation experiments are conducted to investigate traffic behavior in urban networks and support the development of network-level traffic flow relations. Previous use of microscopic traffic simulation has been limited by computational resources to small networks. Supercomputers largely alleviate such limitations and allow more elaborate and realistic simulations of traffic in large networks. This paper provides computational experience in simulating large-scale urban traffic networks on a CRAY supercomputer. In addition to the computational results, the experiments address the effect of intersection traffic control on the two-fluid characterization of vehicular flow in congested networks. Evidence is provided from larger networks that supports conclusions reached in earlier work using smaller test networks.

Linear Programming on Cray Supercomputers

In this note we present computational experience with a linear programming code on two Cray super-computers (Cray-1S, Cray X-MP/48). Results are given for problems involving up to 640,000 variables (1600 constraints) and 10,000 constraints (99,969 variables).

Linear Programming on Cray Supercomputers

In this note we present computational experience with a linear programming code on two Cray super-computers (Cray-1S, Cray X-MP/48). Results are given for problems involving up to 640,000 variables (1600 constraints) and 10,000 constraints (99,969 variables).

Linear Programming on Cray Supercomputers

Linear Programming on Cray Supercomputers

Visualization tools at Lawrence Livermore National Laboratory

An examination is made of three projects underway at LLNL. The Magic (Machine Graphics in Color) Project is developing techniques to move data from a Cray supercomputer and effectively display it on high-resolution color terminal in the scientist's office. In a follow-on to this project, laser disk technology provides a real-time animation capability. The Graphics Workstation Project has examined visualization applications of workstations with graphics accelerators and is currently extending this work to 'graphics minisupercomputers'. The Advanced Visualization Research Project uses high-end graphics devices to perform algorithm development, especially in the implementation of volumetric methods. >

The use of supercomputers for the variational calculation of ro-vibrationally excited states of floppy molecules

The advent of supercomputers has led to great advances in electronic structure calculations and to the ab initio calculation of molecular spectra. Recent theoretical developments have allowed us to develop a two-step variational algorithm for the calculation of rotationally highly excited states of floppy molecules. This algorithm allows highly accurate nuclear motion calculations to be performed on low-lying ro-vibrational states and greatly extends the range of states that can practicably be considered. The algorithm has been adapted to run efficiently on the Cray supercomputers. Analysis of the timings suggest that construction of the secular matrix is highly vectorised and that the special structure of secular matrix can be used to give rapid diagonalisation. The limiting factor on these calculations is the available fast storage, but analysis suggests that this bottleneck could be removed by use of a Solid State Device (SSD). Sample results are given for calculations involving a range of rotational excitation. An adaptation of the algorithm to a loop of parallel processors is also suggested.

Cray Supercomputer Axed, Superstar Departs

Cray Supercomputer Axed, Superstar Departs

The impact of supercomputers

The first and second glances at Cray Research’s manufacturing plant (in the distinctly un‐Silicon Valley milieu of Chippewa Falls, Wisconsin) do not trigger images of the high technology frontier. The long, low, aesthetically neutral building looks very much like the shoe factory it, until quite recently, was; inside platoons of women, working shoulder to shoulder in a rain forest of stringy material draped from the ceiling, evoke memories of the infamous sweat shops of the early years of the century. These knee‐jerk mental pictures are stunningly inappropriate. The products of this obsolete‐looking assembly line are Cray supercomputers, the biggest sellers in that increasingly mindwarping arena of computer science.

Digital scene simulationsm: The synergy of computer technology and human creativity

Digital Scene Simulation is Digital Productions' philosophy for creating visual excellence in computer-generated imagery and simulation. The approach it advocates requires the use of powerful hardware, sophisticated software, and top creative talent. With a CRAY supercomputer at the heart of its computer network and its own proprietary image rendering and simulation software, Digital Productions is revolutionizing state-of-the-art computer graphics. At the forefront of computer graphics technology, Digital Productions is redefining traditional methods of visual communications and creating new forms of self-expression, instruction, and entertainment.

What Should Artificial Intelligence Want from the Supercomputers

While some proposals for supercomputers increase the powers of existing machines like CDC and Cray supercomputers, others suggest radical changes of architecture to speed up non-traditional operations such as logical inference in PROLOG, recognition/ action in production systems, or message passing. We examine the case of parallel PROLOG to identify several related computations which subsume those of parallel PROLOG, but which have much wider interest, and which may have roughly the same difficulty of mechanization. Similar considerations apply to some other proposed architectures as well, raising the possibility that current efforts may be limiting their aims unnecessarily.