Commit d988363f authored by AUMAGE Olivier's avatar AUMAGE Olivier

- rafraichissement du style de la page StarPU

git-svn-id: svn+ssh://scm.gforge.inria.fr/svn/starpu/website@7701 176f6dd6-97d6-42f4-bd05-d3db9ad07c7a
parent 6ce4a01d
......@@ -9,11 +9,13 @@
<body>
<div class="title">
<h1><a href="./">StarPU</a></h1>
<h1 class="sub">A Unified Runtime System for Heterogeneous Multicore Architectures</h1>
<h2>A Unified Runtime System for Heterogeneous Multicore Architectures</h2>
</div>
<div class="menu"> <hr class="menu"/>
<a href="/Runtime/">RUNTIME homepage</a> |
<div class="menu">
<a href="/Runtime/">RUNTIME TEAM</a> |
<a href="#news">News</a> |
<a href="#summary">Summary</a> |
<a href="#features">Features</a> |
......@@ -23,10 +25,40 @@
<a href="tutorials">Tutorials</a> |
<a href="#contact">Contact</a> |
<a href="https://wiki.bordeaux.inria.fr/runtime/doku.php?id=starpu">Intranet</a>
<hr class="menu"/>
</div>
<div class="section" id="news">
<div class="section" id="summary">
<h3>Description</h3>
<p>
<span class="important">StarPU is a task programming library for hybrid architectures</span>
<ol>
<li><b>The application provides algorithms and constraints</b>
<ul>
<li>CPU/GPU implementations of tasks</li>
<li>A graph of tasks, using either the StarPU's high level <b>GCC plugin</b> pragmas or StarPU's rich <b>C API</b></li>
</ul>
<br>
</li>
<li><b>StarPU handles run-time concerns</b>
<ul>
<li>Task dependencies</li>
<li>Optimized heterogeneous scheduling</li>
<li>Optimized data transfers and replication between main memory and discrete memories</li>
<li>Optimized cluster communications</li>
</ul>
</li>
</ol>
</p>
<p>
<span class="important">Rather than handling low-level issues, <b>programmers can concentrate on algorithmic concerns!</b></span>
</p>
<p>
<span class="note">The StarPU documentation is available in <a href="./starpu.pdf">PDF</a> and in <a href="./starpu.html">HTML</a>.</span> Please note that these documents are up-to-date with the latest release of StarPU.
</p>
</div>
<div class="section emphasize newslist" id="news">
<h3>News</h3>
<p>
October 2012 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1570"><b>The
......@@ -50,60 +82,15 @@ August 2012 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1
July 2012 <b>&raquo;&nbsp;</b> StarPU was presented at the <a href="http://gcc.gnu.org/wiki/cauldron2012#StarPU.27s_C_Extensions_for_Hybrid_CPU.2BAC8-GPU_Task_Programming.2C_or.2C_An_Experience_in_Turning_a_Clumsy_API_Into_Language_Extensions">GNU Tools Cauldron 2012</a>
</p>
</div>
<div style="text-align: right; font-style: italic;">
<div class="section emphasizebot" style="text-align: right; font-style: italic;">
Get the latest StarPU news by subscribing to the <a href="http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/starpu-announce">starpu-announce mailing list</a>.
See also the <a href="news/">news archive</a>.
</div>
<hr class="main"/>
<div class="section" id="summary">
<h3>Summary</h3>
<p>
StarPU is a task programming library for hybrid architectures (CPU,
GPU, ...).
</p>
<P>
<b>An application using StarPU needs to provide</b>
<ul>
<li>
a description of the graph of tasks using either the <b>StarPU's high
level GCC plugin</b> pragmas or <b>StarPU's rich C API</b>,
</li>
<li>
and the CPU/GPU implementations for each of the tasks.
</li>
</ul>
</p>
<P>
<b>Then StarPU handles the whole execution.</b> It handles the
numerous run-time challenging issues introduced by heterogeneous
architectures at all levels:
<ul>
<li> Task dependencies and optimized scheduling,
</li>
<li> Optimized data transfers between main memory and e.g. GPU
memory,</li>
<li>
Optimized cluster communications.
</li>
</ul>
</p>
<p>
<b>Rather than handling low-level issues, programmers can concentrate
on algorithmic concerns.</b>
</p>
<div style="text-align: right; font-style: italic;">
<p>
The StarPU documentation is available in <a href="./starpu.pdf">PDF</a> and in <a href="./starpu.html">HTML</a>. Please note that these documents are up-to-date with the latest release of StarPU.
</p>
</div>
</div>
<div class="section" id="contact">
<h3>Contact</h3>
For any questions regarding StarPU, please contact the StarPU developers mailing list.
<p>For any questions regarding StarPU, please contact the StarPU developers mailing list.</p>
<pre>
<a href="mailto:starpu-devel@lists.gforge.inria.fr?subject=StarPU">starpu-devel@lists.gforge.inria.fr</a>
</pre>
......@@ -145,6 +132,8 @@ programmer with best flexibility:
tasks (thus including tasks which have not been created yet),</li>
<li>implicitly from RAW, WAW, and WAR data dependencies.</li>
</ul>
</p>
<p>
StarPU also supports an OpenMP-like <a href="http://runtime.bordeaux.inria.fr/StarPU/starpu.html#Data-reduction">reduction</a> access mode.
</p>
......@@ -237,7 +226,7 @@ corresponding processing unit is busy (resp. idle). The number of ready tasks
is displayed in the curve on top: it appears that with this scheduling policy,
the algorithm suffers a certain lack of parallelism. <b>Measured speed: 175.32
GFlop/s</b>
<a href="./images/greedy-lu-16k-fx5800.png"> <img src="./images/greedy-lu-16k-fx5800.png" alt="LU decomposition (greedy)" width="50%"></a>
<center><a href="./images/greedy-lu-16k-fx5800.png"> <img src="./images/greedy-lu-16k-fx5800.png" alt="LU decomposition (greedy)" width="75%"></a></center>
</p>
<p>
......@@ -247,7 +236,7 @@ performance models and to keep data locality as high as possible. In this
example, the Pajé trace clearly shows that this scheduling strategy outperforms
the previous one in terms of processor usage. <b>Measured speed: 239.60
GFlop/s</b>
<a href="./images/dmda-lu-16k-fx5800.png"><img src="./images/dmda-lu-16k-fx5800.png" alt="LU decomposition (dmda)" width="50%"></a>
<center><a href="./images/dmda-lu-16k-fx5800.png"><img src="./images/dmda-lu-16k-fx5800.png" alt="LU decomposition (dmda)" width="75%"></a></center>
</p>
</div>
......@@ -400,7 +389,6 @@ E. Agullo, C. Augonnet, J. Dongarra, H. Ltaief, R. Namyst, S. Thibault, and S. T
In <em>GPU Computing Gems, volume 2.</em>, September 2010.
<br/>
Available <a href="http://hal.inria.fr/inria-00547847">here</a>.
<ol>
<li>
E. Agullo, C. Augonnet, J. Dongarra, M. Faverge, H. Ltaief, S. Thibault, and S. Tomov.
<br/>
......@@ -426,16 +414,14 @@ In <em>9th ACS/IEEE International Conference on Computer Systems and Application
Available <a href="http://hal.inria.fr/hal-00654193">here</a>
</li>
</ol>
</li>
</ol>
</div>
<hr class="main" />
<div class="section bot">
<p class="updated">
Last updated on 2012/10/03.
</p>
</div>
</body>
</html>
......@@ -9,22 +9,26 @@
<body>
<div class="title">
<h1><a href="./">StarPU</a></h1>
<h1 class="sub">Internships</h1>
<h2>Internships</h2>
</div>
<hr class="main" />
<div>
<div class="menu">
<a href="/Runtime/">RUNTIME TEAM</a> |
<a href="..">Back to the main page</a>
</div>
<hr class="main" />
</div>
<div class="section" id="internships">
<h3>Internship Opportunities</h3>
<div class="section emphasize">
<h3>On-going Internship Opportunities</h3>
There is currently no available internships.
</div>
<div class="section" id="pastinternships">
<h3>2011 M2 Internships</h3>
<div class="section">
<h3>Former Internships</h3>
<div class="section">
<h4>2011 M2 Internships</h4>
<p>
Extending StarPU:
<ul>
......@@ -41,11 +45,12 @@ Using StarPU:
</p>
</div>
<hr class="main" />
<div class="section bot">
<p class="updated">
Last updated on 2012/10/03.
</p>
</div>
</div>
</body>
</html>
......
......@@ -9,20 +9,25 @@
<body>
<h1><a href="./">StarPU</a></h1>
<h1 class="sub">News Archive</h1>
<div class="title">
<h1><a href="../">StarPU</a></h1>
<h2>News Archive</h2>
</div>
<hr class="main" />
<div>
<div class="menu">
<a href="/Runtime/">RUNTIME TEAM</a> |
<a href="..">Back to the main page</a>
</div>
<hr class="main" />
</div>
<div style="text-align: right; font-style: italic;">
<div class="section emphasize">
<h3>News Mailing List</h3>
Get the latest StarPU news by subscribing to the <a href="http://lists.gforge.inria.fr/cgi-bin/mailman/listinfo/starpu-announce">starpu-announce mailing list</a>.
</div>
<div class="section" id="news">
<div class="section">
<h3>News</h3>
<div class="section newslist">
<h4>2012</h4>
<p>
May 2012 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1570"><b>The
v1.0.1 release of StarPU is now available!</b></a>. This release
......@@ -41,12 +46,20 @@ See the announcements
on <a href="http://lwn.net/Articles/489337/">LWN</a>, <a href="http://article.gmane.org/gmane.comp.gcc.devel/125551">
the GCC mailing list</a>, and <a href="http://www.phoronix.com/scan.php?page=news_item&px=MTA4MDI">Phoronix</a>.
</p>
</div>
<div class="section newslist">
<h4>2011</h4>
<p>
May 2011 <b>&raquo;&nbsp;</b> <a href="http://gforge.inria.fr/frs/?group_id=1570"><b>StarPU 0.9.1 is now available !</b></a>
This release provides a reduction mode, an external API for schedulers, theoretical bounds, power-based optimization, parallel
tasks, an MPI DSM, profiling interfaces, an initial support for CUDA4 (GPU-GPU
transfers), improved documentation and of course various fixes.
</p>
</div>
<div class="section newslist">
<h4>2010</h4>
<p>
September 2010 <b>&raquo;&nbsp;</b> Discover how we ported the MAGMA and the PLASMA libraries on top of StarPU in collaboration with ICL/UTK in <a href="http://www.netlib.org/lapack/lawnspdf/lawn230.pdf"><b>this Lapack Working Note</b></a>.
</p>
......@@ -62,6 +75,10 @@ July 2010 <b>&raquo;&nbsp;</b> StarPU was presented during a tutorial entitled "
<p>
May 2010 <b>&raquo;&nbsp;</b> Want to get an overview of StarPU ? Check out our <a href="http://hal.archives-ouvertes.fr/inria-00467677">latest research report</a>!
</p>
</div>
<div class="section newslist">
<h4>2009</h4>
<!--<p>
October 2009 <b>&raquo;&nbsp;</b> <a href="http://gforge.inria.fr/frs/?group_id=1570"><b>StarPU 0.2.901 (0.3-rc1) is now available !</b></a>
This release adds support for asynchronous GPUs and heterogeneous multi-GPU platforms as well as many other improvements.
......@@ -70,13 +87,13 @@ This release adds support for asynchronous GPUs and heterogeneous multi-GPU plat
June 2009 <b>&raquo;&nbsp;</b>NVIDIA granted the StarPU team with a professor partnership and donated several high-end CUDA-capable cards.
</p>
</div>
</div>
<hr class="main" />
<div class="section bot">
<p class="updated">
Last updated on 2012/10/03.
</p>
</div>
</body>
</html>
body {
/* font-size: 15px;*/
/* margin-top: 0px; */
font-family: sans-serif;
color: rgb(0,0,0);
background: rgb(255,255,255);
}
div.menu {
text-align: center;
div.title {
padding: 1em;
margin-top: 12px;
margin-bottom: 3px;
background: #eeeeff;
font-variant: small-caps;
/* position: fixed;*/
width: 100%;
}
div.menu a {
text-decoration: none;
color: #0020a0;
}
div.menu hr.menu {
height: 4px;
background: #fe0;
border: 0px;
margin-top: 0px;
margin-bottom: 0px;
margin-bottom: 12px;
margin-left: auto;
margin-right: auto;
max-width: 80em;
}
h1 {
div.title h1 {
font: bold normal 2.5em sans-serif ;
margin: 0px;
color: #0020a0;
color: rgb(226,0,38);
border: none;
}
h1.sub {
div.title h1 a {
color: rgb(226,0,38);
text-decoration: none;
}
div.title h2 {
color: rgb(38,109,131);
font: bold normal 2em sans-serif ;
text-align: right ;
color: #0020a0;
border: none;
}
h1 a {
color: #0020a0;
div.menu {
text-align: center;
margin-top: 12px;
margin-bottom: 3px;
font-variant: small-caps;
padding-left: 1em;
padding-right: 1em;
padding-top: 1px;
padding-bottom: 1px;
margin-top: 12px;
margin-bottom: 12px;
margin-top:0px;
margin-left: auto;
margin-right: auto;
max-width: 80em;
background: rgb(91,132,152);
border-top: 4px solid rgb(204,209,222);
border-bottom: 4px solid rgb(204,209,222);
}
div.menu a {
text-decoration: none;
color: rgb(255,255,255);
}
h3 {
div.section h1 {
font: bold normal small-caps 2.5em sans-serif;
color: rgb(243,154,38);
}
div.section h2 {
font: bold normal small-caps 2em sans-serif ;
text-align: right;
color: rgb(230,68,57);
border-bottom: 4px solid rgb(28,70,114);
}
div.section h3 {
font: bold normal small-caps 1.5em sans-serif ;
color: #0020a0;
margin-top: 8px;
margin-bottom: 8px;
color: rgb(38,109,131);
border-bottom: 2px solid rgb(71,98,138);
}
h4 {
font: bold normal small-caps 1em sans-serif ;
color: #0020a0;
margin-top: 8px;
margin-bottom: 4px;
div.section h4 {
font: bold normal small-caps 1.2em sans-serif ;
color: rgb(133,146,66);
border-bottom: 1px solid rgb(112,130,164);
}
h6.mirrors {
......@@ -60,38 +83,55 @@ h6.mirrors {
}
div.section {
background: #eeeeff;
padding-left: 2px;
padding-bottom: 2px;
padding: 1em;
margin-top: 12px;
margin-bottom: 12px;
margin-left: auto;
margin-right: auto;
max-width: 80em;
}
div.section.emphasize {
background: rgb(255,245,219);
}
div.section.emphasizebot {
background: rgb(254,236,214);
margin-top: -12px;
}
div.section.bot {
border-top: 8px solid rgb(204,209,222);
}
p {
margin-top: 8px;
margin-bottom: 4px;
margin-left: 6px;
margin-right: 6px;
margin-left: 2em;
margin-right: 2em;
}
hr.main {
height: 8px;
background: #fe0;
border: 0px;
margin-top: 6px;
margin-bottom: 6px;
span.important {
font: bold normal 1.2em sans-serif;
color: rgb(230,68,57);
line-height: 2em;
}
span.note {
line-height: 2em;
background: rgb(238,243,217);
}
pre {
font-size: 12px;
background: #dddddd;
background: #f4fff8;
background: rgb(231,231,215);
padding: 3px;
padding-left: 0px;
padding-left: 1px;
margin-left: 12px;
}
a {
font-weight: bold;
color: rgb(28,70,114);
}
div.publis-desc {
......@@ -106,3 +146,9 @@ p.updated {
font-size: 10px;
font-style: italic;
}
div.section.newslist p+p {
margin-top 2px;
border-top: 1px solid rgb(255,225,138);
padding-top: 4px;
}
......@@ -9,18 +9,22 @@
<body>
<h1><a href="./">StarPU</a></h1>
<h1 class="sub">Hands-on session part 1: Task-based programming model</h1>
<div class="title">
<h1><a href="../../">StarPU</a></h1>
<h2>StarPU Tutorial - ComplexHPC Spring School - 2011</h2>
</div>
<a href="/Runtime/">RUNTIME homepage</a> |
<a href="../">StarPU homepage</a> |
<a href="/Publis/Keyword/StarPU.html">Publications</a>
<div class="menu">
<a href="/Runtime/">RUNTIME TEAM</a> |
<a href="../../">StarPU homepage</a> |
<a href="/Publis/Keyword/StarPU.html">Publications</a>
</div>
<hr class="main"/>
<div class="section">
<h3>Download & Install</h3>
<h2>Download & Install</h2>
<div class="section">
<h4>DAS-4 modules</h4>
<p>The first step is of course to download and install StarPU. Before doing so,
make sure to enable paths to the CUDA and CUBLAS environments on your machine,
......@@ -110,9 +114,11 @@ $ export STARPU_HOSTNAME=das4
a heterogeneous cluster, the cluster launcher script should set various
hostnames for the different node classes, as appropriate.</p>
</div>
</div>
<hr class="main"/>
<div class="section">
<h2>Hands-on session part 1: Task-based programming model</h2>
<div class="section">
<h3>Application example: vector scaling</h3>
......@@ -275,11 +281,11 @@ have been easily ported to StarPU by simply using <tt>starpu_insert_task</tt>.
matrix-matrix multiplication as an example. Try to run it with various numbers
of tasks</p>
</div>
</div>
<hr class="main" />
<h1 class="sub">Hands-on session part 2: Optimizations</h1>
<div class="section">
<h2>Hands-on session part 2: Optimizations</h2>
<p>This is based on StarPU's documentation
......@@ -491,23 +497,19 @@ nodes is decided by the <tt>my_distrib</tt> function, and can thus be changed
trivially.</p>
</div>
<hr class="main" />
<h1 class="sub">Contact</h1>
</div>
<div class="section" id="contact">
<h3>Contact</h3>
<h2>Contact</h2>
For any questions regarding StarPU, please contact the StarPU developers mailing list.
<pre>
<a href="mailto:starpu-devel@lists.gforge.inria.fr?subject=StarPU">starpu-devel@lists.gforge.inria.fr</a>
</pre>
</div>
<hr class="main" />
<div class="section bot">
<p class="updated">
Last updated on 2011/05/11.
Last updated on 2012/10/19.
</p>
</body>
</html>
......@@ -9,18 +9,21 @@
<body>
<h1><a href="./">StarPU</a></h1>
<h1 class="sub">Hands-on session part 1: Task-based programming model</h1>
<a href="/Runtime/">RUNTIME homepage</a> |
<a href="../">StarPU homepage</a> |
<a href="/Publis/Keyword/StarPU.html">Publications</a>
<div class="title">
<h1><a href="../../">StarPU</a></h1>
<h2>StarPU Tutorial - Workshop HPC-GA - 2012</h2>
</div>
<hr class="main"/>
<div class="menu">
<a href="/Runtime/">RUNTIME TEAM</a> |
<a href="../../">StarPU homepage</a> |
<a href="/Publis/Keyword/StarPU.html">Publications</a>
</div>
<div class="section">
<h3>Download & Install</h3>
<h2>Download & Install</h2>
<div class="section">
<h4>DAS-4 modules</h4>
<p>The first step is of course to download and install StarPU. Before doing so,
make sure to enable paths to the CUDA and CUBLAS environments on your machine,
......@@ -110,9 +113,11 @@ $ export STARPU_HOSTNAME=das4
a heterogeneous cluster, the cluster launcher script should set various
hostnames for the different node classes, as appropriate.</p>
</div>
</div>
<hr class="main"/>
<div class="section">
<h2>Hands-on session part 1: Task-based programming model</h2>
<div class="section">
<h3>Application example: vector scaling</h3>
......@@ -275,11 +280,11 @@ have been easily ported to StarPU by simply using <tt>starpu_insert_task</tt>.
matrix-matrix multiplication as an example. Try to run it with various numbers
of tasks</p>
</div>
</div>
<hr class="main" />
<h1 class="sub">Hands-on session part 2: Optimizations</h1>
<div class="section">
<h2>Hands-on session part 2: Optimizations</h2>