User: rhaas Date: 2012/03/06 07:15 PM

Modified: / ET.tex

Log: add some text explaining parallelism in TwoPunctures and Lorene and why TP is usually used on line but Lorene is not

add some LaTeX code to provide a bibfont command indep of whether it has been defined before or not

File Changes:

Directory: / ============

File [modified]: ET.tex Delta lines: +20 -6 =================================================================== --- ET.tex 2012-03-07 00:39:13 UTC (rev 263) +++ ET.tex 2012-03-07 01:15:39 UTC (rev 264) @@ -28,7 +28,14 @@ \usepackage[all]{hypcap}

\usepackage[sort&compress,numbers]{natbib} +% RH: we might have to remove the ifx stuff for submission, but it allows both +% Bruno, Roland and Peter to build this file and all have \bibfont be +% \footnotesize +\ifx\bibfont\relax +\newcommand{\bibfont}{\footnotesize} +\else \renewcommand{\bibfont}{\footnotesize} +\fi

\graphicspath{{figures/} {cactus-benchmarks/} @@ -286,7 +293,7 @@ component's function is explained and, if applicable, the physical system it models is introduced. The next major section, Section~\ref{sec:examples}, contains a set of sample results obtained using the toolkit's components. -Examples from numerical relativity, astrophysics and cosmolology are provided. +Examples from numerical relativity, astrophysics and cosmology are provided.

\section{Requirements}% \label{sec:requirements} @@ -949,7 +956,13 @@ The most widely used routine to generate initial data for these is the \codename{TwoPunctures} code (described originally in~\cite{Ansorg:2004ds}) which solves the binary puncture equations for a pair of BHs~\cite{Brandt:1997tf}. -To do so, one assumes the extrinsic curvature for each BH corresponds to +\codename{TwoPunctures} is a serial code that does not parallelize +beyond a single node through the use of \codename{OpenMP}. Nevertheless it has +proven sufficiently robust and fast that the vacuum initial data problem is +usually solved on-line at the beginning of an inspiral simulation. This +approach allows all information required to set up the run to be kept in the +single \codename{Cactus} parameter file. +\codename{TwoPunctures} assumes the extrinsic curvature for each BH corresponds to the Bowen-York form~\cite{Bowen:1980yu}: \begin{eqnarray} K_{(m)}^{ij}&=&\frac{3}{2r^2}(p_{(m)}^i\hat{N}^j+p_{(m)}^j\hat{N}^i-(\gamma^{ij}-\hat{N}^i\hat{N}^j)p_{(m)}^k\hat{N}_k))\nonumber\ @@ -1007,10 +1020,11 @@ by the {\tt Lorene} code~\cite{Lorene:web,Gourgoulhon:2000nn}, though it does not currently include the capability of generating such data from scratch. For a number of reasons, such functionality is not truly required; in particular, -{\tt Lorene} is a serial code and to call it as -an Einstein Toolkit initial data generator saves no time. Also, it is not guaranteed to be convergent for -an arbitrary set of parameters; thus the initial data routine itself may never -finish its iterative steps. Instead, recommended practice is to let Lorene output +{\tt Lorene} is often used to generate initial data involving neutron stars, +where the initial data problem is sufficiently hard that finding a numerical +solution requires a significant amount of computer time as well as some human +oversight in order to guide the routine to the correct solution. +Therefore, recommended practice is to let Lorene output data into files, and then read those into the Einstein Toolkit at the beginning of a run.

Lorene uses a multigrid spectral approach to solve the conformal thin-sandwich