Theses and Dissertations

Issuing Body

Mississippi State University

Advisor

Janus, Mark

Committee Member

Burg, Clarence

Committee Member

Luke, Edward

Date of Degree

8-3-2002

Document Type

Graduate Thesis - Open Access

Major

Computational Engineering (Program)

Degree Name

Master of Science

College

College of Engineering

Department

Computational Engineering Program

Abstract

A coarse-grain parallel multi-block algorithm was designed for CHEQNS - a multi-block solver for solving chemically reacting flows in local chemical equilibrium and has been implemented using the Message Passing Interface (MPI). The parallel implementation confirms to the Single Program Multiple Data (SPMD) model. The parallel implementation uses synchronous update of fluxes across the block-block boundaries. The solution algorithm consists of block-decoupled Gauss-Seidel iterations. The coupling between the sub-domains on different processors occurs at the Newton iteration level. The parallel implementation is general and can accept an arbitrary arrangement of blocks in multi-block configuration with multiple blocks per processor. The parallel implementation has been verified against the results from the sequential multi-block solver for different types of flows. The parallel performance has been studied in terms of speed-up and efficiency. The influence of parallelization on the convergence was also studied.

URI

https://hdl.handle.net/11668/20765

Share

COinS