===== Distributed systems: paradigms and models ===== This page will host all the **support material** relative to the course: useful links, teacher's notes, articles, etc. The notes relative to the structured parallel programming models that were supposed to be available before Christmas will not be published before end of January (presumably). We apologize for this. All the course lessons are now available on the course web pages (slides and recordings) and we updated the support material web page in such a way most of the arguments dealt with in the course have some pointers to online documentation and/or already published papers. ==== Logistics ==== * [[http://maps.google.com/maps/ms?ie=UTF8&t=h&hl=en&msa=0&ll=43.715969,10.410662&spn=0.02103,0.038838&z=15&msid=112905357480765876314.000473dc17afab3a621d8|This link]] shows you the **position of all the buildings somehow relevant to the Master** * [[cdcFibonacci|Rules to access the Computer facilities at the Polo Fibonacci]] ==== Admin ==== * [[http://compass2.di.unipi.it/didattica/win18/|Link]] to the secretary web page for the Master * [[http://www.di.unipi.it/mcsn2009/Programs/ProgrammiMCSN/Course_Programs_%28English%29.html|Link]] to the page with all the (preliminary) programs of the courses (in English) * [[http://www.mcsn.sssup.it/|Link]] to St. Anna master web page (mainly for registration/enrollment process)===== Course support material and links ===== ==== Slides of all the lessons given in the course ==== Lesson slides for the single lessons given during the course can be taken from the [[http://www.cli.di.unipi.it/doku/doku.php/magistraleinformaticanetworking/spm/sdpm09lessons|lessons]] web page. Here is a [[http://backus.di.unipi.it/~marcod/SPM/AllLessonsB2p.pdf|link]] with a PDF with all the slides used (included handwritings, if available), in a two slides per page format. ==== Course introduction and motivations ==== * [[http://www.top500.org|Link]] to the Top500.org home page * [[http://europar2008.caos.uab.es/documentos/presentation_gonzalez.pdf|Presentation]] by Prof. Antonio González on Elastic Parallel Architectures, at Europar'2009 * Intel [[http://techresearch.intel.com/articles/Tera-Scale/1449.htm|80 core]] teraflop chip * [[http://www.tilera.com/products/processors.php|Tilera]] multicore chip home page * The book //Introduction to parallel computing// ([[http://www.amazon.com/Introduction-Parallel-Computing-Ananth-Grama/dp/0201648652|Amazon pointer]]) by Grama, Karypis, Kumar and Gupta, has in Chapter 5 a good introduction to performance modelling of parallel forms. Most of the analytical modelling presented in SPM lessons is discussed in the book by M. Vanneschi "Architettura degli elaboratori", PLUS ed. 2010 in chapter X (this is in Italian, however). ==== Structured parallel/distributed programming ==== === Skeletons === * the [[http://homepages.inf.ed.ac.uk/mic/Skeletons/index.html|Skeletal parallelism home page]] maintained by [[http://homepages.inf.ed.ac.uk/mic/|Murray Cole]] * Cole's PhD [[http://homepages.inf.ed.ac.uk/mic/Pubs/skeletonbook.ps.gz|thesis]] (gzipped postscript) * [[http://homepages.inf.ed.ac.uk/mic/Pubs/manifesto.pdf|Bringing skeletons out of the closet: a pragmatic manifesto for skeletal parallel programming]] paper (Cole's manifesto on skeletal programming) * Paper //[[http://calvados.di.unipi.it/storage/paper_files/2006_advske_PC.pdf|Algorithmic skeletons meeting grids]]// by Aldinucci, Danelutto, Parallel Computing 32 (2006) 449–462, extending the principles of Cole's manifesto. * P3L paper [[http://calvados.di.unipi.it/storage/paper_files/1995_p3l_concurrency.pdf|Draft]] here. * [[http://backus.di.unipi.it/~marcod/SPM/mdf_paper.pdf|Paper]] introducing the macro data flow implementation model for skeletons * [[http://backus.di.unipi.it/~marcod/SPM/assistmdf.pdf|Paper]] comparing template based and macro data flow based implementation of the ASSIST parmod skeleton (//An alternative implementation schema for ASSIST parmod// by M. Danelutto, C. Migliore & C. Pantaleo). * [[http://www.di.unipi.it/~aldinuc/papers/2007_SCPE_muskel.pdf|Muskel: an expandable skeleton environment]] by M. Aldinucci, M. Danelutto and P. Dazzi (paper on Muskel expandability through macro data flow graphs) * Structured and unstructured parallelism exploitation using macro data flow. [[http://backus.di.unipi.it/~marcod/SPM/IC.59_PAPP06_reading.pdf|//Joint Structured/Unstructured Parallelism Exploitation in muskel//]], by Danelutto, Dazzi, Proc. of ICCS: Intl. Conference on Computational Science, Workshop on Practical Aspects of High-level Parallel Programming, LNCS Springer, 2006. * Skeleton semantics. One [[http://www.di.unipi.it/~aldinuc/paper_files/2005_semantics_CLSS.pdf|paper]] from our group in Pisa (//Skeleton-based parallel programming: Functional and parallel semantics in a single shot// by Aldinucci, Danelutto, Computer Languages, Systems & Structures 33 (2007) pages 179 – 192. Another [[http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=04457103|paper]] by Mario Leyton (//Type Safe Algorithmic Skeletons// by Leyton, Caromel, 16th Euromicro PDP conference, Toulouse 2008). * Performance modeling of skeleton implementations. The notes "ASE" by M. Vanneschi, edited by SEU, have a chapter on analytical modelling of skeleton, which is basically what we used in this course (this is in Italian, unfortunately). The book mentioned above //Introduction to parallel computing// ([[http://www.amazon.com/Introduction-Parallel-Computing-Ananth-Grama/dp/0201648652|Amazon pointer]]) by Grama et al. has also a chapter with the definitions of the measures we used to evaluate performance. * The book by S. Pelagatti [[http://www.amazon.com/Structured-Development-Parallel-Programs-Pelagatti/dp/0748407596|Structured development of parallel programs]] is a quite old book on skeletons that can be found in our library === Design patterns === * [[http://www.amazon.com/Design-Patterns-Elements-Reusable-Object-Oriented/dp/0201633612|Amazon pointer]] to //Design Patterns: Elements of Reusable Object-Oriented Software// book by Erich Gamma, Richard Helm, Ralph Johnson, John M. Vlissides (this is not relative to "parallel" patterns, actually) * [[http://www.amazon.com/Patterns-Parallel-Programming-Timothy-Mattson/dp/0321228111|Amazon pointer]] to //Patterns for Parallel Programming// by Mattson, Sanders and Massingill, this is the book of parallel design patterns mentioned during the lessons * Parallel design pattern [[http://www.cs.uiuc.edu/homes/snir/PPP/|resource page]] * [[http://www.cs.uiuc.edu/homes/snir/PPP/design/massingill.pdf|1]] [[http://www.cs.uiuc.edu/homes/snir/PPP/design/massingill2.pdf|2]] [[http://www.cs.uiuc.edu/homes/snir/PPP/design/massingill3.pdf|3]] [[http://www.cs.uiuc.edu/homes/snir/PPP/design/massingill4.pdf|4]] : pointers to papers by Massingill et al. illustrating parallel design patterns. === NEW! === * A fresh look at the parallel programming problems: [[http://delivery.acm.org/10.1145/1570000/1562783/p56-asanovic.pdf?key1=1562783&key2=4491499521&coll=portal&dl=ACM&ip=131.114.88.241&CFID=66597108&CFTOKEN=59109785|A View of the Parallel Computing Landscape]] by Krste AsAnoViC, rAstisLAV boDiK, JAmes DemmeL, tonY KeAVenY, Kurt KeutZer, John KubiAtoWiCZ, neLson morGAn, DAViD PAtterson, KoushiK sen, John WAWrZYneK, DAViD WesseL, AnD KAtherine YeLiCK. ==== Target architectures (sw) ==== === Non structured === * Sample POSIX/TCP code (pipeline) * [[http://backus.di.unipi.it/~marcod/SPM/main_thread.cpp|Code]] for opening two connections (server and client) with pthreads * [[http://backus.di.unipi.it/~marcod/SPM/main_openmp.cpp|Code]] for opening two connections (server and client) with OpenMP * [[http://backus.di.unipi.it/~marcod/SPM/pipeline1.tgz|Code]] for the Pipeline example (naif version, only integers, connections opened for each communication) * [[http://backus.di.unipi.it/~marcod/SPM/rts.tgz|Code]] for the RTS answering discovery messages * [[http://backus.di.unipi.it/~marcod/SPM/discoverer.cpp|Code]] for the discoverer process * [[http://backus.di.unipi.it/~marcod/SPM/omp_stage.cpp|Code]] for the enhanced, standalone Pipeline example * OpenMP * [[http://openmp.org/wp/|Home page]] * [[http://www.openmp.org/mp-documents/spec30.pdf|Manual]] relative to version 3.0 * [[https://computing.llnl.gov/tutorials/openMP/|Tutorial]] at LLNL * [[http://hpc.cs.tsinghua.edu.cn/research/cfp/iwomp2007/slides/cpp_openmp_terboven.pdf|Slides]] from Tsinghua * MPI * [[http://www.mcs.anl.gov/research/projects/mpich2|MPICH]] home page * [[http://www.mpi-forum.org|MPI Forum]] * [[http://www.lam-mpi.org|LAM/MPI]] home page * TBB * [[http://www.threadingbuildingblocks.org/|Intel Thread Building Block]] library partially adopts skeleton concepts and technology. === Structured === C(++) + MPI based * eSkel * [[http://homepages.inf.ed.ac.uk/abenoit1/eSkel/|eSkel home page]]. This is the skeleton library maintained at the group of Murray Cole in Edinburgh. * Muesli * Muesli is the skeleton library by Herbert Kuchen at Muenster University. [[http://www.wi.uni-muenster.de/pi/forschung/Skeletons/index.html|Here]] is the library homepage. A good paper on Muesli can be found at [[http://www.wi.uni-muenster.de/imperia/md/content/erciscontent/publications/ercis_wp_no_7.pdf|this link]]. * SkeTo * SkeTo is the skeleton library running on top of MPI by Tokyo skeleton group. The [[http://www.ipl.t.u-tokyo.ac.jp/sketo/|home page]] has links to download the library and the relative documentation. * MALLBA * This is a project run by Spanish Univ. Here is the project [[http://www.lsi.upc.edu/~mallba/|home page]] * ASSIST * This is not targeting MPI, actually (it uses a POSIX/TCP target). Most of the concepts in ASSIST come from the skeleton and coordination language frameworks. ASSIST home page is [[http://www.di.unipi.it/Assist.html|here]] Java based * ProActive/Calcium home page * [[http://proactive.inria.fr|Home page]] * Mario Leyton has recently released a new version of Calcium, which is named Skandium. This is the [[http://skandium.niclabs.cl|Skandium home page]] where you can download code (GPL) and documentation. * Behavioural skeletons * Behavioural skeletons run on top of ProActive and provide rule based autonomic managers of non functional features (performance). This is the [[http://gridcomp.ercim.org/content/view/26/34/|home page]] within the GridCOMP project web pages. * Muskel * Muskel [[http://backus.di.unipi.it/~marcod/wiki/doku.php?id=muskel|home page]] * Download muskel from [[http://backus.di.unipi.it/~marcod/SPM/muskel.tgz|this link]] AOP/annotations based * Joao Sobral AOP skeletons * Joao Sobral at Departamento de Informática, Universidade do Minho in Portugal developed some nice libraries merging AOP (Aspect Oriented Programming) and skeleton concepts ([[http://www.di.uminho.pt/~jls/|home page]] and papers [[http://www.di.uminho.pt/~jls/Cluster2007PubFinal.pdf|1]] [[http://www.di.uminho.pt/~jls/64_Cunha_C.pdf|2]] related to the subject) ML based * OcamlP3L * This is an implementation of P3L written in Ocaml, the french dialect of ML ([[http://ocamlp3l.inria.fr/eng.htm|home page]]). * Skipper * Several version of Skipper was developed by J. Serot at LASMEA, here is the [[http://wwwlasmea.univ-bpclermont.fr/Personnel/Jocelyn.Serot/skipperd.html|home page]] of the last version, Skipper-D. Skipper adopts a macro data flow implementation. === Notes === * [[http://backus.di.unipi.it/~marcod/Tools/ProactiveScaLab.pdf|Notes]] on components with ProActive and SCA Tuscany (draft, all the examples detailed have been tested, but there are more recent versions of both ProActive and Tuscany SCA)