NAS researchers present their work at conferences around the world and publish their findings in scientific journals. Some of these become official NAS Technical Reports and are not duplicated on this page. To download the PDF-formatted files you will need the free Adobe Acrobat Reader.

Anisotropic Nanomechanics of Boron Nitron Nanotubes: Nanostructured "Skin" Effect (PDF,145K)
by Deepak Srivastava, Madhu Menon and KyeongJae Cho

The stiffness and plasticity of boron nitride nanotubes are investigated using generalized tight-binding molecular dynamics and ab initio total energy methods. Due to boron nitride (BN) bond rotation effect, compressed zigzag BN nanotubes are found to undergo anisotropic strain release, followed by anisotropic plastic buckling. The strain is preferentially released toward N atoms in the rotated BN bonds. The tubes buckle anistropically toward only one end when uniaxially compressed from both ends. A "skin-effect" model of smart nanocomposite materials, which localizes the structural damage toward the skin or surface side of the material, is proposed.

BN nanotubes (mpeg, 991K)

A Comparison of Three Programming Models for Adaptive Applications on the Origin 2000 (PDF, 210K)
by Hongzhang Shan, Jaswinder Pal Singh, Princeton University; Leonid Oliker, National Energy Research Scientific Computing Center, Rupak Biswas, NASA Ames Research Center.

Adaptive applications have computational workloads and communication patterns which change unpredictably at runtime, requiring dynamic load balancing to achieve scalable performance on parallel machines. This paper compares the performance of and programming effort required for two major classes of adaptive applications (dynamic remeshing and N-body) under three leading parallel programming models (MPI, SHMEM, and CC-SAS) on an SGI Origin 2000 system, which supports all three models efficiently.

Parallelization of a Dynamic Unstructured Application using Three Leading Paradigms (PDF, 1 MB)
principal investigators: Leonid Oliker, National Energy Research Scientific Computing Center. Rupak Biswas, NASA Ames Research Center.

The success of parallel computing in solving real-life computationally intensive problems relies on their efficient mapping and execution on large-scale multiprocessor architectures. Many important applications are both unstructures and dynamic in nature, making their efficient parallel implentation a daunting task. This paper presents the parallelization of a dynamic unstrcutured mesh adaptation using three popular programming paradigms on three leading supercomputers.

Real-Time Lunar Prospector Data Visualization Using Web-Based Java
by D. Glenn Deardorff and Bryan D. Green

For the first time, the World Wide Web was used to graphically display near-real-time data from the Lunar Prospector planetary exploration mission to the global public. Science data from the craft's instruments, as well as engineering data for the spacecraft subsystems, are continuously displayed in time-varying XY plots, with a lag of a few minutes after the initial downlink. In addition, the craft's current location is displayed relative to the whole Moon, and as an off-craft observer would see in the reference frame of the craft, with the lunar terrain scrolling by underneath. This paper describes each of these facets, and address the successes and pitfalls in using the relatively new Java and Web technologies as media for sharing near-real-time planetary exploration data with the public.

Measurement of a Scientific Workload using the IBM Hardware Performance Monitor
by Robert J. Bergeron

Presents data on the performance of an IBM SP2 system using RS6000 hardware monitors and a performance measurement tool. The data collected showed that the SP2 averages about 1.3 Gflops, about 3% of peak. Provides the relative usage for various hardware units over the entire workload measured over a 9-month period. The workload displays moderate parallelism, with the most popular choice of nodes as 16. Although the monitor data provide a good snapshot of workload performance, causal correlations regarding key performance indicators appear difficult to draw from the current data due to the absence of I/O delay measurements.