Universidad de Zaragoza Custodiado por la Biblioteca de la Universidad de Zaragoza Premis-plugin for CDSInvenio, developed by Miguel Martín Miguel Martín González 02408nmm 2200000 a 4500

oai:zaguan.unizar.es:6812 2013-05-23

eng Bosque Arbiol, Ana Llabería Griñó, José María Viñals Yúfera, Víctor Ibáñez Marín, Pablo http://zaguan.unizar.es/record/6812 Nowadays, most computer manufacturers offer chip multiprocessors (CMPs) due to the always increasing chip density. These CMPs have a broad range of characteristics, but all of them support the shared memory programming model. As a result, every CMP implements a coherence protocol to keep local caches coherent. Coherence protocols consume an important fraction of power to determine which coherence action to perform. Specifically, on CMPs with write-through local caches, a shared cache and a directory-based coherence protocol implemented as a duplicate of local caches tags, we have observed that energy is wasted in the directory due to two main reasons. Firstly, an important fraction of directory lookups are useless, because the target block is not located in any local cache. The power consumed by the directory could be reduce by filtering out useless directory lookups. Secondly, useful directory lookups (there are local copies of the target block) are performed over target blocks that are shared by a small number of processors. The directory power consumption could be reduced by limiting the directory lookups to only the directory entries that have a copy of the block. Along this thesis we propose two filtering mechanisms. Each of these mechanisms is focused on one of the problems described above: while our first proposal focuses on reducing number of directory lookups performed, our second proposal aims at reducing the associativity of directory lookups. Several implementations of both filtering approaches have been proposed and evaluated, having all of them a very limited hardware complexity. Our results show that the power consumed by the directory can be reduced as much as 30%. info:eu-repo/semantics/openAccess http://biblioteca.unizar.es/tdr/tdr_acerca.php info:eu-repo/semantics/doctoralThesis info:eu-repo/semantics/publishedVersion application/pdf 2012-01-24 02408nmm 2200000 a 4500 6812 TESIS-2012-003 eng Bosque Arbiol, Ana Filtering directory lookups in CMPS Zaragoza Universidad de Zaragoza, Prensas de la Universidad 2011 98 Tesis de la Universidad de Zaragoza 2011-22 2254-7606 Presentado: 11 11 2011 Tesis-Univ. Zaragoza Zaragoza, Universidad de Zaragoza 2011 Nowadays, most computer manufacturers offer chip multiprocessors (CMPs) due to the always increasing chip density. These CMPs have a broad range of characteristics, but all of them support the shared memory programming model. As a result, every CMP implements a coherence protocol to keep local caches coherent. Coherence protocols consume an important fraction of power to determine which coherence action to perform. Specifically, on CMPs with write-through local caches, a shared cache and a directory-based coherence protocol implemented as a duplicate of local caches tags, we have observed that energy is wasted in the directory due to two main reasons. Firstly, an important fraction of directory lookups are useless, because the target block is not located in any local cache. The power consumed by the directory could be reduce by filtering out useless directory lookups. Secondly, useful directory lookups (there are local copies of the target block) are performed over target blocks that are shared by a small number of processors. The directory power consumption could be reduced by limiting the directory lookups to only the directory entries that have a copy of the block. Along this thesis we propose two filtering mechanisms. Each of these mechanisms is focused on one of the problems described above: while our first proposal focuses on reducing number of directory lookups performed, our second proposal aims at reducing the associativity of directory lookups. Several implementations of both filtering approaches have been proposed and evaluated, having all of them a very limited hardware complexity. Our results show that the power consumed by the directory can be reduced as much as 30%. Arquitectura de ordenadores Llabería Griñó, José María dir. Viñals Yúfera, Víctor dir. Ibáñez Marín, Pablo dir. Universidad de Zaragoza Informática e Ingeniería de Sistemas zaguan@unizar.es http://zaguan.unizar.es/record/6812/files/TESIS-2012-003.pdf Texto completo (eng) oai:zaguan.unizar.es:6812 driver public Tesis tesis Arquitectura y tecn. Computadoras Informática e Ingeniería de Sistemas TESIS URI http://zaguan.unizar.es/record/6812 SUPPORTED 0 MD5 http://zaguan.unizar.es/record/6812/files/TESIS-2012-003.md5 4096 image/x.djvu 6 http://djvu.sourceforge.net/abstract.html DJVU/6 Profile information Lizardtech Document Express Enterprise 5.1 2011-01-19T11:29:27 URI http://zaguan.unizar.es/record/6812/files/TESIS-2012-003.pdf disk Minimum View Print Visualization of DJVU requires specific software, like DjVu Browser Plugin URI http://creativecommons.org/licenses/by-nc/3.0 URI http://creativecommons.org/licenses/by-nc/3.0 license URI http://creativecommons.org/licenses/by-nc/3.0 You are free to adapt, copy, transmite or distribute the work under the following conditions: (1) You must attribute the work in the manner specified by the author or licensor (but not in any way that suggests that they endorse you or your use of the work). (2) You may not use this work for commercial purposes (3) For any reuse or distribution, you must make clear to others the license terms of this work (4) Any of the above conditions can be waived if you get permission from the copyright holder (5) Nothing in this license impairs or restricts the author's moral rights This object is licensed under Creative Common Attribution-NonCommercial 3.0 (further details: http://creativecommons.org/licenses/by-nc/3.0/). Universidad de Zaragoza Automatizacion de Bibliotecas Edif. Matematicas, Pedro Cerbuna 12, 50009 Zaragoza auto.buz@unizar.es