[Commits] [svn:einsteintoolkit] Paper_EinsteinToolkit_2010/ (Rev. 25)

schnetter at cct.lsu.edu schnetter at cct.lsu.edu
Mon Feb 14 09:14:40 CST 2011

User: eschnett
Date: 2011/02/14 09:14 AM


 Describe Carpet

File Changes:

Directory: /

File [modified]: ET.tex
Delta lines: +35 -23
--- ET.tex	2011-02-14 02:15:34 UTC (rev 24)
+++ ET.tex	2011-02-14 15:14:39 UTC (rev 25)
@@ -295,12 +295,15 @@
 \S\ref{broader} describes the growing impact of Cactus in other
 application domains.
+\todo{ES: Say something about component structure and component
+  interfaces; I refer to this in the Carpet subsection below.}
 The Cactus Framework was developed by the
 numerical relativity community, and although it is a general component
 framework that supports different application domains its core user
 group has remained from numerical relativity. The Cactus team has
 traditionally developed and supported a set of core modules for numerical
-relativity, as part of the {\tt CactusEinstein} arrangement. Over the
+relativity, as part of the \texttt{CactusEinstein} arrangement. Over the
 last few years however, the relevance of many of the modules has declined,
 and more and more of the basic infrastructure for numerical relativity
 has been provided by open modules provided and distributed by research
@@ -308,30 +311,40 @@
 The Einstein Toolkit now collects the widely used parts of CactusEinstein,
 combined with contributions from the community.
-\subsection{Carpet Mesh Refinement}
-\todo{1/2 page Erik}
+\subsection{Adaptive Mesh Refinement}
-While Cactus is distributed with a structured-mesh unigrid MPI
-parallel driver (\codename{PUGH}\footnote{Recent results with a
-  numerical relativity benchmark and PUGH show excellent scaling to
-  130,000 BG/P cores}), most GR work is now carried out
-with the adaptive-mesh refinement (AMR) driver
-\textbf{Carpet}\footnote{Other mesh refinement libraries that have
-  been integrated with Cactus include \codename{PAGH}
-  (\codename{GrACE}~\cite{CS_grace_web}), \codename{ParCa}
-  (\codename{PARAMESH}~\cite{CS_PARAMESH_web}) and \codename{Taka}
-  (\codename{SAMRAI}~\cite{CS_SAMRAIweb})}~\cite{CS_Schnetter-etal-03b,
-  CS_carpet_web} whose development is overseen by Schnetter.
-Carpet provides parallel (MPI and OpenMP) AMR capabilities
-on block-structured grids to applications, handling memory management,
-parallelization, I/O, while providing an interface to user application
-code to select and change the desired grid hierarchy.  In addition,
-Carpet provides a multi-block infrastructure, allowing the coupling of
-multiple logically Cartesian, but general grid blocks.  Carpet is open
-source and is openly developed, with the main development located at
-LSU (co-PI Schnetter) and contributions from AEI and others.
+In Cactus, infrastructure capabilities such as memory management,
+parallelisation, time evolution, mesh refinement, and I/O are
+delegated to a set of special \emph{driver} components. This helps
+separate physics code from infrastructure code; in fact, a typical
+physics component (implementing e.g.\ the Einstein or relativistic MHD
+equations) does not contain any code or subroutine calls having to do
+with parallelisation, time evolution, or mesh refinement. The
+information provided in the interface declarations of the individual
+components allows a highly efficient execution of the combined
+The Einstein Toolkit offers two drivers, \emph{PUGH} and
+\emph{Carpet}. PUGH provides domains consisting of a uniform Cartesian
+grid, and is highly scalable (up to more than 130,000 cores on a Blue
+Gene/P \todo{cite}). Carpet \cite{Schnetter:2003rb, Schnetter:2006pg,
+  CarpetCode:web} provides multi-block methods and adaptive mesh
+refinement (AMR\@). Multi-block methods cover the domain with a set of
+distorted Cartesian blocks that exchange information e.g.\ via
+interpolation or penalty methods. The AMR capabilities employ the
+standard Berger-Oliger algorithm \todo{cite} with subcycling in time.
+Carpet is the main driver used today for Cactus-based astrophysical
+simulations. Carpet offers hybrid MPI/OpenMP parallelisation and is
+used in production on up to several thousand cores. We estimate that,
+in 2010, about 7,000 core years of computing time (45 million core
+hours) were used via Carpet by more than a dozen research groups
+world-wide. To date, more than 90 peer-reviewed publication and more
+than 15 student theses are based on Carpet \todo{update these numbers}
+\cite{CarpetCode:web}. Carpet's continued development is overseen by
+E. Schnetter \todo{do we want names here?} and is funded via several
+NSF awards \todo{list them somewhere}.
 \subsection{Simulation Factory}
 \todo{1/3 page Erik}
@@ -610,4 +623,3 @@

More information about the Commits mailing list