Thread injection to make DNS Channelflow run in parallel: part 1

[thumbnail of thread_injection_5-P1a.docx]
Restricted to Repository staff only
[thumbnail of thread_injection_5-P1a.pdf]
Preview
thread_injection_5-P1a.pdf - Published Version (3MB) | Preview

Grossman, Igor and Thorpe, Graham ORCID: 0000-0001-8160-9253 (2016) Thread injection to make DNS Channelflow run in parallel: part 1. Technical Report. Victoria University, Melbourne.

Abstract

The flows of Newtonian fluids are accurately described by the Navier-Stokes equation. The equation is solved simultaneously with the mass continuity equation, and it may also be solved concurrently with equations that govern the flow of scalar quantities. Analytical solutions of the equation cannot be obtained except for the simplest of geometries, and generally when the flow is laminar. As a result the equation must be solved numerically. The problem is compounded when the flow is turbulent as such flows are characterized by their displaying a wide range of temporal and spatial scales that must be resolved. This is impractical for many industrial applications; hence engineers and scientists usually resort to empirical models to account for the effects of turbulence on the flow field. However, it is useful to have benchmark solutions against which the empirical models can be compared, and these solutions can be obtained by direct numerical solution of the Navier-Stokes equation. This is computationally demanding and in this work we demonstrate in detail how thread injection can be used to parallelize the numerical solution. The program is a modified form Channelflow which is open source software produced by chanelflow.org. The present work is quite voluminous and for this reason it has been presented in two fascicles, namely Parts 1 and 2. Readers requiring more details and assistance are most welcome to contact Mr Igor Grossman at igor.grossman@live.vu.edu.au.

Additional Information

Version 0.1.1 -- Initial revision -- May 2016

Item type Monograph (Technical Report)
URI https://vuir.vu.edu.au/id/eprint/31080
Subjects Historical > FOR Classification > 0105 Mathematical Physics
Current > Division/Research > College of Science and Engineering
Keywords DNS; turbulence; parallelisation; computer codes; direct numerical simulation; parallelization
Download/View statistics View download statistics for this item

Search Google Scholar

Repository staff login