Abstract

On the basis of numerous fluid simulation experiences, researchers have discovered that many common operations can be abstracted to form a general fluid simulation framework. These operations include the discretization of equations and variables, the computation of the coefficients, the assembly of the linear or nonlinear systems, the solving of the systems, etc. Furthermore, all of the operations can be attributed to the operations “in the field”, which is an abstract concept derived from the equations and variables. Thus, fluid simulations can be performed under a field-based general framework. Moreover, in response to the urgent need for large-scale fluid simulations, parallelism is integrated into the framework. Due to the convenience of the field operations, parallelization of the framework can be realized on both the OpenMP and MPI levels. In other words, because of the newly defined “fields”, a series of operations in fluid simulations can be simplified and unified. However, very few studies have noted this advantage, and therefore, this work attempts to fill the void. With the help of a field-based general framework, it is anticipated that the parallel codes of fluid simulations can be generated easily and quickly. As an application of the general framework, a parallel 3D simulator for matrix acidization called Masor is developed. The simulation results are regarded as physically reasonable by many studies, which verifies the correctness and effectiveness of the general framework. In addition, it is noteworthy that the parallel performance of Masor is decided by a solver.

Highlights

  • Computational fluid dynamics (CFD) [1, 2] focuses on the numerical simulation of fluids using computer technology

  • Different fluid simulations can be deemed to follow the same kind of pattern, which necessitates building a general framework for all types of fluid simulations, i.e., the field-based general framework

  • The idea of a field-based general framework is summarized as follows—Due to the conservation laws, there are many common terms among the equations of fluid simulations, such as the divergence term and the Laplace term, so that the variables in the equations have common discretization patterns that are recognized by the discretization library

Read more

Summary

Introduction

Computational fluid dynamics (CFD) [1, 2] focuses on the numerical simulation of fluids using computer technology. The simulation procedure, in reality, involves constructing the linear systems and solving them, which is a procedure that applies to most applications of computational fluid dynamics. With this philosophy, a general framework to simulate fluid dynamics is suggested in this work. A Field-based general framework to simulate fluids in parallel advantages of OpenMP and MPI can be leveraged, and the parallel performance is further enhanced. The general framework in this work applies OpenMP in the entire coefficient computing routine, which greatly improves the parallel efficiency on the thread level.

The discretization library
The coefficient library
The linear system library
Parallelization
The mathematical model
The structure of the code
The module “Driver”
The module “Resi”
The module “ConstructMat”
The evaluation of Masor
Findings
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call