The IBM SP1 is IBM`s newest parallel distributed-memory computer. As part of a joint project with IBM, Argonne took delivery of an early system in order to evaluate the software environment and to begin porting programming packages and applications to this machine. This report discusses the results of those efforts once the high-performance switch was installed. An earlier report (ANL/MCS-TM-177) emphasized software usability and the initial ports to the SP1. This report contains performance results and discusses some applications and tools not covered in TM 177.

This document reports on the work done by the Nuclear Waste Programs of the Chemical Technology Division (CMT), Argonne National Laboratory, in the period October 1991-March 1992. In these programs, studies are underway on the performance of waste glass and spent fuel in projected nuclear repository conditions to provide input to the licensing of the nation's high-level waste repositories.

Motion-dependent magnetic forces are the key elements in the study of magnetically levitated vehicle (maglev) system dynamics. In the past, most maglevsystem designs were based on a quasisteady-motion theory of magnetic forces. This report presents an experimental and analytical study that will enhance our understanding of the role of unsteady-motion-dependent magnetic forces and demonstrate an experimental technique that can be used to measure those unsteady magnetic forces directly. The experimental technique provides a useful tool to measure motion-dependent magnetic forces for the prediction and control of maglev systems.

Abstract: Statistical guidelines are described for the development of a pilot scientific observer sampling program to collect turtle interactions data from the Hawaii longline fishery. The best available data are presented and potential sources of variability in longline turtle take rates are discussed. A stratified sampling design for a pilot survey is developed based on time and target species. An array of sample size choices for the pilot program is presented covering a range of tolerance levels and confidence levels for estimation of total turtle take.

The dynamic load-balancing problem for mesh-connected parallel computers can be clearly described by introducing a function that identifies how much work is to be transmitted between neighboring processors. This function is a solution to an elliptic problem for which a wealth of knowledge exists. The non-uniqueness of the solution to the load-balancing problem is made explicit.