X-Git-Url: https://bilbo.iut-bm.univ-fcomte.fr/and/gitweb/book_gpu.git/blobdiff_plain/fd66513f28a83144dbb902570a63fc94787d95d2..ecd01808b5702d940bd77107a2bf829d3832179b:/BookGPU/Chapters/chapter5/ch5.tex?ds=sidebyside

diff --git a/BookGPU/Chapters/chapter5/ch5.tex b/BookGPU/Chapters/chapter5/ch5.tex
index f7eabb4..bdfad6a 100644
--- a/BookGPU/Chapters/chapter5/ch5.tex
+++ b/BookGPU/Chapters/chapter5/ch5.tex
@@ -3,7 +3,7 @@
 %\chapterauthor{Allan S. Nielsen}{Technical University of Denmark}
 %\chapterauthor{Bernd Dammann}{Technical University of Denmark}
 
-\chapter[Software components for heterogeneous manycore architectures]{Development of software components for heterogeneous manycore architectures}\label{ch5}
+\chapter[Software components for heterogeneous many-core architectures]{Development of software components for heterogeneous many-core architectures}\label{ch5}
 
 %Subjects:
 %\begin{itemize}
@@ -160,7 +160,7 @@ We refer the reader to Chapter \ref{ch7} for an example of a scientific applicat
 \subsection{Heat conduction equation}\index{heat conduction}
 First, we consider a two-dimensional heat conduction problem defined on a unit square. The heat conduction equation is a parabolic partial differential diffusion equation, including both spatial and temporal derivatives. It describes how the diffusion of heat in a medium changes with time. Diffusion equations are of great importance in many fields of sciences, e.g., fluid dynamics, where the fluid motion is uniquely described by the Navier-Stokes equations, which include a diffusive viscous term~\cite{ch5:chorin1993,ch5:Ferziger1996}.%, or in financial science where diffusive terms are present in the Black-Scholes equations for estimation of option price trends~\cite{}.
 
-The heat problem is an IVP \index{initial value problem}, it describes how the heat distribution evolves from a specified initial state. Together with homogeneous Dirichlet boundary conditions\index{boundary conditions}, the heat problem in the unit square is given as
+The heat problem is an IVP \index{initial value problem}, it describes how the heat distribution evolves from a specified initial state. Together with homogeneous Dirichlet boundary conditions\index{boundary condition}, the heat problem in the unit square is given as
 \begin{subequations}\begin{align}
 \frac{\partial u}{\partial t} - \kappa\nabla^2u =  0, & \qquad (x,y)\in \Omega([0,1]\times[0,1]),\quad t\geq 0, \label{ch5:eq:heateqdt}\\
 u =  0, & \qquad (x,y) \in \partial\Omega,\label{ch5:eq:heateqbc}
@@ -188,7 +188,7 @@ We use a Method of Lines (MoL)\index{method of lines} approach to solve \eqref{c
 \begin{align}\label{ch5:eq:discreteheateq}
 \frac{\partial u}{\partial t} = \mymat{A}\myvec{u}, \qquad \mymat{A} \in \mathbb{R}^{N\times N}, \quad \myvec{u} \in \mathbb{R}^{N},
 \end{align}
-where $\mymat{A}$ is the sparse finite difference matrix and $N$ is the number of unknowns in the discrete system. The temporal derivative is now free to be approximated by any suitable choice of a time-integration method\index{time integration}. The most simple integration scheme would be the first-order accurate explicit forward Euler method\index{forward Euler},
+where $\mymat{A}$ is the sparse finite difference matrix and $N$ is the number of unknowns in the discrete system. The temporal derivative is now free to be approximated by any suitable choice of a time-integration method\index{time integration}. The most simple integration scheme would be the first-order accurate explicit forward Euler method\index{Euler!forward Euler},
 \begin{align}\label{ch5:eq:forwardeuler}
 \myvec{u}^{n+1} = \myvec{u}^n + \delta t\,\mymat{A}\myvec{u}^n,
 \end{align}
@@ -295,7 +295,7 @@ The spatial derivative in \eqref{ch5:eq:poissoneq} is again approximated with fi
 \begin{align}
 \mymat{A}\myvec{u}=\myvec{f}, \qquad \myvec{u},\myvec{f} \in \mathbb{R}^{N}, \quad \mymat{A} \in \mathbb{R}^{N\times N}, \label{ch5:eq:poissonsystem}
 \end{align}
-where $\mymat{A}$ is the sparse matrix formed by finite difference coefficients, $N$ is the number of unknowns, and $\myvec{f}$ is given by \eqref{ch5:eq:poissonrhs}. Equation \eqref{ch5:eq:poissonsystem} can be solved in numerous ways, but a few observations may help do it more efficiently. Direct solvers based on Gaussian elimination are accurate and use a finite number of operations for a constant problem size. However, the arithmetic complexity grows with the problem size by as much as $\mathcal{O}(N^3)$ and does not exploit the sparsity of $\mymat{A}$. Direct solvers are therefore mostly feasible for dense systems of limited sizes. Sparse direct solvers exist, but they are often difficult to parallelize, or applicable for only certain types of matrices. Regardless of the discretization technique, the discretization of an elliptic PDE into a linear system as in \eqref{ch5:eq:poissonsystem} yields a very sparse matrix $\mymat{A}$ when $N$ is large. Iterative methods\index{iterative methods} for solving large sparse linear systems find broad use in scientific applications, because they require only an implementation of the matrix-vector product, and they often use a limited amount of additional memory. Comprehensive introductions to iterative methods may be found in any of~\cite{ch5:Saad2003,ch5:Kelley1995,ch5:Barrett1994}.
+where $\mymat{A}$ is the sparse matrix formed by finite difference coefficients, $N$ is the number of unknowns, and $\myvec{f}$ is given by \eqref{ch5:eq:poissonrhs}. Equation \eqref{ch5:eq:poissonsystem} can be solved in numerous ways, but a few observations may help do it more efficiently. Direct solvers based on Gaussian elimination are accurate and use a finite number of operations for a constant problem size. However, the arithmetic complexity grows with the problem size by as much as $\mathcal{O}(N^3)$ and does not exploit the sparsity of $\mymat{A}$. Direct solvers are therefore mostly feasible for dense systems of limited sizes. Sparse direct solvers exist, but they are often difficult to parallelize, or applicable for only certain types of matrices. Regardless of the discretization technique, the discretization of an elliptic PDE into a linear system as in \eqref{ch5:eq:poissonsystem} yields a very sparse matrix $\mymat{A}$ when $N$ is large. Iterative methods\index{iterative method} for solving large sparse linear systems find broad use in scientific applications, because they require only an implementation of the matrix-vector product, and they often use a limited amount of additional memory. Comprehensive introductions to iterative methods may be found in any of~\cite{ch5:Saad2003,ch5:Kelley1995,ch5:Barrett1994}.
 
 One benefit of the high abstraction level and the template-based library design is to allow developers to implement their own components, such as iterative methods for solving sparse linear systems. The library includes three popular iterative methods: conjugate gradient, defect correction\index{defect correction}, and geometric multigrid. The conjugate gradient method is applicable only to systems with symmetric positive definite matrices. This is true for the two-dimensional Poisson problem, when it is discretized with a five-point finite difference stencil, because then there will be no off-centered approximations near the boundary. For high-order approximations ($\alpha>1$), we use the defect correction method with multigrid preconditioning. See e.g., \cite{ch5:Trottenberg2001} for details on multigrid methods.