<?xml version="1.0"?>
<feed xmlns="http://www.w3.org/2005/Atom" xml:lang="en">
		<id>http://www.colloquiam.com/wd/index.php?action=history&amp;feed=atom&amp;title=Benoit_et_al_2011a</id>
		<title>Benoit et al 2011a - Revision history</title>
		<link rel="self" type="application/atom+xml" href="http://www.colloquiam.com/wd/index.php?action=history&amp;feed=atom&amp;title=Benoit_et_al_2011a"/>
		<link rel="alternate" type="text/html" href="http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;action=history"/>
		<updated>2026-05-11T21:30:39Z</updated>
		<subtitle>Revision history for this page on the wiki</subtitle>
		<generator>MediaWiki 1.27.0-wmf.10</generator>

	<entry>
		<id>http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=186130&amp;oldid=prev</id>
		<title>Scipediacontent at 16:16, 25 January 2021</title>
		<link rel="alternate" type="text/html" href="http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=186130&amp;oldid=prev"/>
				<updated>2021-01-25T16:16:22Z</updated>
		
		<summary type="html">&lt;p&gt;&lt;/p&gt;
&lt;table class=&quot;diff diff-contentalign-left&quot; data-mw=&quot;interface&quot;&gt;
				&lt;col class='diff-marker' /&gt;
				&lt;col class='diff-content' /&gt;
				&lt;col class='diff-marker' /&gt;
				&lt;col class='diff-content' /&gt;
				&lt;tr style='vertical-align: top;' lang='en'&gt;
				&lt;td colspan='2' style=&quot;background-color: white; color:black; text-align: center;&quot;&gt;← Older revision&lt;/td&gt;
				&lt;td colspan='2' style=&quot;background-color: white; color:black; text-align: center;&quot;&gt;Revision as of 16:16, 25 January 2021&lt;/td&gt;
				&lt;/tr&gt;&lt;tr&gt;&lt;td colspan=&quot;2&quot; class=&quot;diff-lineno&quot; id=&quot;mw-diff-left-l2&quot; &gt;Line 2:&lt;/td&gt;
&lt;td colspan=&quot;2&quot; class=&quot;diff-lineno&quot;&gt;Line 2:&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;== Abstract ==&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;== Abstract ==&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications are composed of a sequence of consecutive tasks linked as a linear graph (pipeline), with a type associated to each task. The challenge is to specialize the machines of a target platform to process only one task type, given that every machine is able to process all the types before being specialized, to avoid costly context or setup changes. Each instance can thus be performed by any machine specialized in its type and the workload of the system can be shared among a set of specialized machines. For identical machines, we prove that an optimal solution can be computed in polynomial time. However, the problem becomes NP-hard when two machines can compute the same task type at different speeds. Several polynomial time heuristics are presented for the most realistic specialized settings. Experimental results show that the best heuristics obtain a good throughput, much better than the throughput obtained with a random mapping, and close to the optimal throughput in the particular cases on which the optimal throughput can be computed.; Dans ce rapport, nous &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Ã©tudions &lt;/del&gt;le &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;problÃ¨me &lt;/del&gt;de l'optimisation du &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bit &lt;/del&gt;d'applications de type pipeline dans un environnement &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;hÃ©tÃ©rogÃ¨ne &lt;/del&gt;sujet &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Ã  &lt;/del&gt;des pannes. Les applications sont &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;constituÃ©es &lt;/del&gt;d'un ensemble de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢ches consÃ©cu&lt;/del&gt;\-tives &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;typÃ©es &lt;/del&gt;et &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;organisÃ©es &lt;/del&gt;selon une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;chaÃ®ne linÃ©aire &lt;/del&gt;ou pipeline. Le but est ici de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;spÃ©cialiser &lt;/del&gt;les machines de la plate-forme d'&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;exÃ©cution &lt;/del&gt;afin qu'elles ne traitent qu'un seul type de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢ches&lt;/del&gt;, sachant qu'au &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©part &lt;/del&gt;elles peuvent &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;exÃ©cuter &lt;/del&gt;tous les types. Cela permet d'&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Ã©viter &lt;/del&gt;des reconfigurations &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;coÃ»teuses &lt;/del&gt;entre &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢ches &lt;/del&gt;de types &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;diffÃ©rents &lt;/del&gt;sur une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;mÃªme &lt;/del&gt;machine. Ainsi, les instances d'une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;mÃªme tÃ¢che &lt;/del&gt;peuvent &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Ãªtre distribuÃ©es &lt;/del&gt;sur plusieurs machines &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;spÃ©cialisÃ©es &lt;/del&gt;pour le type de cette &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢che&lt;/del&gt;, ce qui permet une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©partition &lt;/del&gt;de la charge du &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;systÃ¨me &lt;/del&gt;sur un ensemble de machines &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;spÃ©cialisÃ©es&lt;/del&gt;. Lorsque la plate-forme est &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;composÃ©e &lt;/del&gt;de machines identiques, nous prouvons qu'une solution optimale peut &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Ãªtre trouvÃ©e &lt;/del&gt;en temps polynomial. Par contre, le &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;problÃ¨me &lt;/del&gt;devient NP-complet &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dÃ¨s &lt;/del&gt;lors que deux machines peuvent traiter une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;mÃªme tÃ¢che Ã  &lt;/del&gt;des vitesses &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;diffÃ©rentes&lt;/del&gt;. Ce faisant, plusieurs heuristiques sont &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;prÃ©sentÃ©es &lt;/del&gt;dans le cas le plus &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©aliste &lt;/del&gt;d'un &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;systÃ¨me spÃ©cialisÃ©&lt;/del&gt;. Les &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©sultats expÃ©rimentaux &lt;/del&gt;montrent que les meilleures heuristiques obtiennent de bons &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©sultats &lt;/del&gt;en terme de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bit&lt;/del&gt;, meilleurs qu'avec une allocation &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;alÃ©atoire&lt;/del&gt;, et que les &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bits &lt;/del&gt;atteints sont &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;trÃ¨s &lt;/del&gt;proches des &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bits &lt;/del&gt;optimaux dans les cas particuliers pour lesquels une solution optimale peut &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Ãªtre calculÃ©e&lt;/del&gt;.&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications are composed of a sequence of consecutive tasks linked as a linear graph (pipeline), with a type associated to each task. The challenge is to specialize the machines of a target platform to process only one task type, given that every machine is able to process all the types before being specialized, to avoid costly context or setup changes. Each instance can thus be performed by any machine specialized in its type and the workload of the system can be shared among a set of specialized machines. For identical machines, we prove that an optimal solution can be computed in polynomial time. However, the problem becomes NP-hard when two machines can compute the same task type at different speeds. Several polynomial time heuristics are presented for the most realistic specialized settings. Experimental results show that the best heuristics obtain a good throughput, much better than the throughput obtained with a random mapping, and close to the optimal throughput in the particular cases on which the optimal throughput can be computed.; Dans ce rapport, nous &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;étudions &lt;/ins&gt;le &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;problème &lt;/ins&gt;de l'optimisation du &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;débit &lt;/ins&gt;d'applications de type pipeline dans un environnement &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;hétérogène &lt;/ins&gt;sujet &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;à &lt;/ins&gt;des pannes. Les applications sont &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;constituées &lt;/ins&gt;d'un ensemble de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tâches consécu&lt;/ins&gt;\-tives &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;typées &lt;/ins&gt;et &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;organisées &lt;/ins&gt;selon une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;chaîne linéaire &lt;/ins&gt;ou pipeline. Le but est ici de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;spécialiser &lt;/ins&gt;les machines de la plate-forme d'&lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;exécution &lt;/ins&gt;afin qu'elles ne traitent qu'un seul type de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tâches&lt;/ins&gt;, sachant qu'au &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;départ &lt;/ins&gt;elles peuvent &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;exécuter &lt;/ins&gt;tous les types. Cela permet d'&lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;éviter &lt;/ins&gt;des reconfigurations &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;coûteuses &lt;/ins&gt;entre &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tâches &lt;/ins&gt;de types &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;différents &lt;/ins&gt;sur une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;même &lt;/ins&gt;machine. Ainsi, les instances d'une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;même tâche &lt;/ins&gt;peuvent &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;être distribuées &lt;/ins&gt;sur plusieurs machines &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;spécialisées &lt;/ins&gt;pour le type de cette &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tâche&lt;/ins&gt;, ce qui permet une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;répartition &lt;/ins&gt;de la charge du &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;système &lt;/ins&gt;sur un ensemble de machines &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;spécialisées&lt;/ins&gt;. Lorsque la plate-forme est &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;composée &lt;/ins&gt;de machines identiques, nous prouvons qu'une solution optimale peut &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;être trouvée &lt;/ins&gt;en temps polynomial. Par contre, le &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;problème &lt;/ins&gt;devient NP-complet &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dès &lt;/ins&gt;lors que deux machines peuvent traiter une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;même tâche à &lt;/ins&gt;des vitesses &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;différentes&lt;/ins&gt;. Ce faisant, plusieurs heuristiques sont &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;présentées &lt;/ins&gt;dans le cas le plus &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;réaliste &lt;/ins&gt;d'un &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;système spécialisé&lt;/ins&gt;. Les &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;résultats expérimentaux &lt;/ins&gt;montrent que les meilleures heuristiques obtiennent de bons &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;résultats &lt;/ins&gt;en terme de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;débit&lt;/ins&gt;, meilleurs qu'avec une allocation &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;aléatoire&lt;/ins&gt;, et que les &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;débits &lt;/ins&gt;atteints sont &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;très &lt;/ins&gt;proches des &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;débits &lt;/ins&gt;optimaux dans les cas particuliers pour lesquels une solution optimale peut &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;être calculée&lt;/ins&gt;.&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;

&lt;!-- diff cache key mw_drafts_scipedia-sc_mwd_:diff:version:1.11a:oldid:182350:newid:186130 --&gt;
&lt;/table&gt;</summary>
		<author><name>Scipediacontent</name></author>	</entry>

	<entry>
		<id>http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=182350&amp;oldid=prev</id>
		<title>Scipediacontent at 13:40, 21 January 2021</title>
		<link rel="alternate" type="text/html" href="http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=182350&amp;oldid=prev"/>
				<updated>2021-01-21T13:40:17Z</updated>
		
		<summary type="html">&lt;p&gt;&lt;/p&gt;
&lt;table class=&quot;diff diff-contentalign-left&quot; data-mw=&quot;interface&quot;&gt;
				&lt;col class='diff-marker' /&gt;
				&lt;col class='diff-content' /&gt;
				&lt;col class='diff-marker' /&gt;
				&lt;col class='diff-content' /&gt;
				&lt;tr style='vertical-align: top;' lang='en'&gt;
				&lt;td colspan='2' style=&quot;background-color: white; color:black; text-align: center;&quot;&gt;← Older revision&lt;/td&gt;
				&lt;td colspan='2' style=&quot;background-color: white; color:black; text-align: center;&quot;&gt;Revision as of 13:40, 21 January 2021&lt;/td&gt;
				&lt;/tr&gt;&lt;tr&gt;&lt;td colspan=&quot;2&quot; class=&quot;diff-lineno&quot; id=&quot;mw-diff-left-l2&quot; &gt;Line 2:&lt;/td&gt;
&lt;td colspan=&quot;2&quot; class=&quot;diff-lineno&quot;&gt;Line 2:&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;== Abstract ==&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;== Abstract ==&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications are composed of a sequence of consecutive tasks linked as a linear graph (pipeline), with a type associated to each task. The challenge is to specialize the machines of a target platform to process only one task type, given that every machine is able to process all the types before being specialized, to avoid costly context or setup changes. Each instance can thus be performed by any machine specialized in its type and the workload of the system can be shared among a set of specialized machines. For identical machines, we prove that an optimal solution can be computed in polynomial time. However, the problem becomes NP-hard when two machines can compute the same task type at different speeds. Several polynomial time heuristics are presented for the most realistic specialized settings. Experimental results show that the best heuristics obtain a good throughput, much better than the throughput obtained with a random mapping, and close to the optimal throughput in the particular cases on which the optimal throughput can be computed.; Dans ce rapport, nous &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;étudions &lt;/del&gt;le &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;problème &lt;/del&gt;de l'optimisation du &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;débit &lt;/del&gt;d'applications de type pipeline dans un environnement &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;hétérogène &lt;/del&gt;sujet &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;à &lt;/del&gt;des pannes. Les applications sont &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;constituées &lt;/del&gt;d'un ensemble de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tâches consécu&lt;/del&gt;\-tives &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;typées &lt;/del&gt;et &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;organisées &lt;/del&gt;selon une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;chaîne linéaire &lt;/del&gt;ou pipeline. Le but est ici de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;spécialiser &lt;/del&gt;les machines de la plate-forme d'&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;exécution &lt;/del&gt;afin qu'elles ne traitent qu'un seul type de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tâches&lt;/del&gt;, sachant qu'au &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;départ &lt;/del&gt;elles peuvent &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;exécuter &lt;/del&gt;tous les types. Cela permet d'&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;éviter &lt;/del&gt;des reconfigurations &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;coûteuses &lt;/del&gt;entre &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tâches &lt;/del&gt;de types &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;différents &lt;/del&gt;sur une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;même &lt;/del&gt;machine. Ainsi, les instances d'une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;même tâche &lt;/del&gt;peuvent &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;être distribuées &lt;/del&gt;sur plusieurs machines &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;spécialisées &lt;/del&gt;pour le type de cette &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;tâche&lt;/del&gt;, ce qui permet une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;répartition &lt;/del&gt;de la charge du &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;système &lt;/del&gt;sur un ensemble de machines &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;spécialisées&lt;/del&gt;. Lorsque la plate-forme est &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;composée &lt;/del&gt;de machines identiques, nous prouvons qu'une solution optimale peut &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;être trouvée &lt;/del&gt;en temps polynomial. Par contre, le &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;problème &lt;/del&gt;devient NP-complet &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;dès &lt;/del&gt;lors que deux machines peuvent traiter une &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;même tâche à &lt;/del&gt;des vitesses &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;différentes&lt;/del&gt;. Ce faisant, plusieurs heuristiques sont &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;présentées &lt;/del&gt;dans le cas le plus &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;réaliste &lt;/del&gt;d'un &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;système spécialisé&lt;/del&gt;. Les &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;résultats expérimentaux &lt;/del&gt;montrent que les meilleures heuristiques obtiennent de bons &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;résultats &lt;/del&gt;en terme de &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;débit&lt;/del&gt;, meilleurs qu'avec une allocation &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;aléatoire&lt;/del&gt;, et que les &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;débits &lt;/del&gt;atteints sont &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;très &lt;/del&gt;proches des &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;débits &lt;/del&gt;optimaux dans les cas particuliers pour lesquels une solution optimale peut &lt;del class=&quot;diffchange diffchange-inline&quot;&gt;être calculée&lt;/del&gt;.&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications are composed of a sequence of consecutive tasks linked as a linear graph (pipeline), with a type associated to each task. The challenge is to specialize the machines of a target platform to process only one task type, given that every machine is able to process all the types before being specialized, to avoid costly context or setup changes. Each instance can thus be performed by any machine specialized in its type and the workload of the system can be shared among a set of specialized machines. For identical machines, we prove that an optimal solution can be computed in polynomial time. However, the problem becomes NP-hard when two machines can compute the same task type at different speeds. Several polynomial time heuristics are presented for the most realistic specialized settings. Experimental results show that the best heuristics obtain a good throughput, much better than the throughput obtained with a random mapping, and close to the optimal throughput in the particular cases on which the optimal throughput can be computed.; Dans ce rapport, nous &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;Ã©tudions &lt;/ins&gt;le &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;problÃ¨me &lt;/ins&gt;de l'optimisation du &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bit &lt;/ins&gt;d'applications de type pipeline dans un environnement &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;hÃ©tÃ©rogÃ¨ne &lt;/ins&gt;sujet &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;Ã  &lt;/ins&gt;des pannes. Les applications sont &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;constituÃ©es &lt;/ins&gt;d'un ensemble de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢ches consÃ©cu&lt;/ins&gt;\-tives &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;typÃ©es &lt;/ins&gt;et &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;organisÃ©es &lt;/ins&gt;selon une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;chaÃ®ne linÃ©aire &lt;/ins&gt;ou pipeline. Le but est ici de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;spÃ©cialiser &lt;/ins&gt;les machines de la plate-forme d'&lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;exÃ©cution &lt;/ins&gt;afin qu'elles ne traitent qu'un seul type de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢ches&lt;/ins&gt;, sachant qu'au &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©part &lt;/ins&gt;elles peuvent &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;exÃ©cuter &lt;/ins&gt;tous les types. Cela permet d'&lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;Ã©viter &lt;/ins&gt;des reconfigurations &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;coÃ»teuses &lt;/ins&gt;entre &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢ches &lt;/ins&gt;de types &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;diffÃ©rents &lt;/ins&gt;sur une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;mÃªme &lt;/ins&gt;machine. Ainsi, les instances d'une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;mÃªme tÃ¢che &lt;/ins&gt;peuvent &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;Ãªtre distribuÃ©es &lt;/ins&gt;sur plusieurs machines &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;spÃ©cialisÃ©es &lt;/ins&gt;pour le type de cette &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;tÃ¢che&lt;/ins&gt;, ce qui permet une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©partition &lt;/ins&gt;de la charge du &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;systÃ¨me &lt;/ins&gt;sur un ensemble de machines &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;spÃ©cialisÃ©es&lt;/ins&gt;. Lorsque la plate-forme est &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;composÃ©e &lt;/ins&gt;de machines identiques, nous prouvons qu'une solution optimale peut &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;Ãªtre trouvÃ©e &lt;/ins&gt;en temps polynomial. Par contre, le &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;problÃ¨me &lt;/ins&gt;devient NP-complet &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dÃ¨s &lt;/ins&gt;lors que deux machines peuvent traiter une &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;mÃªme tÃ¢che Ã  &lt;/ins&gt;des vitesses &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;diffÃ©rentes&lt;/ins&gt;. Ce faisant, plusieurs heuristiques sont &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;prÃ©sentÃ©es &lt;/ins&gt;dans le cas le plus &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©aliste &lt;/ins&gt;d'un &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;systÃ¨me spÃ©cialisÃ©&lt;/ins&gt;. Les &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©sultats expÃ©rimentaux &lt;/ins&gt;montrent que les meilleures heuristiques obtiennent de bons &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;rÃ©sultats &lt;/ins&gt;en terme de &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bit&lt;/ins&gt;, meilleurs qu'avec une allocation &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;alÃ©atoire&lt;/ins&gt;, et que les &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bits &lt;/ins&gt;atteints sont &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;trÃ¨s &lt;/ins&gt;proches des &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;dÃ©bits &lt;/ins&gt;optimaux dans les cas particuliers pour lesquels une solution optimale peut &lt;ins class=&quot;diffchange diffchange-inline&quot;&gt;Ãªtre calculÃ©e&lt;/ins&gt;.&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&amp;#160;&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;Document type: Part of book or chapter of book&lt;/del&gt;&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&amp;#160;&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;== Full document ==&lt;/del&gt;&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;−&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #ffe49c; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;del class=&quot;diffchange diffchange-inline&quot;&gt;&amp;lt;pdf&amp;gt;Media:Draft_Content_843678797-beopen249-8427-document.pdf&amp;lt;/pdf&amp;gt;&lt;/del&gt;&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot; class=&quot;diff-lineno&quot; id=&quot;mw-diff-left-l17&quot; &gt;Line 17:&lt;/td&gt;
&lt;td colspan=&quot;2&quot; class=&quot;diff-lineno&quot;&gt;Line 12:&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;* [https://hal.inria.fr/inria-00565151/file/RR-7532.pdf https://hal.inria.fr/inria-00565151/file/RR-7532.pdf]&lt;/div&gt;&lt;/td&gt;&lt;td class='diff-marker'&gt;&amp;#160;&lt;/td&gt;&lt;td style=&quot;background-color: #f9f9f9; color: #333333; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #e6e6e6; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;* [https://hal.inria.fr/inria-00565151/file/RR-7532.pdf https://hal.inria.fr/inria-00565151/file/RR-7532.pdf]&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;* [http://link.springer.com/content/pdf/10.1007/978-3-642-23400-2_23 http://link.springer.com/content/pdf/10.1007/978-3-642-23400-2_23],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [http://dx.doi.org/10.1007/978-3-642-23400-2_23 http://dx.doi.org/10.1007/978-3-642-23400-2_23] under the license http://www.springer.com/tdm&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;* [https://hal.inria.fr/inria-00565151 https://hal.inria.fr/inria-00565151],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://hal.inria.fr/inria-00565151/document https://hal.inria.fr/inria-00565151/document],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://hal.inria.fr/inria-00565151/file/RR-7532.pdf https://hal.inria.fr/inria-00565151/file/RR-7532.pdf]&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;* [https://link.springer.com/chapter/10.1007/978-3-642-23400-2_23 https://link.springer.com/chapter/10.1007/978-3-642-23400-2_23],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [http://graal.ens-lyon.fr/~abenoit/papers/RR-7532.pdf http://graal.ens-lyon.fr/~abenoit/papers/RR-7532.pdf],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://hal.inria.fr/inria-00565151 https://hal.inria.fr/inria-00565151],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://www.scipedia.com/public/Benoit_et_al_2011a https://www.scipedia.com/public/Benoit_et_al_2011a],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://hal.inria.fr/inria-00565151/document https://hal.inria.fr/inria-00565151/document],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://hal.archives-ouvertes.fr/hal-01222766 https://hal.archives-ouvertes.fr/hal-01222766],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://dblp.uni-trier.de/db/conf/europar/europar2011-1.html#BenoitDNP11 https://dblp.uni-trier.de/db/conf/europar/europar2011-1.html#BenoitDNP11],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://rd.springer.com/chapter/10.1007%2F978-3-642-23400-2_23 https://rd.springer.com/chapter/10.1007%2F978-3-642-23400-2_23],&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;
&lt;tr&gt;&lt;td colspan=&quot;2&quot;&gt;&amp;#160;&lt;/td&gt;&lt;td class='diff-marker'&gt;+&lt;/td&gt;&lt;td style=&quot;color:black; font-size: 88%; border-style: solid; border-width: 1px 1px 1px 4px; border-radius: 0.33em; border-color: #a3d3ff; vertical-align: top; white-space: pre-wrap;&quot;&gt;&lt;div&gt;&lt;ins style=&quot;font-weight: bold; text-decoration: none;&quot;&gt;: [https://academic.microsoft.com/#/detail/21985951 https://academic.microsoft.com/#/detail/21985951]&lt;/ins&gt;&lt;/div&gt;&lt;/td&gt;&lt;/tr&gt;

&lt;!-- diff cache key mw_drafts_scipedia-sc_mwd_:diff:version:1.11a:oldid:173564:newid:182350 --&gt;
&lt;/table&gt;</summary>
		<author><name>Scipediacontent</name></author>	</entry>

	<entry>
		<id>http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=173564&amp;oldid=prev</id>
		<title>Scipediacontent: Scipediacontent moved page Draft Content 843678797 to Benoit et al 2011a</title>
		<link rel="alternate" type="text/html" href="http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=173564&amp;oldid=prev"/>
				<updated>2020-10-14T10:57:18Z</updated>
		
		<summary type="html">&lt;p&gt;Scipediacontent moved page &lt;a href=&quot;/public/Draft_Content_843678797&quot; class=&quot;mw-redirect&quot; title=&quot;Draft Content 843678797&quot;&gt;Draft Content 843678797&lt;/a&gt; to &lt;a href=&quot;/public/Benoit_et_al_2011a&quot; title=&quot;Benoit et al 2011a&quot;&gt;Benoit et al 2011a&lt;/a&gt;&lt;/p&gt;
&lt;table class=&quot;diff diff-contentalign-left&quot; data-mw=&quot;interface&quot;&gt;
				&lt;tr style='vertical-align: top;' lang='en'&gt;
				&lt;td colspan='1' style=&quot;background-color: white; color:black; text-align: center;&quot;&gt;← Older revision&lt;/td&gt;
				&lt;td colspan='1' style=&quot;background-color: white; color:black; text-align: center;&quot;&gt;Revision as of 10:57, 14 October 2020&lt;/td&gt;
				&lt;/tr&gt;&lt;tr&gt;&lt;td colspan='2' style='text-align: center;' lang='en'&gt;&lt;div class=&quot;mw-diff-empty&quot;&gt;(No difference)&lt;/div&gt;
&lt;/td&gt;&lt;/tr&gt;&lt;/table&gt;</summary>
		<author><name>Scipediacontent</name></author>	</entry>

	<entry>
		<id>http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=173563&amp;oldid=prev</id>
		<title>Scipediacontent: Created page with &quot; == Abstract ==  In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications...&quot;</title>
		<link rel="alternate" type="text/html" href="http://www.colloquiam.com/wd/index.php?title=Benoit_et_al_2011a&amp;diff=173563&amp;oldid=prev"/>
				<updated>2020-10-14T10:57:15Z</updated>
		
		<summary type="html">&lt;p&gt;Created page with &amp;quot; == Abstract ==  In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications...&amp;quot;&lt;/p&gt;
&lt;p&gt;&lt;b&gt;New page&lt;/b&gt;&lt;/p&gt;&lt;div&gt;&lt;br /&gt;
== Abstract ==&lt;br /&gt;
&lt;br /&gt;
In this report, we study the problem of optimizing the throughput of applications for heterogeneous platforms subject to failures. The considered applications are composed of a sequence of consecutive tasks linked as a linear graph (pipeline), with a type associated to each task. The challenge is to specialize the machines of a target platform to process only one task type, given that every machine is able to process all the types before being specialized, to avoid costly context or setup changes. Each instance can thus be performed by any machine specialized in its type and the workload of the system can be shared among a set of specialized machines. For identical machines, we prove that an optimal solution can be computed in polynomial time. However, the problem becomes NP-hard when two machines can compute the same task type at different speeds. Several polynomial time heuristics are presented for the most realistic specialized settings. Experimental results show that the best heuristics obtain a good throughput, much better than the throughput obtained with a random mapping, and close to the optimal throughput in the particular cases on which the optimal throughput can be computed.; Dans ce rapport, nous étudions le problème de l'optimisation du débit d'applications de type pipeline dans un environnement hétérogène sujet à des pannes. Les applications sont constituées d'un ensemble de tâches consécu\-tives typées et organisées selon une chaîne linéaire ou pipeline. Le but est ici de spécialiser les machines de la plate-forme d'exécution afin qu'elles ne traitent qu'un seul type de tâches, sachant qu'au départ elles peuvent exécuter tous les types. Cela permet d'éviter des reconfigurations coûteuses entre tâches de types différents sur une même machine. Ainsi, les instances d'une même tâche peuvent être distribuées sur plusieurs machines spécialisées pour le type de cette tâche, ce qui permet une répartition de la charge du système sur un ensemble de machines spécialisées. Lorsque la plate-forme est composée de machines identiques, nous prouvons qu'une solution optimale peut être trouvée en temps polynomial. Par contre, le problème devient NP-complet dès lors que deux machines peuvent traiter une même tâche à des vitesses différentes. Ce faisant, plusieurs heuristiques sont présentées dans le cas le plus réaliste d'un système spécialisé. Les résultats expérimentaux montrent que les meilleures heuristiques obtiennent de bons résultats en terme de débit, meilleurs qu'avec une allocation aléatoire, et que les débits atteints sont très proches des débits optimaux dans les cas particuliers pour lesquels une solution optimale peut être calculée.&lt;br /&gt;
&lt;br /&gt;
Document type: Part of book or chapter of book&lt;br /&gt;
&lt;br /&gt;
== Full document ==&lt;br /&gt;
&amp;lt;pdf&amp;gt;Media:Draft_Content_843678797-beopen249-8427-document.pdf&amp;lt;/pdf&amp;gt;&lt;br /&gt;
&lt;br /&gt;
&lt;br /&gt;
== Original document ==&lt;br /&gt;
&lt;br /&gt;
The different versions of the original document can be found in:&lt;br /&gt;
&lt;br /&gt;
* [https://hal.archives-ouvertes.fr/hal-01222766 https://hal.archives-ouvertes.fr/hal-01222766]&lt;br /&gt;
&lt;br /&gt;
* [https://hal.inria.fr/inria-00565151/file/RR-7532.pdf https://hal.inria.fr/inria-00565151/file/RR-7532.pdf]&lt;/div&gt;</summary>
		<author><name>Scipediacontent</name></author>	</entry>

	</feed>