index.html 45 KB
Newer Older
1
<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" "http://www.w3.org/TR/REC-html40/loose.dtd">
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
2 3 4 5 6
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en">
<HEAD>
<meta http-equiv="content-type" content="text/html; charset=UTF-8" />
<TITLE>StarPU</TITLE>
<link rel="stylesheet" type="text/css" href="style.css" />
7
<link rel="Shortcut icon" href="http://www.inria.fr/extension/site_inria/design/site_inria/images/favicon.ico" type="image/x-icon" />
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
8 9 10 11
</HEAD>

<body>

12
<div class="title">
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
13
<h1><a href="./">StarPU</a></h1>
14 15
<h2>A Unified Runtime System for Heterogeneous Multicore Architectures</h2>
</div>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
16

17
<div class="menu">
Nathalie Furmento's avatar
Nathalie Furmento committed
18
<a href="https://www.inria.fr/en/teams/storm/">STORM TEAM</a> |
19
&nbsp; &nbsp; &nbsp;
20
|
21
<a href="#overview">Overview</a> |
Nathalie Furmento's avatar
Nathalie Furmento committed
22
<a href="#news">News</a> |
23
<a href="#contact">Contact</a> |
Nathalie Furmento's avatar
Nathalie Furmento committed
24
<a href="people/">People</a> |
25
<a href="#features">Features</a> |
26
<a href="#software">Software</a> |
THIBAULT Samuel's avatar
THIBAULT Samuel committed
27
<a href="#tryit">Try it!</a> |
28
<a href="#publications">Publications</a> |
29
<a href="internships/">Jobs/Interns</a> |
30
<a href="files/">Download</a> |
Nathalie Furmento's avatar
Nathalie Furmento committed
31
<a href="tutorials/">Tutorials</a> |
32
<a href="https://gforge.inria.fr/plugins/mediawiki/wiki/starpu/index.php/Main_Page">Intranet</a>
Nathalie Furmento's avatar
Nathalie Furmento committed
33
</div>
34

35 36
<div class="section" id="overview">
<h3>Overview</h3>
37 38 39 40 41 42
  <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>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
43
    <li>A graph of tasks, using either the StarPU's high level <b>GCC plugin</b> pragmas, StarPU's rich <b>C API</b>, or <b>OpenMP pragmas</b>.</li>
44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61
    </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>
THIBAULT Samuel's avatar
fix URL  
THIBAULT Samuel committed
62
<span class="note">The StarPU documentation is available in <a href="./doc/starpu.pdf">PDF</a> and in <a href="./doc/html/">HTML</a>.</span> Please note that these documents are up-to-date with the latest release of StarPU.
63 64 65 66
</p>
</div>

<div class="section emphasize newslist" id="news">
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
67 68
<h3>News</h3>
<p>
69 70 71 72 73
April 2017 <b>&raquo;&nbsp;</b>A <a href="https://events.prace-ri.eu/event/618/">tutorial</a>
      "Runtime systems for heterogeneous platform programming" will be
      given at the Maison de la Simulation in May 2017.
</p>
<p>
74 75 76 77 78 79 80
March 2017 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1570"><b>The
      v1.1.7 release of StarPU is now available!</b></a>. This release notably brings the concept of
      scheduling contexts which allows to separate computation
      resources. This is intented to be the last release for the
      branch 1.1.
</p>
<p>
81 82 83 84 85 86 87
March 2017 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1570"><b>The
      1.2.1 release of StarPU is now available!</b></a>.
      This release notably brings an out-of-core support, a MIC Xeon
      Phi support, an OpenMP runtime support, and a new internal
      communication system for MPI.
</p>
<p>
Nathalie Furmento's avatar
Nathalie Furmento committed
88 89 90 91 92 93
August 2016 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1570"><b>The
      v1.1.6 release of StarPU is now available!</b></a>. This release notably brings the concept of
      scheduling contexts which allows to separate computation
      resources.
</p>
<p>
Nathalie Furmento's avatar
Nathalie Furmento committed
94 95 96 97 98 99
August 2016 <b>&raquo;&nbsp;</b><a href="http://gforge.inria.fr/frs/?group_id=1570"><b>The
      1.2.0 release of StarPU is now available!</b></a>.
      This release notably brings an out-of-core support, a MIC Xeon
      Phi support, an OpenMP runtime support, and a new internal
      communication system for MPI.
</p>
100
</div>
101 102

<div class="section emphasizebot" style="text-align: right; font-style: italic;">
103
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>.
104
See also the full <a href="news/">news</a>.
105 106
</div>

107 108 109 110 111 112 113 114 115 116 117
<div class="section" id="video">
<h3>Video Conference</h3>
<p>
A video recording (26') of a <a href=http://www.x.org/wiki/Events/XDC2014/XDC2014ThibaultStarPU/>presentation at the XDC2014 conference</a> gives an overview of StarPU
(<a href=http://www.x.org/wiki/Events/XDC2014/XDC2014ThibaultStarPU/xdc_starpu.pdf>slides</a>):
</p>
<center>
<iframe width="420" height="315" src="https://www.youtube.com/embed/frsWSqb8UJU" frameborder="0" allowfullscreen></iframe>
</center>
</div>

118 119
<div class="section" id="contact">
<h3>Contact</h3>
120
<p>For any questions regarding StarPU, please contact the StarPU developers mailing list.</p>
121 122 123
<pre>
<a href="mailto:starpu-devel@lists.gforge.inria.fr?subject=StarPU">starpu-devel@lists.gforge.inria.fr</a>
</pre>
124
<p>Details of the <a href="people/">StarPU team people</a> are also available.</p>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
125 126
</div>

127 128
<div class="section" id="features">
<h3>Features</h3>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
129

130
<h4>Portability</h4>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
131
  <p>
132 133 134 135 136 137 138 139
Portability is obtained by the means of a unified abstraction of the machine.
StarPU offers a unified offloadable task abstraction named <em>codelet</em>. Rather
than rewriting the entire code, programmers can encapsulate existing functions
within codelets. In case a codelet can run on heterogeneous architectures, <b>it
is possible to specify one function for each architectures</b> (e.g. one function
for CUDA and one function for CPUs). StarPU takes care of scheduling and
executing those codelets as efficiently as possible over the entire machine, include
multiple GPUs.
140 141 142 143
One can even specify <b>several functions for each architecture</b> (new in
v1.0) as well as
<b>parallel implementations</b> (e.g. in OpenMP), and StarPU will
automatically determine which version is best for each input size (new in v0.9).
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
144 145
  </p>

146
<h4>Data transfers</h4>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
147
  <p>
148
To relieve programmers from the burden of explicit data transfers, a high-level
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
149
data management library enforces memory coherency over the machine: before a
150 151
codelet starts (e.g. on an accelerator), all its <b>data are automatically made
available on the compute resource</b>. Data are also kept on e.g. GPUs as long as
THIBAULT Samuel's avatar
THIBAULT Samuel committed
152 153
they are needed for further tasks. When a device runs out of memory, StarPU uses
an LRU strategy to <b>evict unused data</b>. StarPU also takes care of <b>automatically
154
prefetching</b> data, which thus permits to <b>overlap data transfers with computations</b>
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
155
(including <b>GPU-GPU direct transfers</b>) to achieve the most of the architecture.
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
156 157
  </p>

158
<h4>Dependencies</h4>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
159
  <p>
160 161 162
Dependencies between tasks can be given several ways, to provide the
programmer with best flexibility:
  <ul>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
163 164
    <li><b>explicitly</b> between pairs of tasks,</li>
    <li>explicitly through <b>tags</b> which act as rendez-vous points between
165
    tasks (thus including tasks which have not been created yet),</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
166
    <li><b>implicitly</b> from RAW, WAW, and WAR data dependencies.</li>
167
  </ul>
168 169
  </p>
  <p>
170 171 172
  These dependencies are computed in a completely decentralized way.
  </p>
  <p>
173 174 175 176
StarPU also supports an OpenMP-like <a href="doc/html/DataManagement.html#DataReduction">reduction</a> access mode (new in v0.9).
  </p>
  <p>
It also supports a <a href="doc/html/DataManagement.html#DataCommute">commute</a> access mode to allow data access commutativity (new in v1.2).
177 178 179 180 181
  </p>

<h4>Heterogeneous Scheduling</h4>
  <p>
StarPU obtains
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
182
portable performances by efficiently (and easily) using all computing resources
183
at the same time. StarPU also takes advantage of the <b>heterogeneous</b> nature of a
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
184
machine, for instance by using scheduling strategies based on auto-tuned
185 186 187
performance models. These determine the relative performance achieved
by the different processing units for the various kinds of task, and thus
permits to <b>automatically let processing units execute the tasks they are the best for</b>.
188
Various strategies and variants are available. Some of them are centralized, but
THIBAULT Samuel's avatar
THIBAULT Samuel committed
189
most of them are <b>completely distributed</b>. dmda (a data-locality-aware MCT strategy,
190
thus similar to heft but starts executing tasks before the whole task graph is
191 192 193
submitted, thus allowing dynamic task submission and a decentralized scheduler),
eager (dumb centralized queue), decentralized locality-aware work-stealing, ...
The overhead per task is typically around the order of
194 195 196
magnitude of a microsecond. Tasks should thus be a few orders of magnitude
bigger, such as 100 microseconds or 1 millisecond, to make the overhead
negligible.
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
197 198
  </p>

199 200
<h4>Clusters</h4>
  <p>
201
To deal with clusters, StarPU can nicely integrate with <a href="doc/html/MPISupport.html">MPI</a> through
202 203 204 205
explicit network communications, which will then be <b>automatically combined and
overlapped</b> with the intra-node data transfers and computation. The application
can also just provide the whole task graph, a data distribution over MPI nodes, and StarPU
will automatically determine which MPI node should execute which task, and
206
<b>generate all required MPI communications</b> accordingly (new in v0.9). We
THIBAULT Samuel's avatar
THIBAULT Samuel committed
207
have gotten excellent scaling on a 256-node cluster with GPUs, we have not yet
208 209
had the opportunity to test on a yet larger cluster. We have however measured
that with naive task submission, it should scale to a thousand nodes, and with
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
210
pruning-tuned task submission, it should scale to about a <b>million nodes</b>.
211 212 213 214 215 216
  </p>

<h4>Out of core</h4>
  <p>
When memory is not big enough for the working set, one may have to resort to
using disks. StarPU makes this seamless thanks to its <a href="doc/html/OutOfCore.html">out of core support</a> (new in 1.2).
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
217 218
StarPU will <b>automatically evict</b> data from the main memory in advance, and
<b>prefetch back</b> required data before it is needed for tasks.
219 220
  </p>

221 222 223
<h4>Extensions to the C Language</h4>
<p>
  StarPU comes with a GCC plug-in
224
  that <a href="doc/html/cExtensions.html">extends the C programming
225 226
  language</a> with pragmas and attributes that make it easy
  to <b>annotate a sequential C program to turn it into a parallel
227 228 229
  StarPU program</b> (new in v1.0).
</p>

THIBAULT Samuel's avatar
THIBAULT Samuel committed
230 231
<h4>OpenMP 4 -compatible interface</h4>
<p>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
232
  <a href=http://kstar.gforge.inria.fr/>K'Star</a> provides an OpenMP
THIBAULT Samuel's avatar
THIBAULT Samuel committed
233 234 235 236 237 238 239 240 241
  4 -compatible interface on top of StarPU. This allows to just rebuild OpenMP
  applications with the K'Star source-to-source compiler, then build it with the
  usual compiler, and the result will use the StarPU runtime.
</p>
<p>
  K'Star also provides some extensions to the OpenMP 4 standard, to let the
  StarPU runtime perform online optimizations.
</p>

242 243 244 245
<h4>OpenCL-compatible interface</h4>
<p>
  StarPU provides an <a href="doc/html/SOCLOpenclExtensions.html">OpenCL-compatible interface, SOCL</a>
  which allows to simply run OpenCL applications on top of StarPU (new in v1.0).
246 247
</p>

248 249 250 251
<h4>Simulation support</h4>
<p>
  StarPU can very accurately simulate an application execution
  and measure the resulting performance thanks to using the
252
  <a href="http://simgrid.gforge.inria.fr">SimGrid simulator</a> (new in v1.1).  This allows
253 254 255 256 257
  to quickly experiment with various scheduling heuristics, various application
  algorithms, and even various platforms (available GPUs and CPUs, available
  bandwidth)!
</p>

258 259
<h4>All in all</h4>
  <p>
260
All that means that, with the help
261
of <a href="doc/html/cExtensions.html">StarPU's extensions to the C
262 263
language</a>, the following sequential source code of a tiled version of
the classical Cholesky factorization algorithm using BLAS is also valid
THIBAULT Samuel's avatar
THIBAULT Samuel committed
264
StarPU code, possibly running on all the CPUs and GPUs, and given a data
Nathalie Furmento's avatar
Nathalie Furmento committed
265
distribution over MPI nodes, it is even a distributed version!
266 267 268 269 270 271 272 273 274 275 276 277 278
  </p>

  <tt><pre>
for (k = 0; k < tiles; k++) {
  potrf(A[k,k])
  for (m = k+1; m < tiles; m++)
    trsm(A[k,k], A[m,k])
  for (m = k+1; m < tiles; m++)
    syrk(A[m,k], A[m, m])
  for (m = k+1, m < tiles; m++)
    for (n = k+1, n < m; n++)
      gemm(A[m,k], A[n,k], A[m,n])
}</pre></tt>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
279

280
<h4>Supported Architectures</h4>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
281 282
<ul>
<li>SMP/Multicore Processors (x86, PPC, ...) </li>
283
<li>NVIDIA GPUs (e.g. heterogeneous multi-GPU), with pipelined and concurrent kernel execution support (new in v1.2) and GPU-GPU direct transfers (new in 1.1)</li>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
284 285 286
<li>OpenCL devices</li>
<li>Cell Processors (experimental)</li>
</ul>
287
and soon (in v1.2)
288 289
<ul>
<li>Intel SCC</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
290
<li>Intel MIC / Xeon Phi</li>
291
</ul>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
292

293
<h4>Supported Operating Systems</h4>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
294
<ul>
Ludovic Courtès's avatar
Ludovic Courtès committed
295 296
<li>GNU/Linux</li>
<li>Mac OS X</li>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
297 298 299
<li>Windows</li>
</ul>

300
<h4>Performance analysis tools</h4>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
301 302 303 304 305 306 307 308 309
  <p>
In order to understand the performance obtained by StarPU, it is helpful to
visualize the actual behaviour of the applications running on complex
heterogeneous multicore architectures.  StarPU therefore makes it possible to
generate Pajé traces that can be visualized thanks to the <a
href="http://vite.gforge.inria.fr/"><b>ViTE</b> (Visual Trace Explorer) open
source tool.</a>
  </p>

310
<p>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
311 312 313 314 315 316
<b>Example:</b> LU decomposition on 3 CPU cores and a GPU using a very simple
greedy scheduling strategy. The green (resp. red) sections indicate when the
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>
317
<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>
318 319
</p>

Nathalie Furmento's avatar
website  
Nathalie Furmento committed
320 321 322 323 324 325
<p>
This second trace depicts the behaviour of the same application using a
scheduling strategy trying to minimize load imbalance thanks to auto-tuned
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
326
GFlop/s</b>
327
<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>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
328 329
</p>

THIBAULT Samuel's avatar
THIBAULT Samuel committed
330 331
<p>
<a href="http://www.hlrs.de/temanejo">Temanejo</a> can be used to debug the task
332
graph, as shown below (new in v1.1).
THIBAULT Samuel's avatar
THIBAULT Samuel committed
333 334 335
</p>

<center>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
336
<a href="images/temanejo.png"><img src="images/temanejo.png" width="50%"/></a>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
337 338
</center>

Nathalie Furmento's avatar
website  
Nathalie Furmento committed
339 340
</div>

341 342 343 344 345
<div class="section" id="software">
<h3>Software using StarPU</h3>

<p>
Some software is known for being able to use StarPU to tackle heterogeneous
346 347
architectures, here is a non-exhaustive list (feel free to ask to be in the
list!):
348 349 350 351
</p>

<ul>
	<li><a href="http://icl.cs.utk.edu/magma/">MAGMA</a>, dense linear algebra library, starting from version 1.1</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
352
	<li><a href="https://project.inria.fr/chameleon/">Chameleon</a>, dense linear algebra library</li>
353 354
	<li><a href="http://www.ida.liu.se/~chrke/skepu/">SkePU</a>, a skeleton programming framework.</li>
	<li><a href="http://pastix.gforge.inria.fr/">PaStiX</a>, sparse linear algebra library, starting from version 5.2.1</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
355
	<li><a href="http://buttari.perso.enseeiht.fr/qr_mumps/">qr_mumps</a>, sparse linear algebra library</li>
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
356
	<li><a href="http://scalfmm-public.gforge.inria.fr/doc/">ScalFMM</a>, N-body interaction simulation using the Fast Multipole Method. </li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
357
	<li><a href="https://gitlab.inria.fr/solverstack/maphys">MaPHyS</a>, Massively Parallel Hybrid Solver</li>
358
	<li><a href="https://hal.archives-ouvertes.fr/hal-01086246">SignalPU</a>, a Dataflow-Graph-specific programming model. </li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
359
	<li><a href="https://tel.archives-ouvertes.fr/tel-01410049/">SCHNAPS</a>, Solver for Conservative Hypebolic Non-linear systems Applied to PlasmaS. </li>
360 361
</ul>

362 363 364 365 366
<p>
You can find below the list of publications related to applications
using StarPU.
</p>

367 368
</div>

THIBAULT Samuel's avatar
THIBAULT Samuel committed
369 370 371 372
<div class="section" id="tryit">
<h3>Give it a try!</h3>
<p>
You can easily try the performance on the Cholesky factorization for
THIBAULT Samuel's avatar
THIBAULT Samuel committed
373 374 375
instance. Make sure to have the pkg-config and
<a href="http://www.open-mpi.org/projects/hwloc/">hwloc</a>
software installed for
THIBAULT Samuel's avatar
THIBAULT Samuel committed
376 377 378
proper CPU control and BLAS kernels for your computation units and configured in
your environment (e.g. MKL for CPUs and CUBLAS for GPUs).
</p>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
379 380

<tt><pre>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
381 382 383 384 385 386
$ wget http://starpu.gforge.inria.fr/files/starpu-someversion.tar.gz
$ tar xf starpu-someversion.tar.gz
$ cd starpu-someversion
$ ./configure
$ make -j 12
$ STARPU_SCHED=dmdas ./examples/cholesky/cholesky_implicit -size $((960*40)) -nblocks 40
THIBAULT Samuel's avatar
THIBAULT Samuel committed
387
$ STARPU_SCHED=dmdas mpirun -np 4 -machinefile mymachines ./mpi/examples/matrix_decomposition/mpi_cholesky_distributed -size $((960*40*4)) -nblocks $((40*4))</pre></tt>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
388 389 390 391 392

<p>Note that the dmdas scheduler uses performance models, and thus needs
calibration execution before exhibiting optimized performance (until the "model
something is not calibrated enough" messages go away).</p>

THIBAULT Samuel's avatar
THIBAULT Samuel committed
393 394 395 396 397 398 399 400 401 402 403 404 405 406
<p>To get a glimpse at what happened, you can get an execution trace by
installing
<a href="http://savannah.nongnu.org/projects/fkt">FxT</a>
and <a href="http://vite.gforge.inria.fr/">ViTE</a>, and enabling traces:
</p>

<tt><pre>
$ ./configure --with-fxt
$ make -j 12
$ STARPU_SCHED=dmdas ./examples/cholesky/cholesky_implicit -size $((960*40)) -nblocks 40
$ ./tools/starpu_fxt_tool -i /tmp/prof_file_${USER}_0
$ vite paje.trace
</pre></tt>

THIBAULT Samuel's avatar
THIBAULT Samuel committed
407 408 409 410 411 412 413 414 415 416
<p>
Starting with StarPU 1.1, it is also possible to reproduce the performance that
we show in our articles on our machines, by installing simgrid, and then using
the simulation mode of StarPU using the performance models of our machines:
</p>
  <tt><pre>
$ ./configure --enable-simgrid
$ make -j 12
$ STARPU_PERF_MODEL_DIR=$PWD/tools/perfmodels/sampling STARPU_HOSTNAME=mirage STARPU_SCHED=dmdas ./examples/cholesky/cholesky_implicit -size $((960*40)) -nblocks 40
# size	ms	GFlops
THIBAULT Samuel's avatar
THIBAULT Samuel committed
417
38400	9915	1903.7</pre></tt>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
418 419
<p>(MPI simulation is not supported yet)</p>

420 421 422 423
<div class="section" id="publications">
<h3>Publications</h3>
<p>
All StarPU related publications are also
424
listed <a href="./publications">here</a>
425 426 427 428 429 430 431
with the corresponding Bibtex entries.
</p>

<p>A good overview is available in
the following <a href="http://hal.archives-ouvertes.fr/inria-00467677">Research Report</a>.
</p>

432
<h4>General Presentations</h4> 
433 434
<ol>
<li>
435 436 437 438 439 440 441
<a name="agullo:hal-01283949"></a>Emmanuel Agullo, Olivier Aumage, Mathieu Faverge, Nathalie Furmento, Florent Pruvost, Marc Sergent,  and Samuel Thibault<br/>
<strong>Harnessing clusters of hybrid nodes with a sequential task-based programming model</strong><br/>
In <em>8th International Workshop on Parallel Matrix Algorithms and Applications</em>, July 2014<br/>
[<a href="https://hal.inria.fr/hal-01283949">WWW</a>]
[<a href="https://hal.inria.fr/hal-01283949/file/pmaa14.pdf">PDF</a>]
</li>
<li>
442 443 444
<a name="Aug11Thesis"></a>Cédric Augonnet<br/>
<strong>Scheduling Tasks over Multicore machines enhanced with Accelerators: a Runtime System's Perspective</strong><br/>
PhD thesis, Université Bordeaux 1, 351 cours de la Libération --- 33405 TALENCE cedex, December 2011<br/>
445
[<a href="http://tel.archives-ouvertes.fr/tel-00777154">WWW</a>]
446 447
</li>
<li>
448 449 450
<a name="AugThiNamWac11CCPE"></a>Cédric Augonnet, Samuel Thibault, Raymond Namyst,  and Pierre-André Wacrenier<br/>
<strong>StarPU: A Unified Platform for Task Scheduling on Heterogeneous Multicore Architectures</strong><br/>
<em>Concurrency and Computation: Practice and Experience, Special Issue: Euro-Par 2009</em>, 23:187-198, February 2011<br/>
451
[<a href="http://hal.inria.fr/inria-00550877">WWW</a>]
452
[doi:<a href="http://dx.doi.org/10.1002/cpe.1631">10.1002/cpe.1631</a>]
453
</li>
454
<li>
455 456 457 458
<a name="AugThiNamWac10RR7240"></a>Cédric Augonnet, Samuel Thibault,  and Raymond Namyst<br/>
<strong>StarPU: a Runtime System for Scheduling Tasks over Accelerator-Based Multicore Machines</strong><br/>
Technical Report 7240, INRIA, March 2010<br/>
[<a href="http://hal.inria.fr/inria-00467677">WWW</a>]
459 460
</li>
<li>
461 462 463 464
<a name="Aug09Renpar19"></a>Cédric Augonnet<br/>
<strong>StarPU: un support exécutif unifié pour les architectures multicoeurs hétérogènes</strong><br/>
In <em>19èmes Rencontres Francophones du Parallélisme</em>, Toulouse / France, September 2009<br/>
Note: Best Paper Award<br/>
465
[<a href="http://hal.inria.fr/inria-00411581">WWW</a>]
466 467
</li>
<li>
468 469 470 471
<a name="AugThiNamWac09Europar"></a>Cédric Augonnet, Samuel Thibault, Raymond Namyst,  and Pierre-André Wacrenier<br/>
<strong>StarPU: A Unified Platform for Task Scheduling on Heterogeneous Multicore Architectures</strong><br/>
In <em>Proceedings of the 15th International Euro-Par Conference</em>, volume 5704 of <em>Lecture Notes in Computer Science</em>, Delft, The Netherlands, pages 863-874, August 2009<br/>
Springer<br/>
472
[<a href="http://hal.inria.fr/inria-00384363">WWW</a>]
473 474 475 476 477 478 479 480
[doi:<a href="http://dx.doi.org/10.1007/978-3-642-03869-3_80">10.1007/978-3-642-03869-3_80</a>]
</li>
<li>
<a name="AugNam08HPPC"></a>Cédric Augonnet and Raymond Namyst<br/>
<strong>A unified runtime system for heterogeneous multicore architectures</strong><br/>
In <em>Proceedings of the International Euro-Par Workshops 2008, HPPC'08</em>, volume 5415 of <em>Lecture Notes in Computer Science</em>, Las Palmas de Gran Canaria, Spain, pages 174-183, August 2008<br/>
Springer<br/>
<strong>ISBN:</strong> 978-3-642-00954-9<br/>
481
[<a href="http://hal.inria.fr/inria-00326917">WWW</a>]
482 483 484 485 486 487
[doi:<a href="http://dx.doi.org/10.1007/978-3-642-00955-6_22">10.1007/978-3-642-00955-6_22</a>]
</li>
<li>
<a name="Aug08Master"></a>Cédric Augonnet<br/>
<strong>Vers des supports d'exécution capables d'exploiter les machines multicoeurs hétérogènes</strong><br/>
Mémoire de DEA, Université Bordeaux 1, June 2008<br/>
488
[<a href="http://hal.inria.fr/inria-00289361">WWW</a>]
489 490
</li>
</ol>
491
<h4>On Composability</h4> 
Nathalie Furmento's avatar
Nathalie Furmento committed
492 493
<ol>
<li>
494 495 496
<a name="AH13Renpar"></a>Andra Hugo<br/>
<strong>Le problème de la composition parallèle : une approche supervisée</strong><br/>
In <em>21èmes Rencontres Francophones du Parallélisme (RenPar'21)</em>, Grenoble, France, January 2013<br/>
497
[<a href="http://hal.inria.fr/hal-00773610">WWW</a>]
Nathalie Furmento's avatar
Nathalie Furmento committed
498 499
</li>
<li>
500 501 502
<a name="hugo:hal-00824514"></a>Andra Hugo, Abdou Guermouche, Raymond Namyst,  and Pierre-André Wacrenier<br/>
<strong>Composing multiple StarPU applications over heterogeneous machines: a supervised approach</strong><br/>
In <em>Third International Workshop on Accelerators and Hybrid Exascale Systems</em>, Boston, USA, May 2013<br/>
503
[<a href="http://hal.inria.fr/hal-00824514">WWW</a>]
504 505 506 507 508
</li>
<li>
<a name="AH11Master"></a>Andra Hugo<br/>
<strong>Composabilité de codes parallèles sur architectures hétérogènes</strong><br/>
Mémoire de Master, Université Bordeaux 1, June 2011<br/>
509
[<a href="http://hal.inria.fr/inria-00619654/en/">WWW</a>]
Nathalie Furmento's avatar
Nathalie Furmento committed
510 511
</li>
</ol>
512
<h4>On Scheduling</h4> 
513
<ol>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
514
<li>
515 516 517 518 519 520 521 522
<a name="agullo:hal-01223573"></a>Emmanuel Agullo, Olivier Beaumont, Lionel Eyraud-Dubois,  and Suraj Kumar<br/>
<strong>Are Static Schedules so Bad ? A Case Study on Cholesky Factorization</strong><br/>
In <em>IPDPS'16</em>, Proceedings of the 30th IEEE International Parallel & Distributed Processing Symposium, IPDPS'16, Chicago, IL, United States, May 2016<br/>
IEEE<br/>
[<a href="https://hal.inria.fr/hal-01223573">WWW</a>]
[<a href="https://hal.inria.fr/hal-01223573/file/heteroprioCameraReady-ieeeCompatiable.pdf">PDF</a>]
</li>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
523 524 525 526
<a name="beaumont:hal-01361992"></a>Olivier Beaumont, Terry Cojean, Lionel Eyraud-Dubois, Abdou Guermouche,  and Suraj Kumar<br/>
<strong>Scheduling of Linear Algebra Kernels on Multiple Heterogeneous Resources</strong><br/>
In <em>International Conference on High Performance Computing, Data, and Analytics (HiPC)</em>, Hyderabad, India, December 2016<br/>
[<a href="https://hal.inria.fr/hal-01361992">WWW</a>]
THIBAULT Samuel's avatar
THIBAULT Samuel committed
527
[<a href="https://hal.inria.fr/hal-01361992v2/document">PDF</a>]
THIBAULT Samuel's avatar
THIBAULT Samuel committed
528 529
</li>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
530 531 532 533 534 535 536
<a name="garciapinto:hal-01353962"></a>Vinicius Garcia Pinto, Luka Stanisic, Arnaud Legrand, Lucas Mello Schnorr, Samuel Thibault,  and Vincent Danjean<br/>
<strong>Analyzing Dynamic Task-Based Applications on Hybrid Platforms: An Agile Scripting Approach</strong><br/>
In <em>3rd Workshop on Visual Performance Analysis (VPA)</em>, Salt Lake City, United States, November 2016<br/>
Note: Held in conjunction with SC16<br/>
[<a href="https://hal.inria.fr/hal-01353962">WWW</a>]
[<a href="https://hal.inria.fr/hal-01353962/file/VPA_2016_paper_3.pdf">PDF</a>]
</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
537
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
538 539 540
<a name="JaBlHU2016a"></a>Johan Janzén, David Black-Schaffer,  and Andra Hugo<br/>
<strong>Partitioning GPUs for Improved Scalability</strong><br/>
In <em>IEEE 28th International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD)</em>, October 2016<br/>
541 542 543
[doi:<a href="http://dx.doi.org/10.1109/SBAC-PAD.2016.14">10.1109/SBAC-PAD.2016.14</a>]
</li>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
544 545 546 547 548 549
<a name="beaumont:hal-01386174"></a>Olivier Beaumont, Lionel Eyraud-Dubois,  and Suraj Kumar<br/>
<strong>Approximation Proofs of a Fast and Efficient List Scheduling Algorithm for Task-Based Runtime Systems on Multicores and GPUs</strong><br/>
Note: Working paper or preprint, October 2016<br/>
[<a href="https://hal.inria.fr/hal-01386174">WWW</a>]
[<a href="https://hal.inria.fr/hal-01386174/file/heteroPrioApproxProofsRR.pdf">PDF</a>]
</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
550
<li>
551 552 553
<a name="agullo:hal-01120507"></a>Emmanuel Agullo, Olivier Beaumont, Lionel Eyraud-Dubois, Julien Herrmann, Suraj Kumar, Loris Marchal,  and Samuel Thibault<br/>
<strong>Bridging the Gap between Performance and Bounds of Cholesky Factorization on Heterogeneous Platforms</strong><br/>
In <em>Heterogeneity in Computing Workshop 2015</em>, Hyderabad, India, May 2015<br/>
554
[<a href="https://hal.inria.fr/hal-01120507">WWW</a>]
THIBAULT Samuel's avatar
THIBAULT Samuel committed
555
</li>
556
<li>
557 558 559
<a name="sergent:hal-00978364"></a>Marc Sergent and Simon Archipoff<br/>
<strong>Modulariser les ordonnanceurs de tâches : une approche structurelle</strong><br/>
In <em>Compas'2014</em>, Neuchâtel, Suisse, April 2014<br/>
560 561
[<a href="http://hal.inria.fr/hal-00978364">WWW</a>]
[<a href="http://hal.inria.fr/hal-00978364/PDF/ordonnanceurs_modulaires.pdf">PDF</a>]
562 563
</li>
</ol>
564
<h4>On The C Extensions</h4> 
565
<ol>
566 567 568 569
<li>
<a name="LC13Report"></a>Ludovic Courtès<br/>
<strong>C Language Extensions for Hybrid CPU/GPU Programming with StarPU</strong><br/>
Research Report RR-8278, INRIA, April 2013<br/>
570 571
[<a href="http://hal.inria.fr/hal-00807033">WWW</a>]
[<a href="http://hal.inria.fr/hal-00807033/PDF/RR-8278.pdf">PDF</a>]
572 573
</li>
</ol>
574
<h4>On OpenMP Support on top of StarPU</h4> 
THIBAULT Samuel's avatar
THIBAULT Samuel committed
575
<ol>
576
<li>
577 578 579 580 581 582 583
<a name="agullo:hal-01372022"></a>Emmanuel Agullo, Olivier Aumage, Berenger Bramas, Olivier Coulaud,  and Samuel Pitoiset<br/>
<strong>Bridging the gap between OpenMP 4<br/>0 and native runtime systems for the fast multipole method</strong><br/>
Research Report RR-8953, Inria, March 2016<br/>
[<a href="https://hal.inria.fr/hal-01372022">WWW</a>]
[<a href="https://hal.inria.fr/hal-01372022/file/RR-8953.pdf">PDF</a>]
</li>
<li>
584 585 586 587
<a name="virouleau:hal-01081974"></a>Philippe Virouleau, Pierrick BRUNET, François Broquedis, Nathalie Furmento, Samuel Thibault, Olivier Aumage,  and Thierry Gautier<br/>
<strong>Evaluation of OpenMP Dependent Tasks with the KASTORS Benchmark Suite</strong><br/>
In <em>10th International Workshop on OpenMP, IWOMP2014</em>, 10th International Workshop on OpenMP, IWOMP2014, Salvador, Brazil, France, pages 16 - 29, September 2014<br/>
Springer<br/>
588
[<a href="https://hal.inria.fr/hal-01081974">WWW</a>]
589
[doi:<a href="http://dx.doi.org/10.1007/978-3-319-11454-5_2">10.1007/978-3-319-11454-5_2</a>]
THIBAULT Samuel's avatar
THIBAULT Samuel committed
590 591
</li>
</ol>
592
<h4>On MPI Support</h4> 
593 594
<ol>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
595 596 597 598 599 600 601
<a name="agullo:hal-01332774"></a>Emmanuel Agullo, Olivier Aumage, Mathieu Faverge, Nathalie Furmento, Florent Pruvost, Marc Sergent,  and Samuel Thibault<br/>
<strong>Achieving High Performance on Supercomputers with a Sequential Task-based Programming Model</strong><br/>
Research Report RR-8927, Inria Bordeaux Sud-Ouest ; Bordeaux INP ; CNRS ; Université de Bordeaux ; CEA, June 2016<br/>
[<a href="https://hal.inria.fr/hal-01332774">WWW</a>]
[<a href="https://hal.inria.fr/hal-01332774/file/RR-8927.pdf">PDF</a>]
</li>
<li>
602 603 604 605 606 607 608
<a name="augonnet:hal-00992208"></a>Cédric Augonnet, Olivier Aumage, Nathalie Furmento, Samuel Thibault,  and Raymond Namyst<br/>
<strong>StarPU-MPI: Task Programming over Clusters of Machines Enhanced with Accelerators</strong><br/>
Rapport de recherche RR-8538, INRIA, May 2014<br/>
[<a href="http://hal.inria.fr/hal-00992208">WWW</a>]
[<a href="http://hal.inria.fr/hal-00992208/PDF/RR-8538.pdf">PDF</a>]
</li>
<li>
609 610 611 612 613
<a name="AugAumFurNamThi2012EuroMPI"></a>Cédric Augonnet, Olivier Aumage, Nathalie Furmento, Raymond Namyst,  and Samuel Thibault<br/>
<strong>StarPU-MPI: Task Programming over Clusters of Machines Enhanced with Accelerators</strong><br/>
In Siegfried Benkner Jesper Larsson Träff and Jack Dongarra, editors, <em>EuroMPI 2012</em>, volume 7490 of <em>LNCS</em>, September 2012<br/>
Springer<br/>
Note: Poster Session<br/>
614
[<a href="http://hal.inria.fr/hal-00725477">WWW</a>]
615
</li>
616
</ol>
Nathalie Furmento's avatar
Nathalie Furmento committed
617
<h4>On Memory Control</h4> 
THIBAULT Samuel's avatar
THIBAULT Samuel committed
618 619
<ol>
<li>
Nathalie Furmento's avatar
Nathalie Furmento committed
620
<a name="sergent:hal-01284004"></a>Marc Sergent, David Goudin, Samuel Thibault,  and Olivier Aumage<br/>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
621
<strong>Controlling the Memory Subscription of Distributed Applications with a Task-Based Runtime System</strong><br/>
Nathalie Furmento's avatar
Nathalie Furmento committed
622
In <em>21st International Workshop on High-Level Parallel Programming Models and Supportive Environments</em>, Chicago, United States, May 2016<br/>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
623 624 625
[<a href="https://hal.inria.fr/hal-01284004">WWW</a>]
[<a href="https://hal.inria.fr/hal-01284004/file/PID4127657.pdf">PDF</a>]
</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
626 627 628 629 630 631 632
<li>
<a name="agullo:hal-01332774"></a>Emmanuel Agullo, Olivier Aumage, Mathieu Faverge, Nathalie Furmento, Florent Pruvost, Marc Sergent,  and Samuel Thibault<br/>
<strong>Achieving High Performance on Supercomputers with a Sequential Task-based Programming Model</strong><br/>
Research Report RR-8927, Inria Bordeaux Sud-Ouest ; Bordeaux INP ; CNRS ; Université de Bordeaux ; CEA, June 2016<br/>
[<a href="https://hal.inria.fr/hal-01332774">WWW</a>]
[<a href="https://hal.inria.fr/hal-01332774/file/RR-8927.pdf">PDF</a>]
</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
633
</ol>
634
<h4>On Data Transfer Management</h4> 
635 636
<ol>
<li>
637 638 639
<a name="AugCleThiNam10ICPADS"></a>Cédric Augonnet, Jérôme Clet-Ortega, Samuel Thibault,  and Raymond Namyst<br/>
<strong>Data-Aware Task Scheduling on Multi-Accelerator based Platforms</strong><br/>
In <em>The 16th International Conference on Parallel and Distributed Systems (ICPADS)</em>, Shanghai, China, December 2010<br/>
640
[<a href="http://hal.inria.fr/inria-00523937">WWW</a>]
641
[doi:<a href="http://dx.doi.org/10.1109/ICPADS.2010.129">10.1109/ICPADS.2010.129</a>]
642 643
</li>
</ol>
644
<h4>On Performance Model Tuning</h4> 
645 646
<ol>
<li>
647 648 649 650
<a name="AugThiNam09HPPC"></a>Cédric Augonnet, Samuel Thibault,  and Raymond Namyst<br/>
<strong>Automatic Calibration of Performance Models on Heterogeneous Multicore Architectures</strong><br/>
In <em>Proceedings of the International Euro-Par Workshops 2009, HPPC'09</em>, volume 6043 of <em>Lecture Notes in Computer Science</em>, Delft, The Netherlands, pages 56-65, August 2009<br/>
Springer<br/>
651
[<a href="http://hal.inria.fr/inria-00421333">WWW</a>]
652
[doi:<a href="http://dx.doi.org/10.1007/978-3-642-14122-5_9">10.1007/978-3-642-14122-5_9</a>]
653 654
</li>
</ol>
655
<h4>On The Simulation Support through SimGrid</h4> 
THIBAULT Samuel's avatar
THIBAULT Samuel committed
656
<ol>
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
657
<li>
658 659 660
<a name="stanisic:hal-01147997"></a>Luka Stanisic, Samuel Thibault, Arnaud Legrand, Brice Videau,  and Jean-François Méhaut<br/>
<strong>Faithful Performance Prediction of a Dynamic Task-Based Runtime System for Heterogeneous Multi-Core Architectures</strong><br/>
<em>Concurrency and Computation: Practice and Experience</em>, pp 16, May 2015<br/>
661 662
[<a href="https://hal.inria.fr/hal-01147997">WWW</a>]
[<a href="https://hal.inria.fr/hal-01147997/file/CCPE14_article.pdf">PDF</a>]
663
[doi:<a href="http://dx.doi.org/10.1002/cpe">10.1002/cpe</a>]
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
664
</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
665
<li>
666 667 668 669
<a name="stanisic:hal-01011633"></a>Luka Stanisic, Samuel Thibault, Arnaud Legrand, Brice Videau,  and Jean-François Méhaut<br/>
<strong>Modeling and Simulation of a Dynamic Task-Based Runtime System for Heterogeneous Multi-Core Architectures</strong><br/>
In <em>Euro-par - 20th International Conference on Parallel Processing</em>, Porto, Portugal, August 2014<br/>
Springer-Verlag<br/>
670 671
[<a href="http://hal.inria.fr/hal-01011633">WWW</a>]
[<a href="http://hal.inria.fr/hal-01011633/PDF/StarPUSG_article.pdf">PDF</a>]
THIBAULT Samuel's avatar
THIBAULT Samuel committed
672 673
</li>
</ol>
674
<h4>On The Cell Support</h4> 
675 676
<ol>
<li>
677 678 679
<a name="AugThiNamNij09Samos"></a>Cédric Augonnet, Samuel Thibault, Raymond Namyst,  and Maik Nijhuis<br/>
<strong>Exploiting the Cell/BE architecture with the StarPU unified runtime system</strong><br/>
In <em>SAMOS Workshop - International Workshop on Systems, Architectures, Modeling, and Simulation</em>, volume 5657 of <em>Lecture Notes in Computer Science</em>, Samos, Greece, July 2009<br/>
680
[<a href="http://hal.inria.fr/inria-00378705">WWW</a>]
681
[doi:<a href="http://dx.doi.org/10.1007/978-3-642-03138-0_36">10.1007/978-3-642-03138-0_36</a>]
682 683
</li>
</ol>
684
<h4>On Applications</h4> 
685
<ol>
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
686
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
687 688 689 690 691 692 693 694
<a name="couteyencarpaye:hal-01507613"></a>Jean Marie Couteyen Carpaye, Jean Roman,  and Pierre Brenner<br/>
<strong>Design and Analysis of a Task-based Parallelization over a Runtime System of an Explicit Finite-Volume CFD Code with Adaptive Time Stepping</strong><br/>
<em>International Journal of Computational Science and Engineering</em>, pp 1 - 22, 2017<br/>
[<a href="https://hal.inria.fr/hal-01507613">WWW</a>]
[<a href="https://hal.inria.fr/hal-01507613/file/flusepa-task-hal-inria-preprint.pdf">PDF</a>]
[doi:<a href="http://dx.doi.org/10.1016/j.jocs.2017.03.008">10.1016/j.jocs.2017.03.008</a>]
</li>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
695 696 697 698 699 700 701
<a name="agullo:hal-01473475"></a>Emmanuel Agullo, Alfredo Buttari, Mikko Byckling, Abdou Guermouche,  and Ian Masliah<br/>
<strong>Achieving high-performance with a sparse direct solver on Intel KNL</strong><br/>
Research Report RR-9035, Inria Bordeaux Sud-Ouest ; CNRS-IRIT ; Intel corporation ; Université Bordeaux, February 2017<br/>
[<a href="https://hal.inria.fr/hal-01473475">WWW</a>]
[<a href="https://hal.inria.fr/hal-01473475/file/RR-9035.pdf">PDF</a>]
</li>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
702 703 704 705 706 707 708
<a name="agullo:hal-01387482"></a>Emmanuel Agullo, Bérenger Bramas, Olivier Coulaud, Martin Khannouz,  and Luka Stanisic<br/>
<strong>Task-based fast multipole method for clusters of multicore processors</strong><br/>
Research Report RR-8970, Inria Bordeaux Sud-Ouest, October 2016<br/>
[<a href="https://hal.inria.fr/hal-01387482">WWW</a>]
[<a href="https://hal.inria.fr/hal-01387482/file/report-8970.pdf">PDF</a>]
</li>
<li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
709 710 711 712 713 714 715
<a name="agullo:hal-01316982"></a>E Agullo, L Giraud, A Guermouche, S Nakov,  and Jean Roman<br/>
<strong>Task-based Conjugate Gradient: from multi-GPU towards heterogeneous architectures</strong><br/>
Research Report 8912, Inria Bordeaux Sud-Ouest, May 2016<br/>
[<a href="https://hal.inria.fr/hal-01316982">WWW</a>]
[<a href="https://hal.inria.fr/hal-01316982/file/RR-8912.pdf">PDF</a>]
</li>
<li>
716 717 718
<a name="rossignon:tel-01230876"></a>Corentin Rossignon<br/>
<strong>A fine grain model programming for parallelization of sparse linear solver</strong><br/>
PhD thesis, Université de Bordeaux, July 2015<br/>
719 720
[<a href="https://tel.archives-ouvertes.fr/tel-01230876">WWW</a>]
[<a href="https://tel.archives-ouvertes.fr/tel-01230876/file/ROSSIGNON_CORENTIN_2015.pdf">PDF</a>]
THIBAULT Samuel's avatar
update  
THIBAULT Samuel committed
721
</li>
THIBAULT Samuel's avatar
THIBAULT Samuel committed
722
<li>
723 724 725
<a name="MaMiDuAuThiAoNa15"></a>Vìctor Martìnez, David Michéa, Fabrice Dupros, Olivier Aumage, Samuel Thibault, Hideo Aochi,  and Philippe Olivier Alexandre Navaux<br/>
<strong>Towards seismic wave modeling on heterogeneous many-core architectures using task-based runtime system</strong><br/>
In <em>27th International Symposium on Computer Architecture and High Performance Computing (SBAC-PAD)</em>, Florianopolis, Brazil, October 2015<br/>
726 727
[<a href="https://hal.inria.fr/hal-01182746">WWW</a>]
[<a href="https://hal.inria.fr/hal-01182746/file/sbac2015_soumission.pdf">PDF</a>]
728 729
</li>
<li>
730 731 732 733
<a name="sylvain:hal-01005765"></a>Sylvain Henry, Alexandre Denis, Denis Barthou, Marie-Christine Counilh,  and Raymond Namyst<br/>
<strong>Toward OpenCL Automatic Multi-Device Support</strong><br/>
In Fernando Silva, Ines Dutra,  and Vitor Santos Costa, editors, <em>Euro-Par 2014</em>, Porto, Portugal, August 2014<br/>
Springer<br/>
734 735
[<a href="http://hal.inria.fr/hal-01005765">WWW</a>]
[<a href="http://hal.inria.fr/hal-01005765/PDF/final.pdf">PDF</a>]
736 737
</li>
<li>
738 739 740 741 742
<a name="lacoste:hal-00987094"></a>Xavier Lacoste, Mathieu Faverge, Pierre Ramet, Samuel Thibault,  and George Bosilca<br/>
<strong>Taking advantage of hybrid systems for sparse direct solvers via task-based runtimes</strong><br/>
In <em>HCW'2014 workshop of IPDPS</em>, Phoenix, États-Unis, May 2014<br/>
IEEE<br/>
Note: RR-8446 RR-8446<br/>
743 744
[<a href="http://hal.inria.fr/hal-00987094">WWW</a>]
[<a href="http://hal.inria.fr/hal-00987094/PDF/sparsegpus.pdf">PDF</a>]
745 746
</li>
<li>
747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771
<a name="lacoste:hal-00925017"></a>Xavier Lacoste, Mathieu Faverge, Pierre Ramet, Samuel Thibault,  and George Bosilca<br/>
<strong>Taking advantage of hybrid systems for sparse direct solvers via task-based runtimes</strong><br/>
Rapport de recherche RR-8446, INRIA, January 2014<br/>
[<a href="http://hal.inria.fr/hal-00925017">WWW</a>]
[<a href="http://hal.inria.fr/hal-00925017/PDF/RR-8446.pdf">PDF</a>]
</li>
<li>
<a name="sergent:hal-00978602"></a>Emmanuel Agullo, Olivier Aumage, Mathieu Faverge, Nathalie Furmento, Florent Pruvost, Marc Sergent,  and Samuel Thibault<br/>
<strong>Overview of Distributed Linear Algebra on Hybrid Nodes over the StarPU Runtime</strong><br/>
SIAM Conference on Parallel Processing for Scientific Computing, February 2014<br/>
[<a href="http://hal.inria.fr/hal-00978602">WWW</a>]
[<a href="http://hal.inria.fr/hal-00978602/PDF/siampp14.pdf">PDF</a>]
</li>
<li>
<a name="Bor13Thesis"></a>Cyril Bordage<br/>
<strong>Ordonnancement dynamique, adapté aux architectures hétérogènes, de la méthode multipôle pour les équations de Maxwell, en électromagnétisme</strong><br/>
PhD thesis, Université Bordeaux 1, 351 cours de la Libération --- 33405 TALENCE cedex, December 2013<br/>
</li>
<li>
<a name="Hen13Thesis"></a>Sylvain Henry<br/>
<strong>Modèles de programmation et supports exécutifs pour architectures hétérogènes</strong><br/>
PhD thesis, Université Bordeaux 1, 351 cours de la Libération --- 33405 TALENCE cedex, November 2013<br/>
[<a href="http://tel.archives-ouvertes.fr/tel-00948309">WWW</a>]
</li>
<li>
772 773 774
<a name="hen13fhpc"></a>Sylvain Henry<br/>
<strong>ViperVM: a Runtime System for Parallel Functional High-Performance Computing on Heterogeneous Architectures</strong><br/>
In <em>2nd Workshop on Functional High-Performance Computing (FHPC'13)</em>, Boston, États-Unis, September 2013<br/>
775 776
[<a href="http://hal.inria.fr/hal-00851122">WWW</a>]
[<a href="http://hal.inria.fr/hal-00851122/PDF/fhpc13.pdf">PDF</a>]
777 778
</li>
<li>
779 780 781
<a name="odajima:hal-00920915"></a>Tetsuya Odajima, Taisuke Boku, Mitsuhisa Sato, Toshihiro Hanawa, Yuetsu Kodama, Raymond Namyst, Samuel Thibault,  and Olivier Aumage<br/>
<strong>Adaptive Task Size Control on High Level Programming for GPU/CPU Work Sharing</strong><br/>
In <em>The 2013 International Symposium on Advances of Distributed and Parallel Computing (ADPC 2013)</em>, Vietri sul Mare, Italie, December 2013<br/>
782 783
[<a href="http://hal.inria.fr/hal-00920915">WWW</a>]
[<a href="http://hal.inria.fr/hal-00920915/PDF/ADPC2013-117.pdf">PDF</a>]
784 785 786 787 788 789
</li>
<li>
<a name="ohshima:hal-00926144"></a>Satoshi Ohshima, Satoshi Katagiri, Kengo Nakajima, Samuel Thibault,  and Raymond Namyst<br/>
<strong>Implementation of FEM Application on GPU with StarPU</strong><br/>
In <em>SIAM CSE13 - SIAM Conference on Computational Science and Engineering 2013</em>, Boston, États-Unis, February 2013<br/>
SIAM<br/>
790
[<a href="http://hal.inria.fr/hal-00926144">WWW</a>]
791 792 793 794 795
</li>
<li>
<a name="Ros13Renpar"></a>Corentin Rossignon<br/>
<strong>Optimisation du produit matrice-vecteur creux sur architecture GPU pour un simulateur de reservoir</strong><br/>
In <em>21èmes Rencontres Francophones du Parallélisme (RenPar'21)</em>, Grenoble, France, January 2013<br/>
796
[<a href="http://hal.inria.fr/hal-00773571">WWW</a>]
797 798
</li>
<li>
799 800 801
<a name="rossignon:hal-00858350"></a>Corentin Rossignon, Pascal Hénon, Olivier Aumage,  and Samuel Thibault<br/>
<strong>A NUMA-aware fine grain parallelization framework for multi-core architecture</strong><br/>
In <em>PDSEC - 14th IEEE International Workshop on Parallel and Distributed Scientific and Engineering Computing - 2013</em>, Boston, États-Unis, May 2013<br/>
802 803
[<a href="http://hal.inria.fr/hal-00858350">WWW</a>]
[<a href="http://hal.inria.fr/hal-00858350/PDF/taggre_pdsec_2013.pdf">PDF</a>]
804
</li>
805
<li>
806 807 808 809 810 811 812 813 814 815 816 817
<a name="HenDenBar2012TSI"></a>Sylvain Henry, Alexandre Denis,  and Denis Barthou<br/>
<strong>Programmation unifiée multi-accélérateur OpenCL</strong><br/>
<em>Techniques et Sciences Informatiques</em>, (8-9-10):1233-1249, 2012<br/>
[<a href="http://hal.inria.fr/hal-00772742">WWW</a>]
</li>
<li>
<a name="MahManAugThi12TSI"></a>Sidi Ahmed Mahmoudi, Pierre Manneback, Cédric Augonnet,  and Samuel Thibault<br/>
<strong>Traitements d'Images sur Architectures Parallèles et Hétérogènes</strong><br/>
<em>Technique et Science Informatiques</em>, 2012<br/>
[<a href="http://hal.inria.fr/hal-00714858/">WWW</a>]
</li>
<li>
818 819 820
<a name="BenkBajMarSanNamThiEuroPar2012"></a>Siegfried Benkner, Enes Bajrovic, Erich Marth, Martin Sandrieser, Raymond Namyst,  and Samuel Thibault<br/>
<strong>High-Level Support for Pipeline Parallelism on Many-Core Architectures</strong><br/>
In <em>Europar - International European Conference on Parallel and Distributed Computing - 2012</em>, Rhodes Island, Grèce, August 2012<br/>
821 822
[<a href="http://hal.inria.fr/hal-00697020">WWW</a>]
[<a href="http://hal.inria.fr/hal-00697020/PDF/europar2012-submitted.pdf">PDF</a>]
823 824
</li>
<li>
825 826 827
<a name="kessler:hal-00776610"></a>Christoph Kessler, Usman Dastgeer, Samuel Thibault, Raymond Namyst, Andrew Richards, Uwe Dolinsky, Siegfried Benkner, Jesper Larsson Träff,  and Sabri Pllana<br/>
<strong>Programmability and Performance Portability Aspects of Heterogeneous Multi-/Manycore Systems</strong><br/>
In <em>Design, Automation and Test in Europe (DATE)</em>, Dresden, Allemagne, March 2012<br/>
828 829
[<a href="http://hal.inria.fr/hal-00776610">WWW</a>]
[<a href="http://hal.inria.fr/hal-00776610/PDF/date12-paper.pdf">PDF</a>]
830 831
</li>
<li>
832 833 834 835 836 837 838 839
<a name="BenPllTraTsiDolAugBacKesMolOsi11IEEEMicro"></a>Siegfried Benkner, Sabri Pllana, Jesper Larsson Träff, Philippas Tsigas, Uwe Dolinsky, Cédric Augonnet, Beverly Bachmayer, Christoph Kessler, David Moloney,  and Vitaly Osipov<br/>
<strong>PEPPHER: Efficient and Productive Usage of Hybrid Computing Systems</strong><br/>
<em>IEEE Micro</em>, 31(5):28-41, September 2011<br/>
<strong>ISSN:</strong> 0272-1732<br/>
[<a href="http://hal.inria.fr/hal-00648480">WWW</a>]
[doi:<a href="http://dx.doi.org/10.1109/MM.2011.67">10.1109/MM.2011.67</a>]
</li>
<li>
840 841 842
<a name="AguAugDonFavLanLtaTomAICCSA11"></a>Emmanuel Agullo, Cédric Augonnet, Jack Dongarra, Mathieu Faverge, Julien Langou, Hatem Ltaief,  and Stanimire Tomov<br/>
<strong>LU factorization for accelerator-based systems</strong><br/>
In <em>9th ACS/IEEE International Conference on Computer Systems and Applications (AICCSA 11)</em>, Sharm El-Sheikh, Egypt, June 2011<br/>
843
[<a href="http://hal.inria.fr/hal-00654193">WWW</a>]
844 845 846 847 848
</li>
<li>
<a name="AguAugDonFavLtaThiTom11IPDPS"></a>Emmanuel Agullo, Cédric Augonnet, Jack Dongarra, Mathieu Faverge, Hatem Ltaief, Samuel Thibault,  and Stanimire Tomov<br/>
<strong>QR Factorization on a Multicore Node Enhanced with Multiple GPU Accelerators</strong><br/>
In <em>25th IEEE International Parallel & Distributed Processing Symposium (IEEE IPDPS 2011)</em>, Anchorage, Alaska, USA, May 2011<br/>
849
[<a href="http://hal.inria.fr/inria-00547614">WWW</a>]
850 851 852 853 854 855
[doi:<a href="http://dx.doi.org/10.1109/IPDPS.2011.90">10.1109/IPDPS.2011.90</a>]
</li>
<li>
<a name="DasKesThi11ParCo"></a>Usman Dastgeer, Christoph Kessler,  and Samuel Thibault<br/>
<strong>Flexible runtime support for efficient skeleton programming on hybrid systems</strong><br/>
In <em>Proceedings of the International Conference on Parallel Computing (ParCo), Applications, Tools and Techniques on the Road to Exascale Computing</em>, volume 22 of <em>Advances of Parallel Computing</em>, Gent, Belgium, pages 159-166, August 2011<br/>
856
[<a href="http://hal.inria.fr/inria-00606200/">WWW</a>]
857 858 859 860 861
</li>
<li>
<a name="Hen11Renpar"></a>Sylvain Henry<br/>
<strong>Programmation multi-accélérateurs unifiée en OpenCL</strong><br/>
In <em>20èmes Rencontres Francophones du Parallélisme (RenPar'20)</em>, Saint Malo, France, May 2011<br/>
862
[<a href="http://hal.archives-ouvertes.fr/hal-00643257">WWW</a>]
863 864 865 866 867
</li>
<li>
<a name="MahManAugThi11Renpar20"></a>Sidi Ahmed Mahmoudi, Pierre Manneback, Cédric Augonnet,  and Samuel Thibault<br/>
<strong>Détection optimale des coins et contours dans des bases d'images volumineuses sur architectures multicoeurs hétérogènes</strong><br/>
In <em>20èmes Rencontres Francophones du Parallélisme</em>, Saint-Malo / France, May 2011<br/>
868
[<a href="http://hal.inria.fr/inria-00606195">WWW</a>]
869 870
</li>
<li>
871 872 873 874 875 876 877
<a name="AguAugDonLtaNamThiTomGPUgems"></a>Emmanuel Agullo, Cédric Augonnet, Jack Dongarra, Hatem Ltaief, Raymond Namyst, Samuel Thibault,  and Stanimire Tomov<br/>
<strong>A Hybridization Methodology for High-Performance Linear Algebra Software for GPUs</strong><br/>
In Wen-mei W<br/> Hwu, editor, <em>GPU Computing Gems</em>, volume 2<br/>
Morgan Kaufmann, September 2010<br/>
[<a href="http://hal.inria.fr/inria-00547847">WWW</a>]
</li>
<li>
878 879 880
<a name="AguAugDonLtaNamRomThiTom10SAAHPC"></a>Emmanuel Agullo, Cédric Augonnet, Jack Dongarra, Hatem Ltaief, Raymond Namyst, Jean Roman, Samuel Thibault,  and Stanimire Tomov<br/>
<strong>Dynamically scheduled Cholesky factorization on multicore architectures with GPU accelerators</strong><br/>
In <em>Symposium on Application Accelerators in High Performance Computing (SAAHPC)</em>, Knoxville, USA, July 2010<br/>
881
[<a href="http://hal.inria.fr/inria-00547616">WWW</a>]
882
</li>
883
</ol>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
884

885
<div class="section bot">
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
886
<p class="updated">
887
  Last updated on 2016/04/13.
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
888
</p>
889
</div>
Nathalie Furmento's avatar
website  
Nathalie Furmento committed
890 891 892

</body>
</html>