Accession Number : ADA186583

Title :   Supercomputers for Solving PDE (Partial Differential Equations) Problems.

Descriptive Note : Final rept. 20 Sep 86-11 Aug 87,

Corporate Author : UNIVERSITY OF SOUTHERN CALIFORNIA LOS ANGELES COMPUTER RESEARCH INST

Personal Author(s) : Hwang, Kai

PDF Url : ADA186583

Report Date : 11 Aug 1987

Pagination or Media Count : 10

Abstract : This project investigated parallel/vector supercomputer architectures for solving Air Force problems, which demand the solution of partial differential equations (PDEs). We have developed an orthogonal multiprocessor (omp) architecture for efficiently implementing the SLOR and ADI methods in solving PDEs. Another parallel PDE machine architecture, called the V-tree multiprocessor, has been developed for mapping the multigrid algorithms. This V-tree is shown to be more effective than the well-known hypercube and mesh architectures. Both the OMP and the V-tree architectures can demonstrate linear speedup by exploiting parallelism and vectorization. Continued efforts are needed to expand these initial studies into real hardware experiments and software simulations to verify the theoretical predictions on speedup performance.

Descriptors :   *COMPUTER ARCHITECTURE, *MULTIPROCESSORS, *SUPERCOMPUTERS, AIR FORCE, COMPUTER PROGRAMS, COMPUTERIZED SIMULATION, MESH, ORTHOGONALITY, PARTIAL DIFFERENTIAL EQUATIONS, PREDICTIONS, PROBLEM SOLVING, THEORY

Subject Categories : Computer Hardware

Distribution Statement : APPROVED FOR PUBLIC RELEASE