diff options
author | Erik Schnetter <schnetter@cct.lsu.edu> | 2011-02-15 11:21:40 -0500 |
---|---|---|
committer | Barry Wardell <barry.wardell@gmail.com> | 2011-12-14 18:26:01 +0000 |
commit | ae74a639e4a11249793442de694ae84d806d96f4 (patch) | |
tree | 9c1c90108215aa10e900e1657a87def879d192c1 /Carpet/CarpetWeb | |
parent | 4412d33c2c56fe068b8d9d5bcfec686bcdf043d7 (diff) |
CarpwetWeb: Update
Diffstat (limited to 'Carpet/CarpetWeb')
-rw-r--r-- | Carpet/CarpetWeb/get-carpet.html | 214 | ||||
-rw-r--r-- | Carpet/CarpetWeb/index.html | 181 | ||||
-rw-r--r-- | Carpet/CarpetWeb/olds.html | 126 |
3 files changed, 313 insertions, 208 deletions
diff --git a/Carpet/CarpetWeb/get-carpet.html b/Carpet/CarpetWeb/get-carpet.html index 3783c958d..8c46a21a0 100644 --- a/Carpet/CarpetWeb/get-carpet.html +++ b/Carpet/CarpetWeb/get-carpet.html @@ -15,44 +15,45 @@ <h2>Available Versions</h2> - <p>Carpet is distributed under the <a - href="http://www.gnu.org/licenses/licenses.html">GNU General - Public License (GPL)</a>. It might be released under the GNU - Lesser General Public License (LGPL) in the future, to match the - distribution terms of Cactus.</p> - - <p>There are currently three stable versions of Carpet available, - plus the current development version. Versions 1 and 2 have been - unchanged for quite some time, and should be considered outdated. - There are no plans to make any further changes to these - versions.</p> - - <p>Version 3 is the current stable version. There are no plans to - make further changes to this version unless a serious error is - detected. We recommend this version for the casual users and for - production runs.</p> + <p>Carpet is distributed under + the <a href="http://www.gnu.org/licenses/licenses.html">GNU + General Public License (GPL)</a>. It might be released under the + GNU Lesser General Public License (LGPL) in the future, to match + the distribution terms of Cactus.</p> + + <p>There is a stable version (version 4) and a development version + of Carpet available, plus several outdated versions. Versions 1, + 2, and 3 have been unchanged for quite some time, and should be + considered outdated. There are no plans to make any further + changes to these versions.</p> + + <p>Version 4 is the current stable version. There are no plans to + develop this version further, but errors will be corrected. We + recommend this version for the casual users and for production + runs.</p> <p>The development version will always see changes, some of which might surprise you. You should not use it without keeping in - close contact with the developers.</p> + close contact with the developers, i.e. following the relvant + mailing lists.</p> <h2>Downloading the Code</h2> - <p>Carpet is a driver for Cactus. It works as a part of Cactus, - and you will need to have the developers' version of Cactus - installed before you can use Carpet. Please look at the <a - href="http://www.cactuscode.org/">Cactus web pages</a> for an - introduction to Cactus and for installation instructions.</p> + <p>Carpet is a driver for Cactus. It works as a part of Cactus, + and you will need to have the Cactus installed before you can use + Carpet. Please look at + the <a href="http://www.cactuscode.org/">Cactus web pages</a> for + an introduction to Cactus and for install instructions.</p> <p>Carpet consists of several arrangements, each living in a - directory. The arrangement <code>Carpet</code> contains the basic - driver part that everybody needs. The arrangement + directory. The arrangement <code>Carpet</code> contains the basic + driver part that everybody needs. The arrangement <code>CarpetExtra</code> contains useful add-ons and some example - code. Development of experimental thorns happens in the + code. Development of experimental thorns happens in the <code>CarpetDev</code> arrangement, which means that the code in - there is not to be trusted. And finally, there is a graveyard + there is not to be trusted. And finally, there is a graveyard arrangement <code>CarpetAttic</code> of things that only used to be useful and are now in a state of decay.</p> @@ -79,13 +80,13 @@ cd arrangements ln -s ../carpet-stable-2/Carpet* .</pre> <p>(Don't miss the dot after the <code>Carpet*</code> in the last - line.) Instructions for using darcs are - given <a href="#darcs">below</a>. You can also have a look at + line.) Instructions for using darcs are + given <a href="#darcs">below</a>. You can also have a look at the <a href="http://www.carpetcode.org/~darcs/carpet-stable-2/">version 2 source tree</a> in your web browser.</p> - <h3>Version 3 (current stable version)</h3> + <h3>Version 3 (outdated)</h3> <p>Version 3 of Carpet is available via anonymous <a href="http://www.darcs.net/">darcs</a>:</p> @@ -94,43 +95,49 @@ cd arrangements ln -s ../carpet-stable-3/Carpet* .</pre> <p>(Don't miss the dot after the <code>Carpet*</code> in the last - line.) Instructions for using darcs are - given <a href="#darcs">below</a>. You can also have a look at + line.) Instructions for using darcs are + given <a href="#darcs">below</a>. You can also have a look at the <a href="http://www.carpetcode.org/~darcs/carpet-stable-3/">version 3 source tree</a> in your web browser.</p> <p>You can also obtain the darcs repository using <tt>wget</tt> - instead of <tt>darcs</tt>. For this, use the command</p> + instead of <tt>darcs</tt>. For this, use the command</p> <pre> wget -r -nH -np --cut-dirs=1 -R "index.html*" http://www.carpetcode.org/\~darcs/carpet-stable-3/</pre> <p>This copies the darcs repository into a subdirectory called <tt>carpet-stable-3</tt>, in much the same way as - the <tt>darcs get</tt> command above would. That is, you also end + the <tt>darcs get</tt> command above would. That is, you also end up with a fully functional local repository.</p> - <h3>Development Version</h3> + <h3>Version 4 (current stable version)</h3> - <p>The development version of Carpet is available via - <a href="http://git.or.cz/">git</a>:</p> + <p>Version 4 of Carpet is available + via <a href="http://git.or.cz/">git</a>:</p> <pre> cd Cactus git clone -o carpet git://carpetcode.org/carpet.git cd arrangements ln -s ../carpet/Carpet* .</pre> <p>(Don't miss the dot after the <code>Carpet*</code> in the last - line.) Instructions for using git are + line.) Instructions for using git are given <a href="#git">below</a>.</p> -<!-- This doesn't work yet - You can also have a look at - the <a href="http://www.carpetcode.org/~carpet/git/">development - source tree</a> in your web browser.</p> ---> -<!-- Should we also allow download via wget? --> + <h3>Development Version</h3> + + <p>The development version of Carpet is available + via <a href="http://mercurial.selenic.com/">Mercurial</a>:</p> +<pre> cd Cactus + hg clone https://carpetcode.googlecode.com/hg/ carpetcode + cd arrangements + ln -s ../carpetcode/Carpet* .</pre> + <p>(Don't miss the dot after the <code>Carpet*</code> in the last + line.) Instructions for using Mercurial are + given <a href="#hg">below</a>.</p> <h2>Write Access</h2> +<!-- darcs is outdated <h3>Darcs Repositories</h3> <p>Write access to Carpet darcs repositories is handled via ssh. @@ -141,42 +148,42 @@ cd arrangements ln -s ../carpet-stable-3/Carpet* .</pre> <p>(Don't miss the dot after the <code>Carpet*</code> in the last - line.) Further instructions for using darcs are + line.) Further instructions for using darcs are given <a href="#darcs">below</a>.</p> <p>You can also obtain the darcs repository using <tt>rsync</tt> - instead of <tt>darcs</tt>. For this, use the command</p> + instead of <tt>darcs</tt>. For this, use the command</p> <pre> rsync -Paz darcs@cvs.carpetcode.org:carpet-stable-3 .</pre> <p>This copies the darcs repository into a subdirectory called <tt>carpet-stable-3</tt>, in much the same way as - the <tt>darcs get</tt> command above would. That is, you also end + the <tt>darcs get</tt> command above would. That is, you also end up with a fully functional local repository.</p> <p>We thank the <a href="http://www.tat.physik.cct.lsu.edu/">Institut für Astronomie und Astrophysik</a> of the Universität Tübingen for hosting the CVS and darcs servers.</p> +--> <h3>Git Repository</h3> - <p>Write access to Carpet git repositories is also handled via - ssh. Once you have an account set up, you obtain e.g. the - development version with</p> + <p>Write access to Carpet git repositories is handled via + ssh. Once you have an account set up, you obtain Carpet via</p> <pre> cd Cactus git clone carpetgit@carpetcode.org:carpet.git cd arrangements ln -s ../carpet/Carpet* .</pre> <p>(Don't miss the dot after the <code>Carpet*</code> in the last - line.) Further instructions for using git are + line.) Further instructions for using git are given <a href="#git">below</a>.</p> <!-- rsync access is not yet set up <p>You can also obtain the darcs repository using <tt>rsync</tt> - instead of <tt>darcs</tt>. For this, use the command</p> + instead of <tt>darcs</tt>. For this, use the command</p> <pre> rsync -Paz darcs@cvs.carpetcode.org:carpet-stable-3 .</pre> <p>This copies the darcs repository into a subdirectory called <tt>carpet-stable-3</tt>, in much the same way as - the <tt>darcs get</tt> command above would. That is, you also end + the <tt>darcs get</tt> command above would. That is, you also end up with a fully functional local repository.</p> --> @@ -184,16 +191,36 @@ of <a href="http://www.tapir.caltech.edu/~cott/">Christian D. Ott</a>.</p> + <h3>Mercurial Repository</h3> + + <p>We use two different Carpet Mercurial repositories for code + development that we keep manually in sync. The Google Code + repository is mainly used to proved public read-only access; + development itself occurs at <code>carpetcode.dyndns.org</code>. + Access to this Carpet repository is handled via ssh: +<pre> cd Cactus + hg glone ssh://carpetmercurial@carpetcode.dyndns.org/carpet + cd arrangements + ln -s ../carpet/Carpet* .</pre> + <p>(Don't miss the dot after the <code>Carpet*</code> in the last + line.) Further instructions for using git are + given <a href="#hgt">below</a>.</p> + + <p>The Carpet Mercurial server is also a courtesy + of <a href="http://www.tapir.caltech.edu/~cott/">Christian + D. Ott</a>.</p> + <hr /> <h2>Modern Version Control Systems</h2> - <p>Carpet is managed in <a href="http://darcs.net/">darcs</a> and - <a href="http://git.or.cz/">git</a> repositories instead of a CVS - repository. Darcs and git have a number of advantages over CVS - for developers, such as:</p> + <p>Carpet is managed in <a href="http://git.or.cz/">git</a> + and <a href="http://mercurial.selenic.com/">Mercurial</a> + repositories instead of CVS or SVN repositories. Git and Mercurial + have a number of advantages over CVS and SVN for developers, such + as:</p> <ul> <li>You have a local copy of the repository, and can therefore @@ -202,18 +229,19 @@ so that you can omit dangerous changes, or keep changes to yourself until you are ready to publish them</li> <li>You can undo all changes</li> - <li>You can easily rename files and directories</li> <li>You can work in a decentralised manner, which suits large collaborations which may want to avoid a central point of control</li> </ul> - <p>and then some more, as described in - the <a href="http://darcs.net/manual/">darcs manual</a> and - the <a href="http://git.or.cz/">git web pages</a>.</p> + <p>and then some more, as described on + the <a href="http://git.or.cz/">git web pages</a> and + the <a href="http://mercurial.selenic.com/">Mercurial web + pages</a>.</p> +<!-- <h3 id="darcs">Using Darcs</h3> <p>The foremost source of information about darcs is @@ -226,39 +254,40 @@ Asked Questions</a>.</p> <p>If darcs is not already installed on your system, you need to - do so yourself. This is described on + do so yourself. This is described on the <a href="http://darcs.net/">darcs home page</a>, and some links to binaries are given in the darcs wiki.</p> <h3>Updating the Repository from the Master Repository</h3> <p>At some time you will want to update your version of Carpet and - incorporate some changes from the main Carpet repository. You do + incorporate some changes from the main Carpet repository. You do this with the command</p> <pre>cd Cactus/carpet-stable-3 darcs pull</pre> <p>which will look for new changes, and then ask you which of these - you want to obtain. Normally, you will want all changes.</p> + you want to obtain. Normally, you will want all changes.</p> <h3>Working with Darcs</h3> <p>We have some instructions on how to <a href="work-with-darcs.html">develop Carpet with darcs</a>.</p> +--> <h3 id="git">Using Git</h3> <p>The <a href="http://git.or.cz/">git web site</a> contains - introductions and documentation for git. The Linux kernel + introductions and documentation for git. The Linux kernel developers also maintain a <a href="http://www.kernel.org/pub/software/scm/git/docs/gittutorial.html">tutorial</a> for - git. Git should be available for all modern operating systems. + git. Git should be available for all modern operating systems. It is also not difficult to install manually.</p> <p>Git comes with a convenient graphical user interface - called <code>git-gui</code>. It allows you to update your code + called <code>git-gui</code>. It allows you to update your code from the master, commit local changes, compare branches, or push local changes back to the master repository.</p> @@ -268,35 +297,62 @@ darcs pull</pre> <h3>Updating the Repository from the Master Repository</h3> <p>At some time you will want to update your version of Carpet and - incorporate some changes from the main Carpet repository. If you + incorporate some changes from the main Carpet repository. If you are not using the graphical user interface, then you do this with the command</p> <pre>cd Cactus/carpet git pull</pre> - <p>which will download and merge the current version. Git will + <p>which will download and merge the current version. Git will refuse to overwrite any conflicting local changes that you may have.</p> + <h3 id="hg">Using Mercurial</h3> + + <p>The <a href="http://mercurial.selenic.com/">Mercurial web + site</a> contains introductions and documentation for Mercurial. + Mercurial should be available for all modern operating systems. It + is also not difficult to install manually.</p> + + <p>On Mac operating systems, Mercurial comes with a convenient + graphical user interface called <code>MacHG</code>. This allows + you to update your code from the master, commit local changes, + compare branches, or push local changes back to the master + repository.</p> + + <h3>Updating the Repository from the Master Repository</h3> + + <p>At some time you will want to update your version of Carpet and + incorporate some changes from the main Carpet repository. If you + are not using the graphical user interface, then you use the + command</p> +<pre>cd Cactus/carpet +hg pull -u</pre> + <p>which will download and merge the current version. Mercurial + will refuse to overwrite any conflicting local changes that you + may have.</p> + + + <h3>Convenient SSH Key Management</h3> <p>SSH has two mechanisms for authentication, typing a password, - or using ssh keys. When you use ssh keys, your private key is - (<em>should</em>) be protected by a password. That means that you + or using ssh keys. When you use ssh keys, your private key is + (<em>should</em>) be protected by a password. That means that you have to type this password every time you log into a different - machine. Some people protect their private ssh key with an empty + machine. Some people protect their private ssh key with an empty password --- in this way, they don't have to type a password, but - this is not very secure. If somebody is able to copy the private - ssh key, they have access to your remote accounts. Intruders can - use this hop from one machine to the next. Please do not use + this is not very secure. If somebody is able to copy the private + ssh key, they have access to your remote accounts. Intruders can + use this hop from one machine to the next. Please do not use empty passwords on your ssh keys.</p> - <p>SSH-agent is a convenient way to make things safe. It is an + <p>SSH-agent is a convenient way to make things safe. It is an agent that starts automatically when you log in, and asks you for - your ssh key password once. It remembers this password in memory, + your ssh key password once. It remembers this password in memory, and whenever you use ssh to log into a remote account, ssh - contacts the ssh-agent for the password to the key. If this + contacts the ssh-agent for the password to the key. If this password is accepted, you don't have to type anything.</p> <p>I use the following lines in my <code>.bash_profile</code> to @@ -304,7 +360,7 @@ git pull</pre> <pre>keychain id_dsa test -f $HOME/.keychain/$(hostname)-sh && source $HOME/.keychain/$(hostname)-sh > /dev/null </pre> - <p>Keychain starts the ssh-agent. Keychain can also handle gpg + <p>Keychain starts the ssh-agent. Keychain can also handle gpg key passwords for your encrypted and/or signed emails.</p> @@ -325,7 +381,7 @@ test -f $HOME/.keychain/$(hostname)-sh && source $HOME/.keychain/$(hostn <address><a href="mailto:schnetter@cct.lsu.edu">Erik Schnetter</a></address> <!-- Created: Tue Sep 28 16:52:20 CEST 2004 --> <!-- hhmts start --> -Last modified: Sat Mar 01 2008 +Last modified: Tue Feb 15 2011 <!-- hhmts end --> </body> </html> diff --git a/Carpet/CarpetWeb/index.html b/Carpet/CarpetWeb/index.html index 00cf19354..3912d095d 100644 --- a/Carpet/CarpetWeb/index.html +++ b/Carpet/CarpetWeb/index.html @@ -33,14 +33,14 @@ <a href="humour.html">Other Carpets</a></p> <p><b>Mailing Lists</b><br /> -<a href="http://lists.carpetcode.org/listinfo/developers/">Subscribe</a><br /> -<a href="http://lists.carpetcode.org/archives/developers/">List Archive</a><br /> -<a href="http://lists.carpetcode.org/listinfo/carpet-cvs/">CVS messages</a><br /> -<a href="http://lists.carpetcode.org/listinfo/carpet-darcs/">darcs/git messages</a></p> +<a href="http://cactuscode.org/community/mailinglists/">List Management</a><br /> +<a href="http://cactuscode.org/mailman/listinfo/developers">Subscribe</a><br /> +<a href="http://cactuscode.org/pipermail/developers/">List Archive</a><br /> +</p> <p><b>Development</b><br /> <a href="get-carpet.html">Download</a><br /> -<a href="http://bugs.carpetcode.org/">Bug Reports</a><br /> +<a href="http://trac.einsteintoolkit.org/">Bug Reports</a><br /> <a href="contributors.html">Contributors</a></p> <!-- These are outdated <a href="feature-requests.html">Missing features</a><br /> @@ -121,127 +121,35 @@ Barcelona<br /> <h2>News</h2> - <table><tr><td valign="top"> - <p><b>March 30, 2009:</b> We have ported Carpet to - the <a href="http://www-03.ibm.com/systems/deepcomputing/bluegene/">BlueGene/P</a> - architecture, using - the <a href="http://www.alcf.anl.gov/resources/storage.php">Surveyor</a> - system at the <a href="http://www.alcf.anl.gov/">ALCF</a>. The - graph to the right shows preliminary performance and scaling - results, comparing different compilers and options - (<a href="http://gcc.gnu.org/">gcc</a>, <a href="http://www.ibm.com/software/awdtools/xlcpp/">IBM's - XL compilers</a> without OpenMP, and IBM's XL compilers - with <a href="http://www.openmp.org/">OpenMP</a>, which required - reducing the optimisation level). For these benchmarks, the - problem size was reduced to about one eighth of the standard - size, using 13<sup>3</sup> grid points per core. The results - show that Carpet scales fine up to the size of the total machine - (4k cores), but further work on compiler options is - required.</p> - </td><td valign="top"> - <p><a href="scaling-surveyor/results-surveyor.pdf"><img - src="scaling-surveyor/results-surveyor.png" - width="180" alt="AMR benchmark results" /></a></p> - </td></tr></table> - - <table><tr><td valign="top"> - <p><b>March 20, 2009:</b> Carpet can now perform <i>performance - experiments</i> by artificially increasing the size or the - number of MPI messages exchanged between processes. This can - help determine whether either the communication bandwidth or the - communication latency are a bottleneck of a particular - simulation. The figure to the right shows results for the - standard <a href="http://www.cct.lsu.edu/~eschnett/McLachlan/">McLachlan</a> - AMR benchmark run on - the <a href="http://en.wikipedia.org/wiki/Cray_XT4">Cray XT4</a> - <a href="http://www.nics.tennessee.edu/computing-resources/kraken">Kraken</a>, using 25<sup>3</sup> grid points per core. These - results indicate that the additional latency from increasing the - number of messages has no significant effect, and hence the - benchmark is bandwidth limited for this problem size.</p> - </td><td valign="top"> - <p><a href="scaling-whatif/results-whatif.pdf"><img - src="scaling-whatif/results-whatif.png" - width="180" alt="AMR benchmark results" /></a></p> - </td></tr></table> - - <table><tr><td valign="top"> - <p><b>March 16, 2009:</b> Erik Schnetter and Steve Brandt - published a white - paper <a href="http://www.cct.lsu.edu/CCT-TR/CCT-TR-2009-4"><i>Relativistic - Astrophysics on the SiCortex Architecture</i></a>. This paper - expands on a - <a href="http://www.sicortex.com/news_events/campaigns/lsu_webinar">webinar</a> - by Erik and Steve that was hosted - by <a href="http://www.sicortex.com/">SiCortex</a>.</p> - <p>The graph at the right shows Carpet's parallel scalability - using - the <a href="http://www.cct.lsu.edu/~eschnett/McLachlan/">McLachlan</a> - code with nine levels of AMR for a set of current HPC systems. - The results have been rescaled to the architectures' theoretical - single-core peak performance. This makes it possible to compare - Carpet's scalability on different architectures. (It is not - possible to compare the systems' absolute performance in this - figure.)</p> - </td><td valign="top"> - <p><a href="sicortex/results-scaled.pdf"><img - src="sicortex/results-scaled.png" - width="180" alt="AMR benchmark results" /></a></p> - </td></tr></table> - - <table><tr><td valign="top"> - <p><b>November 9, 2008:</b> In the context of - the <a href="http://www.cct.lsu.edu/xirel/">XiRel project</a>, - we re-designed Carpet's communication layer to avoid many - operations that had a cost of O(<var>N</var>), growing linearly - with the number of MPI processes. Such costs are generally not - acceptable when running on several thousand cores, and have to - be reduced e.g. to O(log <var>N</var>). Carpet now stores the - communication schedule (mostly) in a distributed manner, - increasing performance and reducing its memory requirement. - These improvements are currently being tested; preliminary - scaling results are shown in the figure to the right.</p> - </td><td valign="top"> - <p><a href="scaling-improved/results-best.pdf"><img - src="scaling-improved/results-best.png" - width="180" alt="AMR benchmark results" /></a></p> - </td></tr></table> - - <p><b>June 25, 2008:</b> We are happy to announce - the <a href="http://www.cct.lsu.edu/~eschnett/SimFactory"><i>Simulation - Factory</i></a>, a tool to help access remote HPC systems, - manage source trees, and submit and control simulations. The - Simulation Factory contains a set of abstractions of the tasks - which are necessary to set up and successfully finish numerical - simulations using the Cactus framework. These abstractions hide - tedious low-level management tasks, they capture "best - practices" of experienced users, and they create a log trail - ensuring repeatable and well-documented scientific results. - Using these abstractions, many types of potentially disastrous - user errors are avoided, and different supercomputers can be - used in a uniform manner.</p> - - <table><tr><td valign="top"> - <p><b>March 29, 2008:</b> We have benchmarked McLachlan, a new - BSSN-type vacuum Einstein code, using Carpet for unigrid and AMR - calculations. We compare several current large machines: - <a href="http://www.nersc.gov/nusers/systems/franklin/">Franklin</a> - (NERSC), <a href="http://www.loni.org/systems/system.php?system=QueenBee">Queen - Bee</a> (LONI), - and <a href="http://www.tacc.utexas.edu/services/userguides/ranger/">Ranger</a> - (TACC). - <!-- These machines have different architectures and - interconnects.--></p> - </td><td valign="top"> - <p><a - href="scaling-amr/results-carpet-1lev.pdf"><img - src="scaling-amr/results-carpet-1lev.png" width="180" - alt="Unigrid benchmark results" /></a></p> - </td><td valign="top"> - <p><a - href="scaling-amr/results-carpet-9lev.pdf"><img - src="scaling-amr/results-carpet-9lev.png" width="180" - alt="AMR benchmark results" /></a></p> - </td></tr></table> + <p><b>February 15, 2011:</b> + The <a href="get-carpet.html">download instructions</a> for Carpet + now also point to <a href="http://code.google.com/">Google + Code</a>, where the current development version is availble for + download.</p> + + <p><b>November 23, 2010:</b> We are pleased to announce the second + release (code name + "<a href="http://en.wikipedia.org/wiki/Subrahmanyan_Chandrasekhar">Chandrasekhar</a>") + of the Einstein Toolkit, an open, community developed software + infrastructure for relativistic astrophysics. This release is + mainly a maintenance release incorporating fixes accumulated since + the previous release in June 2010, as well as additional test + suites.</p> + + <p><b>August 30, 2011:</b> Notes from our + <a href="http://ccrg.rit.edu/~carpet/wiki/Main_Page">Carpet + Developer Workshop</a> at RIT are now available.</p> + + <p><b>June 17, 2010:</b> We are pleased to announce the first + release (code name + "<a href="http://en.wikipedia.org/wiki/Bohr-Einstein_debates">Bohr</a>") + of the Einstein Toolkit, an open, community developed software + infrastructure for relativistic astrophysics. The Einstein Toolkit + is a collection of over 130 software components and tools for + simulating and analyzing general relativistic astrophysical + systems that builds on numerous software efforts in the numerical + relativity community including CactusEinstein, the Whisky + hydrodynamics code, and the Carpet AMR infrastructure.</p> <p><a href="olds.html"><b>Old News...</b></a></p> @@ -291,6 +199,7 @@ Barcelona<br /> <h2>Interacting with the developers</h2> +<!-- outdated <p>Most discussions about Carpet, i.e. user questions, feature requests, and bug reports, are held on the Carpet developers' mailing list <a @@ -300,7 +209,17 @@ Barcelona<br /> You will also find the mailing list archive there. We thank <a href="http://www.tat.physik.uni-tuebingen.de/~kobras/">Daniel Kobras</a> for managing the mailing list server.</p> - +--> + <p>Most discussions about Carpet, i.e. user questions, feature + requests, and bug reports, are held on the Cactus developers' + mailing + list <a href="mailto:developers@cactuscode.org">developers@cactuscode.org</a>. + You can subscribe and unsubscribe from + our <a href="http://cactuscode.org/community/mailinglists/">list + management web page</a>. You will also find the mailing list + archive there.</p> + +<!-- outdated <p>We have started to use <a href="http://www.bugzilla.org/">Bugzilla</a> to keep track of requested features or reported bugs in Carpet. You can submit or @@ -309,6 +228,12 @@ Barcelona<br /> have created an account there. The old <a href="feature-requests.html">list of missing features</a> have not yet been moved over to Bugzilla.</p> +--> + + <p>We use <a href="http://trac.edgewall.org/">TRAC</a> to keep + track of requested features or reported bugs in Carpet. You can + submit or comment on issues from + our <a href="http://trac.einsteintoolkit.org/">TRAC site</a>.</p> <hr /> @@ -493,7 +418,7 @@ Barcelona<br /> <p> <!-- Created: Tue Aug 12 12:12:08 CEST 2003 --> <!-- hhmts start --> -Last modified: Mon Mar 30 2009 +Last modified: Tue Feb 15 2011 <!-- hhmts end --> </p> diff --git a/Carpet/CarpetWeb/olds.html b/Carpet/CarpetWeb/olds.html index 1d29d2cc8..0b6f5e474 100644 --- a/Carpet/CarpetWeb/olds.html +++ b/Carpet/CarpetWeb/olds.html @@ -15,6 +15,130 @@ <p><a href="index.html"><b>New News...</b></a></p> <table><tr><td valign="top"> + <p><b>March 30, 2009:</b> We have ported Carpet to + the <a href="http://www-03.ibm.com/systems/deepcomputing/bluegene/">BlueGene/P</a> + architecture, using + the <a href="http://www.alcf.anl.gov/resources/storage.php">Surveyor</a> + system at the <a href="http://www.alcf.anl.gov/">ALCF</a>. The + graph to the right shows preliminary performance and scaling + results, comparing different compilers and options + (<a href="http://gcc.gnu.org/">gcc</a>, <a href="http://www.ibm.com/software/awdtools/xlcpp/">IBM's + XL compilers</a> without OpenMP, and IBM's XL compilers + with <a href="http://www.openmp.org/">OpenMP</a>, which required + reducing the optimisation level). For these benchmarks, the + problem size was reduced to about one eighth of the standard + size, using 13<sup>3</sup> grid points per core. The results + show that Carpet scales fine up to the size of the total machine + (4k cores), but further work on compiler options is + required.</p> + </td><td valign="top"> + <p><a href="scaling-surveyor/results-surveyor.pdf"><img + src="scaling-surveyor/results-surveyor.png" + width="180" alt="AMR benchmark results" /></a></p> + </td></tr></table> + + <table><tr><td valign="top"> + <p><b>March 20, 2009:</b> Carpet can now perform <i>performance + experiments</i> by artificially increasing the size or the + number of MPI messages exchanged between processes. This can + help determine whether either the communication bandwidth or the + communication latency are a bottleneck of a particular + simulation. The figure to the right shows results for the + standard <a href="http://www.cct.lsu.edu/~eschnett/McLachlan/">McLachlan</a> + AMR benchmark run on + the <a href="http://en.wikipedia.org/wiki/Cray_XT4">Cray XT4</a> + <a href="http://www.nics.tennessee.edu/computing-resources/kraken">Kraken</a>, using 25<sup>3</sup> grid points per core. These + results indicate that the additional latency from increasing the + number of messages has no significant effect, and hence the + benchmark is bandwidth limited for this problem size.</p> + </td><td valign="top"> + <p><a href="scaling-whatif/results-whatif.pdf"><img + src="scaling-whatif/results-whatif.png" + width="180" alt="AMR benchmark results" /></a></p> + </td></tr></table> + + <table><tr><td valign="top"> + <p><b>March 16, 2009:</b> Erik Schnetter and Steve Brandt + published a white + paper <a href="http://www.cct.lsu.edu/CCT-TR/CCT-TR-2009-4"><i>Relativistic + Astrophysics on the SiCortex Architecture</i></a>. This paper + expands on a + <a href="http://www.sicortex.com/news_events/campaigns/lsu_webinar">webinar</a> + by Erik and Steve that was hosted + by <a href="http://www.sicortex.com/">SiCortex</a>.</p> + <p>The graph at the right shows Carpet's parallel scalability + using + the <a href="http://www.cct.lsu.edu/~eschnett/McLachlan/">McLachlan</a> + code with nine levels of AMR for a set of current HPC systems. + The results have been rescaled to the architectures' theoretical + single-core peak performance. This makes it possible to compare + Carpet's scalability on different architectures. (It is not + possible to compare the systems' absolute performance in this + figure.)</p> + </td><td valign="top"> + <p><a href="sicortex/results-scaled.pdf"><img + src="sicortex/results-scaled.png" + width="180" alt="AMR benchmark results" /></a></p> + </td></tr></table> + + <hr /> + + <table><tr><td valign="top"> + <p><b>November 9, 2008:</b> In the context of + the <a href="http://www.cct.lsu.edu/xirel/">XiRel project</a>, + we re-designed Carpet's communication layer to avoid many + operations that had a cost of O(<var>N</var>), growing linearly + with the number of MPI processes. Such costs are generally not + acceptable when running on several thousand cores, and have to + be reduced e.g. to O(log <var>N</var>). Carpet now stores the + communication schedule (mostly) in a distributed manner, + increasing performance and reducing its memory requirement. + These improvements are currently being tested; preliminary + scaling results are shown in the figure to the right.</p> + </td><td valign="top"> + <p><a href="scaling-improved/results-best.pdf"><img + src="scaling-improved/results-best.png" + width="180" alt="AMR benchmark results" /></a></p> + </td></tr></table> + + <p><b>June 25, 2008:</b> We are happy to announce + the <a href="http://www.cct.lsu.edu/~eschnett/SimFactory"><i>Simulation + Factory</i></a>, a tool to help access remote HPC systems, + manage source trees, and submit and control simulations. The + Simulation Factory contains a set of abstractions of the tasks + which are necessary to set up and successfully finish numerical + simulations using the Cactus framework. These abstractions hide + tedious low-level management tasks, they capture "best + practices" of experienced users, and they create a log trail + ensuring repeatable and well-documented scientific results. + Using these abstractions, many types of potentially disastrous + user errors are avoided, and different supercomputers can be + used in a uniform manner.</p> + + <table><tr><td valign="top"> + <p><b>March 29, 2008:</b> We have benchmarked McLachlan, a new + BSSN-type vacuum Einstein code, using Carpet for unigrid and AMR + calculations. We compare several current large machines: + <a href="http://www.nersc.gov/nusers/systems/franklin/">Franklin</a> + (NERSC), <a href="http://www.loni.org/systems/system.php?system=QueenBee">Queen + Bee</a> (LONI), + and <a href="http://www.tacc.utexas.edu/services/userguides/ranger/">Ranger</a> + (TACC). + <!-- These machines have different architectures and + interconnects.--></p> + </td><td valign="top"> + <p><a + href="scaling-amr/results-carpet-1lev.pdf"><img + src="scaling-amr/results-carpet-1lev.png" width="180" + alt="Unigrid benchmark results" /></a></p> + </td><td valign="top"> + <p><a + href="scaling-amr/results-carpet-9lev.pdf"><img + src="scaling-amr/results-carpet-9lev.png" width="180" + alt="AMR benchmark results" /></a></p> + </td></tr></table> + + <table><tr><td valign="top"> <p><b>March 1, 2008:</b> Carpet has a logo! This logo is a <a href="http://en.wikipedia.org/wiki/Sierpinski_carpet">Sierpiński carpet</a>, which is a fractal pattern with @@ -417,7 +541,7 @@ <p> <!-- Created: Tue Aug 12 12:12:08 CEST 2003 --> <!-- hhmts start --> -Last modified: Sat Mar 01 2008 +Last modified: Feb 15 2011 <!-- hhmts end --> </p> |