Introduction and Overview

The main goals and important issues in developing parallel algorithms of unstructured mesh calculations


Introduction

Unstructured meshes are a powerful computational tool used in the numerical modeling of physical phenomena on complex, irregular domains. Instead of requiring a uniform distribution of grid points, unstructured meshes allow grid points to be strategically placed in the computational domain. Thus, these meshes are particularly effective in modeling irregular boundaries, multiscale geometries, and rapidly changing solutions.

Much research has been done to design sequential algorithms and techniques to effectively use unstructured meshes in the solution of large-scale applications. Unfortunately, many of these applications cannot take advantage of the power of parallel computing because of a lack of widely available software tools on distributed memory architectures. In addition, no attempt has been made to provide a publicly available software product that integrates key aspects of unstructured mesh computation on parallel computers.

The SUMAA3D (Scalable Unstructured Mesh Algorithms and Applications) project will rectify this situation. Our expertise in the development of parallel algorithms for unstructured mesh computation, our experience in distributing and maintaining software in the public domain, and our commitment to work with application scientists in a variety of disciplines and industries will propel this software tool to the forefront of computational science.


Overview

The SUMAA3D project comprises three primary components (1) algorithmic design and implementation, (2) software development and dissemination, and (3) large-scale application solution.

Algorithmic Design and Implementation

The first component is fundamental research into the development of provably good, provably fast algorithms for all aspects of unstructured mesh computation on distributed memory architectures. In particular, we have identified the following basic problems that are fundamental to massively parallel computation on unstructured domains:


Next: Accomplishements
Back: Unstructured Mesh Home Page


freitag@mcs.anl.gov / jones@cs.utk.edu / plassman@mcs.anl.gov
Argonne National Laboratory / The University of Tennessee