John Ambrosiano

A vectorized particle tracer for unstructured grids

Abstract A vectorized particle tracer for unstructured grids is described. The basic approach is to use elementary properties of the linear basis functions to search for particles on the grid using the element last occupied as an initial guess. To permit vectorization, a simple binary sort of the particles is performed every timestep such that all particles that have as yet not found their host element remain at the top of the list. In this way, vector-loops can be easily formed. Timings taken from a numerical example indicate that speed-ups of the order of 1:14 can be obtained on vector-machines when using this algorithm.

High-order upwind flux correction methods for hyperbolic conservation laws

Abstract Totally one-sided first-order and second-order schemes are presented employing a numerically calculated characteristic speed direction and are combined into a transportive, monotonicity-preserving hybrid scheme using the method of flux correction. The first-order scheme is free of expansion shocks and artificial extrema. The hybrid scheme computes a provisional update from the first-order scheme, and then filters the second-order corrections to prevent occurrence of new extrema. Computed versus analytic results are compared for two different N-wave shocks and for a third case involving linear advection of a square wave. Results are given with and without the second-order correction. The second-order results are always superior to first-order results, with the most dramatic difference occurring in the case of linear advection. The results suggest that higher than second-order upwind differences could be substituted in the hybrid scheme to reduce truncation error even further.

A finite element formulation of the Darwin PIC model for use on unstructured grids

In this paper we introduce a new formulation of the Darwin approximation of Maxwells equations and discuss its domain of applicability. We describe our finite element implementation of this model, allowing the use of unstructured grids, and its coupling with a PIC method for the particles.

CLIMATE SYSTEM MODELING USING A DOMAIN AND TASK DECOMPOSITION MESSAGE-PASSING APPROACH

Abstract We have developed a Climate System Modeling Framework (CSMF) for high-performance computing systems, designed to schedule and couple multiple physics simulation packages in a flexible and transportable manner. Some of the major packages in the CSMF include models of atmospheric and oceanic circulation and chemistry, land surface and sea ice processes, and trace gas biogeochemistry. Parallelism is achieved through both domain decomposition and process-level concurrency, with data transfer and synchronization accomplished through message-passing. Both machine transportability and architecture-dependent optimization are handled through libraries and conditional compile directives. Preliminary experiments with the CSMF have been executed on a number of high-performance platforms, including the Intel Paragon, the TMC CM-5 and the Meiko CS-2, and we are in the very early stages of optimization. Progress to date is presented.

High-performance climate system modeling using a domain and task decomposition message-passing approach

We have developed a climate system modeling framework (CSMF) for high-performance systems, designed to schedule and couple multiple physics simulation packages in a flexible and transportable manner. Some of the major packages in the CSMF include models of atmospheric and oceanic circulation and chemistry, land surface and sea ice processes, and trace gas biogeochemistry. Parallelism is achieved through both domain decomposition and process-level concurrency, with data transfer and synchronization accomplished through message-passing. Both machine transportability and architecture-dependent optimization are handled through libraries and conditional compile directives. Preliminary experiments with the CSMF have been executed on a number of high-performance platforms, including the Intel Paragon, the TMC CM-5 and the Meiko CS-2, and we are in the very early stages of optimization. Progress to date (1994) is presented.<<ETX>>

A one-dimensional PIC-circuit code for simulating a reflex triode

Abstract We describe a computer model capable of simulating an idealized one-dimensional reflex triode. Charged particle motion and electron scattering by a thin foil are calculated self-consistently with the response of an external circuit. The code has many potential applications in pulse power physics. We describe the model in detail and give a computational example. Our test case follows the evolution to a low-impedance steady state of a reflex triode driven symmetrically by a capacitive discharge.

Electromagnetic scattering calculations using a finite—element solver for the Maxwell equations

We describe a pair of finite-element codes which use unstructured meshes to solve the time-dependent Maxwell equations, with particular emphasis on their application to electromagnetic scattering problems. A two-step, flux-corrected transport scheme is used to effect the time integration, while the spatial structure of the field is determined by a Galerkin procedure. The basis functions are piecewise-linear on three-noded triangles in two dimensions and four-noded tetrahedra in three. For the periodic scattering problems with which we are presently concerned, adaptive remeshing is a convenient and powerful method for improving the quality of the solutions. Results for the analytically tractable case of scattering by a perfectly conducting circular cylinder are used to illustrate the performance of the codes.

A finite element particle code on an unstructured grid

Summary form only given. A prototype code called Mad Max has been developed. It applies finite-element techniques on an unstructured grid of triangles to electromagnetic particle simulation. The unstructured grid provides enormous geometric flexibility in fitting the computational nodes to a particular problem and also makes adaptive refinement straightforward

A finite element formulation of the Darwin electromagnetic PIC model for unstructured meshes of triangles

Summary form only given, as follows. The Darwin model for electromagnetic simulation is a reduced form of the Maxwell-Vlasov system that retains all essential physical processes except the propagation of light waves. It is useful in modeling systems for which the light-transit time scales are less important than Alfven wave propagation or quasistatic effects. The Darwin model is elliptic rather than hyperbolic, as is the full set of Maxwells equations. Appropriate boundary conditions must be chosen for the problems to be well-posed. Using finite element techniques to apply this method to unstructured triangular meshes, a mesh made up of unstructured triangles allows realistic device geometries to be modeled without the necessity of using a large number of mesh points. Analyzing the dispersion relation makes it possible to validate the code as well as the Darwin approximation.

Time domain propagation studies with the NPE: Two short movies

The nonlinear progressive wave equation (NPE) model has been used to investigate the propagation of linear and nonlinear broadband signals in a range‐dependent shallow water waveguide. The NPE is the nonlinear time domain equivalent of the frequency domain parabolic wave equation (PE). Results from each of the studies are given in short computer‐generated movies illustrating a four‐cycle sine‐wave pulse propagating a total of 20 km in isovelocity water (c = 1500 m/s) over a bottom (c = 1600 m/s) whose depth profile contains a shelf of slope 0.15 connecting regions of constant depth 225 and 150 m. Effects evident in the movies are: radiation into the bottom resulting in head waves that re‐radiate back into the water column; energy dumping following reflection from the sloping bottom and surface; and mode separation at late times. Early time results suggest that bottom penetration is greater when nonlinearity is included.