Contributions to simulation of modelica models on data-parallel mutli-core architectures / by Kristian Stavåker.

Format:

Book

Author/Creator:

Stavåker, Kristian, author.

Series:

Linköping studies in science and technology. Dissertations ; Number 1666.

Linköping Studies in Science and Technology. Dissertation, 0345-7524 ; Number 1666

Language:

English

Subjects (All):

Object-oriented methods (Computer science).

Computer simulation.

Electronic data processing--Distributed processing.

Electronic data processing.

Physical Description:

1 online resource (262 pages) : illustrations (some color).

Edition:

1st ed.

Place of Publication:

Linköping, Sweden : Department of Computer and Information Science, Linköping University, 2015.

Contents:

Intro

I Prologue

Introduction

Motivation

Research Questions

Research Process

Contributions

Delimitations

List of Publications

Thesis Outline

Part I

Part II

Part III

Part IV

Part VI

Background

The Modelica Modeling and Simulation Language

The OpenModelica Development Environment

Mathematical Concepts

ODE and DAE Representation

ODE and DAE Numerical Integration Methods

Partial Differential Equations (PDEs)

PDE Solving Software

Causalization of Equations

Sorting Example

Sorting Example with Modelica Model

Conversion to Causal Form in Two Steps

Algebraic Loops

Compiler Structure

Compilation and Simulation of Modelica Models

Multi-Core Computing

Graphics Processing Units (GPUs)

The Fermi Architecture

CUDA

OpenCL

OpenACC

Previous Research

Early Work with Compilation of Mathematical Models to Parallel Executable Code

Task Scheduling and Clustering Approach

Task Graphs

Modpar

Inlined and Distributed Solver Approach

Distributed Simulation using Transmission Line Modeling

PDE Modeling with Modelica

Related Research in other Research Groups

II Parallel Simulation of Equation-Based Models on Graphics Processing Units

Simulation of Equation-Based Models on the CELL BE Processor Architecture

The Cell BE Processor Architecture

Implementation

Measurements

Discussion

Simulation of Equation-Based Models with Quantized State Systems on Graphics Processing Units

Quantized State Systems (QSS)

Restricted Set of Modelica Models

Simulation of Equation-Based Models on Graphics Processing Units Utilizing Task Graph Creation

Case Study

Runtime Code and Generated Code

Discussion.

Compilation of Modelica Array Computation into Single Assignment C for Execution on Graphics Processing Units

Single Assignment C (SAC)

Extending the Algorithmic Subset of Modelica with Explicit Parallel Programming Constructs for Multi-Core Simulation

Parallel Variables

Parallel Functions

Kernel Functions

Parallel For-Loops

OpenCL Functionalities

Synchronization and Thread Management

Compilation of Unexpanded Modelica Array Equations

Problems with Expanding Array Equations

Splitting For-Equations with Multiple Equations in their Bodies

Algorithm

Examples

Transforming For-Equations and Array Equations into Slice Equations

Matching and Sorting of Unexpanded Array Equations

III Partial Differential Equation Modeling with Modelica via FMI Import of C++ Components

Partial Differential Equation Modeling with Modelica via FMI Import of C++ Components

Simulation Scenario 1: Heat Equation

Computing the Temperature Distribution

Proportional-Integral-Derivative (PID) Controller

Simulation Scenario 2: Elasticity Equation

Linear Elasticity Model

Coupled Implementation

The HiFlow3 Finite Element Library

The Functional Mock-Up Interface

Simulation Overview for Simulation Scenario 1

HiFlow3-based PDE Component - Simulation Scenario 1

Modelica Model 1

Modelica Model 2

Parallelization Concept

Parallel Execution of the Model

IV Using Parallel Skeletons from Modelica

Using Parallel Skeletons from Modelica

SkePU - Autotunable Multi-Back-end Skeleton Programming Framework for Multi-Core CPU and Multi-GPU Systems.

Containers

User Functions

Skeletons

Use Cases

Use Case 1: Main1

Use Case 2: SPH Fluid Dynamics

Modelica-SkePU Library

OpenModelica Compiler Extensions

Implementation Status

System Settings

Modelica-SkePU Test Suite Models - Serial

Modelica-SkePU Test Suite Models - Parallel

V Epilogue

What Kind of Problems are Graphics Processing Units Suitable for?

Are Graphics Processing Units Suitable for Simulating Equation-Based Modelica Models?

Discussion on the Various Approaches of Simulating Modelica Models on GPUs

Discussion on Modeling of Partial Differential Equations Modeling with Modelica

Discussion on Skeletons and Parallel Patterns in the context of Modelica

Future Work

Simulation of Modelica Models on GPUs

Skeleton and Parallel Pattern Programming in the Context of Modelica

Overview of the FastFlow Parallel Pattern Framework

VI Appendix

QSS Generated CUDA Code

Partial Differential Equation Modeling with Modelica via FMI Import of HiFlow3 C++ Components

Modelica-SkePU Library Code

Modelica-SkePU Test Suite Models

Basic SkePU Models

Mandelbrot Fractals

LU Factorization

Mean Square Error (MSE)

Pearson Product-Moment Correlation Coefficient (PPMCC)

Peak Signal to Noise Ratio (PSNR)

Taylor Series Calculation

Smooth Particle Hydrodynamics (SPH), Fluid Dynamics Shocktube simulation

A Runge-Kutta ODE solver

Glossary

Tom sida.

Notes:

Includes bibliographical references.

Description based on online resource; title from PDF title page (ebrary, viewed June 15, 2015).

ISBN:

91-7519-068-0