DOCSLIB.ORG

Copyrighted Material

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
Copyrighted Material k INDEX k k A benchmarks, 118, 314, 386, 425 aborts, 83 block access, 4 adaptive locking protocol, 466 BlockLib, 139 adaptive scheduling, 422 Brook, 145, 201 algorithmic skeleton, 263 bus interconnection, 431 algorithm view, 35 allocation, 418 C all-or-nothing transaction, 167 CABAC, 282 Amdahl’s law, 372 cache, 72, 237 analytical model, 442 cache behavior, 95 application-centric models, 44 cache coherence, 348 array, 67 cascading aborts, 172 Array-OL, 146 CAVLC, 282 atomicity, 84, 326 Cell/B.E. processor, 39 automated parallelization,COPYRIGHTED 233 Cell Superscalar, MATERIAL 43 auto-tuning, 194–195 Charm++,45 Cilk, 37 B cloud computing, 452 bag, 65 cluster, 113, 311 bandwidth, 425 code optimization, 292 Programming Multicore and Many-core Computing Systems, First Edition. Edited by Sabri Pllana and Fatos Xhafa. © 2017 John Wiley & Sons, Inc. Published 2017 by John Wiley & Sons, Inc. 481 k k 482 Index code profiling, 291 diagnostic tools, 358 collector, 268 dictionary, 68 combinability, 62 dining philosopher, 166 communication, 36 distributed desk checking, 331 communication congestion, 431 distributed programming, 451 communication links, 444 distributed real-time applications, 464 commutativity, 177 distributive review, 325 compiler, 109, 388 DOACROSS parallelism, 205 computational accelerators, 407 DOALL parallelism, 205 computing performance, 3 dynamic data structure, 218 concurrency, 36, 228, 232 dynamic scheduling, 421 concurrent code, 81 concurrent conflicting transactions, 83 E concurrent data structures, 59 eager update, 88, 207 concurrent program, 325 efficiency, 432 concurrent programming, 465 elastic transactions, 179 ConcurrentTesting, 340 embarrassingly parallel applications, 13 concurrent transactions, 91, 173 emitter, 268 conflicts, 94 encoder, 284 connectivity, 237 energy consumption, 352 consensus problem, 62 evolutionary algorithms, 301 consistency, 84 execution flow, 420 containers, 125 execution plan, 130 k contention manager, 91 execution time, 432 k context switch, 345 Corey, 457 F CPU cycles, 352 fairness, 353, 466 CPU-intensive application, 346 false conflicts, 94, 176 crossover operator, 303 farm paradigm, 271 CUDA, 41, 302 FastFlow, 148, 262 CUDPP, 138 fat-pointer, 220 CUFFT library, 108 Flynn’s taxonomy, 5 cyclic dependency deadlock, 328 fragmentation, 69 frame level, 286 D functional-level parallelism, 287 data block, 390 functional parallelism, 4 data compression, 386 data exchanges, 133 G data layout, 35 garbage collection, 70 data locality, 431 GenerOS, 457 data parallelism, 186 genetic programming, 302 data-parallel skeletons, 122 global scheduler, 463 data transfers, 36, 133 global scheduling, 463 deadlocks, 166 GPGPU, 301 debugging, 341 GPMCs, 39 decoder, 284 GPUs, 39 decompressions, 388 granularity, 466 deque, 66 graphics processor, 72 desk checking, 330 Grid, 452 k k Index 483 Grid middleware, 460 load balancing, 73 Grid OS, 460 local data structures, 466 Grid systems, 460 lock-free, 60 lock-free data structure, 63 H locking strategy, 466 hash table, 68 lock table, 95 H.264/AVC, 282 loosely coupled components, 451 heterogeneous architectures, 101 heterogeneous multicores, 19 M high-performance computing, 431, 451 many-core accelerators, 410 H.264/MPEG-4, 281 many-core architectures, 30 homogeneous components, 411 many-cores scalability, 380 Map, 139 I MapOverlap, 139 ILP wall, 11 mappers, 413 instruction set architecture, 238 mapping, 36, 52 Intel TBB, 39 mapping strategy, 435 interconflicts, 94 MapReduce, 139, 246, 411 intercore communication, 156 Map skeleton, 122 interleaving, 333 massively parallel applications, 29 intraconflicts, 94 master, 115 invisible reads, 89 master/worker pattern, 192, 432 invisible read transactions, 174 memory allocator, 69 k I/O latency, 357 memory bandwidth, 431, 466 k island model, 306 memory hierarchy, 466 isolated parallel program, 336 memory latency, 345–346, 355 isolation, 84 memory reclamation, 70 iterator, 150 memory wall, 10 Mercurium, 109 J Message Passing Interface (MPI), 37, 102, Java, 169 229, 432 JavaGrande, 200 message passing libraries, 7 J2EE, 363 message-passing paradigm, 466 joining, 191 metadata storage, 219–220 microkernels, 453 K miscompression rate, 392 kernel, 345 MPI. See Message Passing Interface (MPI) kernel function, 132 MPI applications, 432 MPI communication, 116 L multicore architectures, 9, 101 LAMP stack, 365 multicore clusters, 432 latency, 264, 369, 465, 466 multicore nodes, 431 layered design, 262 multiobjective evolutionary algorithms, 303 lazy update, 88, 207 multiobjective optimization, 302 linearizability, 60, 86 Multiple Instruction Multiple Data, 6 linked lists, 67 Multiple Instruction Single Data, 6 Linux, 170, 453 multiprocessors, 5 Linux kernel, 465 multiprogramming, 5 list, 67 multithreaded applications, 366 k k 484 Index multithreading, 8 power wall, 10 mutation operator, 304 predictability, 367 mutual exclusion, 60, 81 prefetching, 4 priority queue, 67 N productivity, 32 Nanos++, 109 programmability, 143 nested transaction, 91 programmability gap, 32 NUMA, 355 programming models, 6, 32 O Q object-orientation, 185 queue, 66 off-chip communication, 407 offline profiling, 211 R offload function, 418 read sharing, 171 offloading, 253 read-write ratio, 466 off-the-shelf components, 408 real-time applications, 467 OmpSs programming model, 46, 102 real-time scheduling, 463, 465 on-chip memory, 391 Reduce, 139 opacity, 86 reducers, 413 OpenCL, 47, 129 Reduction, 123 OpenCL/CUDA, 102 region tree, 113 OpenMP, 37, 102, 212, 230 regression analysis, 381 OPL, 46 repeatability, 367 optimistic concurrency control, 83 k resource aware, 249 k optimizations, 36 resource configurations, 415 review techniques, 324 P runtime system, 102 parallel bug patterns, 325 parallel design patterns, 191, 265 S parallel implementation, 344 parallelism, 3 scalability, 228, 363, 367, 432 parallel performance, 343 scalability tests, 366 parallel statements, 189 scalar processor, 4 partitioned scheduling, 463 scheduling, 36, 52 performance, 32 semantics, 84 performance aware, 249 sensitivity analysis, 395 performance bottlenecks, 96 sequential implementation, 344 performance metrics, 432, 467 Sequoia, 45 performance optimization, 344 serializability, 85 performance portability, 243 set, 68 performance predictions, 243 shared cache, 431 pessimistic concurrency control, 83 shared memory, 8, 303 PetaBricks, 246 shared-memory communication, 466 pipeline parallelism, 186 shared-memory locking, 466 pipelining, 4, 124 shared-memory multicore processor, 343 polymorphism, 177 shared memory multiprocessor, 6 population, 305 shared-memory paradigm, 466 portability, 32, 121 shared-memory parallel programming, POSIX, 293, 467 343 k k Index 485 shared-memory programming, 231 task parallelism, 186 Simics, 393 test environment, 378 single global lock, 85 testing, 323 Single Instruction Single Data, 5 thread-based programming, 229 single-system image, 455 Threading Building Blocks, 138 skeletal approach, 265 thread-level parallelism, 205 skeleton, 104 threads, 37, 345 skeleton programming, 122 throughput, 380, 465 skip list, 68 timing requirements, 465 SLICES, 161 trace, 106 socket, 348 transaction, 166 software accelerators, 266 transactional abort, 82 software pipelining, 222 transactional boosting, 177 software transactional memory, 81 transactional commit, 82 SP@CE, 44 transactional isolation, 87 SpecFP2000, 393 transactional memory, 119, 165 speculative computation, 209 transactional polymorphism, 178 speculative parallelization, 206 transaction models, 177 speculative thread, 207 transaction nesting, 169 speedup, 310, 381, 432, 435 transformation, 153 splitting, 191 tree, 69 SPMD applications, 433 try-to-eat procedure, 167 stack, 65 tuning, 185 k STAPL, 246 two-phase locking, 171 k Star Superscalar, 232 state separation, 208 U stream graph, 186 unified runtime architecture, 119 StreamIt, 144, 201 unit of transfer, 390 stream parallelism, 263 stream programming, 143, 185 V superscalar processor, 4 video encoding, 281 symmetric multiprocessors, 7 synchronization, 60, 61, 114, 185, 339, 344, W 350, 465 waiting mechanism, 351 synchronization APIs, 358 WebLogic, 363 synchronization overheads, 358 work block, 189 synchronization primitives, 353 work descriptors, 112 synchronous data flow, 143 worksharings, 106 system performance, 364 work unit, 418 write-back policy, 115 T task dependency graph, 113 X task farm, 124 XJava, 187 task graph, 116 XtreemOS, 460 k k k k k.
Load more

Recommended publications
  • Toward Optimised Skeletons for Heterogeneous
    View metadata, citation and similar papers at core.ac.uk brought to you by CORE provided by ROS: The Research Output Service. Heriot-Watt University Edinburgh TOWARD OPTIMISED SKELETONS FOR HETEROGENEOUS PARALLEL ARCHITECTURE WITH PERFORMANCE COST MODEL By Khari A. Armih Submitted for the Degree of Doctor of Philosophy at Heriot-Watt University on Completion of Research in the School of Mathematical and Computer Sciences July, 2013 The copyright in this thesis is owned by the author. Any quotation from the thesis or use of any of the information contained in it must acknowledge this thesis as the source of the quotation or information. I hereby declare that the work presented in this the- sis was carried out by myself at Heriot-Watt University, Edinburgh, except where due acknowledgement is made, and has not been submitted for any other degree. Khari A. Armih (Candidate) Greg Michaelson, Phil Trinder (Supervisors) (Date) ii Abstract High performance architectures are increasingly heterogeneous with shared and distributed memory components, and accelerators like GPUs. Programming such architectures is complicated and performance portability is a major issue as the architectures evolve. This thesis explores the potential for algorithmic skeletons integrating a dynamically parametrised static cost model, to deliver portable performance for mostly regular data parallel programs on heterogeneous archi- tectures. The first contribution of this thesis is to address the challenges of program- ming heterogeneous architectures by providing two skeleton-based programming libraries: i.e. HWSkel for heterogeneous multicore clusters and GPU-HWSkel that enables GPUs to be exploited as general purpose multi-processor devices. Both libraries provide heterogeneous data parallel algorithmic skeletons including hMap, hMapAll, hReduce, hMapReduce, and hMapReduceAll.
    [Show full text]
  • Algorithmic Skeletons and Parallel Design Patterns in Mainstream Parallel Programming
    International Journal of Parallel Programming https://doi.org/10.1007/s10766-020-00684-w(0123456789().,-volV)(0123456789().,-volV) Algorithmic Skeletons and Parallel Design Patterns in Mainstream Parallel Programming 1 1 1 Marco Danelutto • Gabriele Mencagli • Massimo Torquati • 2 3 Horacio Gonza´lez–Ve´lez • Peter Kilpatrick Received: 13 August 2019 / Accepted: 8 October 2020 Ó The Author(s) 2020 Abstract This paper discusses the impact of structured parallel programming methodologies in state-of-the-art industrial and research parallel programming frameworks. We first recap the main ideas underpinning structured parallel programming models and then present the concepts of algorithmic skeletons and parallel design patterns. We then discuss how such concepts have permeated the wider parallel programming community. Finally, we give our personal overview—as researchers active for more than two decades in the parallel programming models and frameworks area—of the process that led to the adoption of these concepts in state-of-the-art industrial and research parallel programming frameworks, and the perspectives they open in relation to the exploitation of forthcoming massively-parallel (both general and special-purpose) architectures. Keywords Algorithmic skeletons Á Parallel design patterns Á High performance computing Á Multi-core architecture Á Parallel computing 1 Introduction In the last two decades, the number of parallel architectures available to the masses has substantially increased. The world has moved from clusters/networks of workstations—composed of individual nodes with a small number of CPUs sharing a common memory hierarchy—to the ubiquitous presence of multi-core CPUs coupled with different many-core accelerators, typically interconnected through high-bandwidth, low-latency networks.
    [Show full text]
  • A Divide-And-Conquer Parallel Skeleton for Unbalanced and Deep Problems
    13th International Symposium on High-Level Parallel Programming and Applications (HLPP 2020) Porto, Portugal July 9-10, 2020 A Divide-and-conquer Parallel Skeleton for Unbalanced and Deep Problems Mill´anA. Mart´ınez Basilio B. Fraguela · · Jos´eC. Cabaleiro Abstract The Divide-and-conquer (D&C) pattern appears in a large number of problems and is highly suitable to exploit parallelism. This has led to much research on its easy and efficient application both in shared and distributed memory parallel systems. One of the most successful approaches explored in this area consists in expressing this pattern by means of parallel skeletons which automate and hide the complexity of the parallelization from the user while trying to provide good performance. In this paper we tackle the develop- ment of a skeleton oriented to the efficient parallel resolution of D&C problems with a high degree of unbalance among the subproblems generated and/or a deep level of recurrence. Our evaluation shows good results achieved both in terms of performance and programmability. Keywords Algorithmic skeletons Divide-and-conquer Template metapro- gramming Load balancing · · · 1 Introduction Divide-and-conquer [1], hence denoted D&C, is a strategy widely used to solve problems whose solution can be obtained by dividing them into subproblems, separately solving those subproblems, and combining their solutions to com- pute the one of the original problem. In this pattern the subproblems have the same nature as the original one, thus this strategy can be recursively ap- plied to them until base cases are found. The independence of the subproblems Mill´anA.
    [Show full text]
  • Towards an Algorithmic Skeleton Framework for Programming the Intel R Xeon Phitm Processor
    Hélder de Almeida Marques Licenciado em Engenharia Informática Towards an Algorithmic Skeleton Framework for Programming the Intel R Xeon PhiTM Processor Dissertação para obtenção do Grau de Mestre em Engenharia Informática Orientador : Hervé Paulino, Prof. Auxiliar, Universidade Nova de Lisboa Júri: Presidente: Prof. Doutor Miguel Pessoa Monteiro Universidade Nova de Lisboa Arguente: Prof. Doutor Nuno Roma Universidade de Lisboa Vogal: Prof. Doutor Hervé Miguel Cordeiro Paulino Universidade Nova de Lisboa Dezembro, 2014 iii Towards an Algorithmic Skeleton Framework for Programming the Intel R Xeon PhiTM Processor Copyright c Hélder de Almeida Marques, Faculdade de Ciências e Tecnologia, Univer- sidade Nova de Lisboa A Faculdade de Ciências e Tecnologia e a Universidade Nova de Lisboa têm o direito, perpétuo e sem limites geográficos, de arquivar e publicar esta dissertação através de ex- emplares impressos reproduzidos em papel ou de forma digital, ou por qualquer outro meio conhecido ou que venha a ser inventado, e de a divulgar através de repositórios científicos e de admitir a sua cópia e distribuição com objectivos educacionais ou de in- vestigação, não comerciais, desde que seja dado crédito ao autor e editor. iv Aos meus pais e amigos que tornaram esta tese possível vi Acknowledgements I would like to start by thanking so much to my adviser Hervé Paulino for all the sup- port, availability and constant meetings throughout the course of this thesis that really helped me a lot to develop this work. I would like to thank the department of Com- puter Sciences of Faculdade de Ciências e Tecnologias da Universidade Nova de Lis- boa, for the facilities and its constant availability to always have a friendly environment to work.
    [Show full text]
  • UNIVERSITY of PISA and SCUOLA SUPERIORE SANTANNA Master
    UNIVERSITY OF PISA AND SCUOLA SUPERIORE SANTANNA Master Degree in Computer Science and Networking Master Thesis Structured programming in MPI A streaming framework experiment Candidate : Leta Melkamu Jifar Supervisor : Prof. Marco Danelluto Academic Year : 2012/2013 To my family, who emphasized the importance of education, who has been my role-model for hard work, persistence and personal sacrifices, and who instilled in me the inspiration to set high goals and the confidence to achieve them. Who has been proud and supportive of my work and who has shared many uncertainties and challenges for completing this thesis and the whole masters program. To Hatoluf, to Dani, to Basho, to Ebo and to mom! Acknowledgments I would like to express my gratitude to my supervisor, Prof. Marco Danelluto, whose expertise, understanding, and patience, added considerably to my graduate experience. I appreciate his vast knowledge and skill in many areas (e.g., parallel and distributed computing, structured Skeleton programming, Multi/Many-Core and Heterogeneous Architectures, Autonomic computing), and his assistance in keeping me in the right track both in the implementation and thesis writing parts. I would also like to mention his humble support in making the experimental machine available and installing all tools that are important for the experiment of this thesis. And I also want to thank him for understanding my situation and be able to available for discus- sion; giving me time from his busy schedule to serve many students, attending conferences and conducting research works. Finally, I would like to thank my family, my classmates and my friends for being supportive and proud.
    [Show full text]
  • A Theoretical Model for Global Optimization of Parallel Algorithms
    mathematics Article A Theoretical Model for Global Optimization of Parallel Algorithms Julian Miller 1,* , Lukas Trümper 1,2 , Christian Terboven 1 and Matthias S. Müller 1 1 Chair for High Performance Computing, IT Center, RWTH Aachen University, 52074 Aachen, Germany; [email protected] (L.T.); [email protected] (C.T.); [email protected] (M.S.M.) 2 Huddly AS, Karenslyst Allé 51, 0279 Oslo, Norway * Correspondence: [email protected] Abstract: With the quickly evolving hardware landscape of high-performance computing (HPC) and its increasing specialization, the implementation of efficient software applications becomes more challenging. This is especially prevalent for domain scientists and may hinder the advances in large-scale simulation software. One idea to overcome these challenges is through software abstraction. We present a parallel algorithm model that allows for global optimization of their synchronization and dataflow and optimal mapping to complex and heterogeneous architectures. The presented model strictly separates the structure of an algorithm from its executed functions. It utilizes a hierarchical decomposition of parallel design patterns as well-established building blocks for algorithmic structures and captures them in an abstract pattern tree (APT). A data-centric flow graph is constructed based on the APT, which acts as an intermediate representation for rich and automated structural transformations. We demonstrate the applicability of this model to three representative algorithms and show runtime speedups between 1.83 and 2.45 on a typical Citation: Miller, J.; Trümper, L.; heterogeneous CPU/GPU architecture. Terboven, C.; Müller, M.S. A Theoretical Model for Global Keywords: parallel programming; parallel patterns; program optimization; optimizing framework; Optimization of Parallel Algorithms.
    [Show full text]
  • Arxiv:2005.04094V1 [Cs.DC] 5 May 2020 National University of Defense Technology E-Mail: {J.Fang, Chunhuang, Taotang84}@Nudt.Edu.Cn Z
    manuscript No. (will be inserted by the editor) Parallel Programming Models for Heterogeneous Many-Cores : A Survey Jianbin Fang · Chun Huang B · Tao Tang · Zheng Wang Received: date / Accepted: date Abstract Heterogeneous many-cores are now an integral part of modern comput- ing systems ranging from embedding systems to supercomputers. While heteroge- neous many-core design offers the potential for energy-efficient high-performance, such potential can only be unlocked if the application programs are suitably par- allel and can be made to match the underlying heterogeneous platform. In this article, we provide a comprehensive survey for parallel programming models for heterogeneous many-core architectures and review the compiling techniques of improving programmability and portability. We examine various software opti- mization techniques for minimizing the communicating overhead between hetero- geneous computing devices. We provide a road map for a wide variety of different research areas. We conclude with a discussion on open issues in the area and po- tential research directions. This article provides both an accessible introduction to the fast-moving area of heterogeneous programming and a detailed bibliography of its main achievements. Keywords Heterogeneous Computing · Many-Core Architectures · Parallel Programming Models 1 Introduction Heterogeneous many-core systems are now commonplace [158,159]. The combina- tion of using a host CPU together with specialized processing units (e.g., GPGPUs, XeonPhis, FPGAs, DSPs and NPUs) has been shown in many cases to achieve orders of magnitude performance improvement. As a recent example, Google's Ten- sor Processing Units (TPUs) are application-specific integrated circuits (ASICs) J. Fang, C. Huang (B), T.
    [Show full text]
  • Skelcl -- a Portable Multi-GPU Skeleton Library
    ANGEWANDTE MATHEMATIK UND INFORMATIK SkelCL – A Portable Multi-GPU Skeleton Library Michel Steuwer, Philipp Kegel, and Sergei Gorlatch 04/10 - I UNIVERSIT A¨ T M UNSTER¨ SkelCL { A Portable Multi-GPU Skeleton Library Michel Steuwer, Philipp Kegel, and Sergei Gorlatch Department of Mathematics and Computer Science University of M¨unster,M¨unster,Germany November 30, 2010 Abstract Modern Graphics Processing Units (GPU) are increasingly used as general- purpose processors. While the two currently most widely used programming models for GPUs, CUDA and OpenCL, are a step ahead as compared to ex- tremely laborious shader programming, they still remain effort-demanding and error-prone. Memory management is complex due to the separate memories of CPU and GPU and GPUs support only a subset of common CPU programming languages. Both, CUDA and OpenCL lack support for multi-GPU systems, such that additional challenges like communication and synchronization arise when using multiple GPUs. We present SkelCL { a novel library for simplifying OpenCL programming of single and multiple GPUs. SkelCL provides for the application programmer two main abstractions: Firstly, a unified memory management that implicitly transfers data transfers between the system's and the GPU's memory. Lazy copying techniques pre- vent unnecessary data transfers and increase performance. Secondly, SkelCL provides so-called algorithmic skeletons to describe computations. Algorithmic skeletons encapsulate parallel implementations of recurring programming pat- terns. Skeletons are configured by providing one or several parameters which adapt a skeleton parallel program pattern according to an application. In ad- dition a set of basic skeletons, SkelCL also provides means to easily compose skeletons, to allow for more complex applications.
    [Show full text]
  • Algorithmic Skeletons and Parallel Design Patterns in Mainstream Parallel Programming
    Algorithmic Skeletons and Parallel Design Patterns in Mainstream Parallel Programming Danelutto, M., Mencagli, G., Torquati, M., González–Vélez, H., & Kilpatrick, P. (2020). Algorithmic Skeletons and Parallel Design Patterns in Mainstream Parallel Programming. International Journal of Parallel Programming. https://doi.org/10.1007/s10766-020-00684-w Published in: International Journal of Parallel Programming Document Version: Publisher's PDF, also known as Version of record Queen's University Belfast - Research Portal: Link to publication record in Queen's University Belfast Research Portal Publisher rights © 2020 The Authors. This is an open access article published under a Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution and reproduction in any medium, provided the author and source are cited. General rights Copyright for the publications made accessible via the Queen's University Belfast Research Portal is retained by the author(s) and / or other copyright owners and it is a condition of accessing these publications that users recognise and abide by the legal requirements associated with these rights. Take down policy The Research Portal is Queen's institutional repository that provides access to Queen's research output. Every effort has been made to ensure that content in the Research Portal does not infringe any person's rights, or applicable UK laws. If you discover content in the Research Portal that you believe breaches copyright or violates any law,
    [Show full text]
  • Nested Parallelism with Algorithmic Skeletons A
    NESTED PARALLELISM WITH ALGORITHMIC SKELETONS A Thesis by ALIREZA MAJIDI Submitted to the Office of Graduate and Professional Studies of Texas A&M University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Chair of Committee, Lawrence Rauchwerger Committee Members, Timothy Davis Diego Donzis Head of Department, Scott Schaefer August 2020 Major Subject: Computer Science and Engineering Copyright 2020 Alireza Majidi ABSTRACT New trend in design of computer architectures, from memory hierarchy design to grouping computing units in different hierarchical levels in CPUs, pushes developers toward algorithms that can exploit these hierarchical designs. This trend makes support of nested-parallelism an impor- tant feature for parallel programming models. It enables implementation of parallel programs that can then be mapped onto the system hierarchy. However, supporting nested-parallelism is not a trivial task due to complexity in spawning nested sections, destructing them and more importantly communication between these nested parallel sections. Structured parallel programming models are proven to be a good choice since while they hide the parallel programming complexities from the programmers, they allow programmers to customize the algorithm execution without going through radical changes to the other parts of the program. In this thesis, nested algorithm com- position in the STAPL Skeleton Library (SSL) is presented, which uses a nested dataflow model as its internal representation. We show how a high level program specification using SSL allows for asynchronous computation and improved locality. We study both the specification and per- formance of the STAPL implementation of Kripke, a mini-app developed by Lawrence Livermore National Laboratory.
    [Show full text]
  • Algorithmic Skeletons for Python Jolan Philippe, Frédéric Loulergue
    PySke: Algorithmic Skeletons for Python Jolan Philippe, Frédéric Loulergue To cite this version: Jolan Philippe, Frédéric Loulergue. PySke: Algorithmic Skeletons for Python. The 2019 Interna- tional Conference on High Performance Computing & Simulation (HPCS), Jul 2019, Dublin, Ireland. 10.1109/HPCS48598.2019.9188151. hal-02317127 HAL Id: hal-02317127 https://hal.archives-ouvertes.fr/hal-02317127 Submitted on 15 Oct 2019 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. PySke: Algorithmic Skeletons for Python Jolan Philippe Fred´ eric´ Loulergue School of Informatics Computing and Cyber Systems School of Informatics Computing and Cyber Systems Northern Arizona University Northern Arizona University Flagstaff, AZ, USA Flagstaff, AZ, USA [email protected] [email protected] Abstract—PySke is a library of parallel algorithmic skeletons b) Approach and Goals: To ease the development of in Python designed for list and tree data structures. Such parallel programming, Murray Cole has introduced the skeletal algorithmic skeletons are high-order functions implemented in parallelism approach [1]. A skeleton is a parallel implementa- parallel. An application developed with PySke is a composition of skeletons. To ease the write of parallel programs, PySke does tion of a computation pattern.
    [Show full text]
  • Algorithmic Skeletons for Exact Combinatorial Search at Scale
    Archibald, Blair (2018) Algorithmic skeletons for exact combinatorial search at scale. PhD thesis. https://theses.gla.ac.uk/31000/ Copyright and moral rights for this work are retained by the author A copy can be downloaded for personal non-commercial research or study, without prior permission or charge This work cannot be reproduced or quoted extensively from without first obtaining permission in writing from the author The content must not be changed in any way or sold commercially in any format or medium without the formal permission of the author When referring to this work, full bibliographic details including the author, title, awarding institution and date of the thesis must be given Enlighten: Theses https://theses.gla.ac.uk/ [email protected] ALGORITHMIC SKELETONS FOR EXACT COMBINATORIAL SEARCH AT SCALE BLAIR ARCHIBALD SUBMITTED IN FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF Doctor of Philosophy SCHOOL OF COMPUTING SCIENCE COLLEGE OF SCIENCE AND ENGINEERING UNIVERSITY OF GLASGOW NOVEMBER 4, 2018 © BLAIR ARCHIBALD Abstract Exact combinatorial search is essential to a wide range of application areas including constraint optimisation, graph matching, and computer algebra. Solutions to combinatorial problems are found by systematically exploring a search space, either to enumerate solutions, determine if a specific solution exists, or to find an optimal solution. Combinatorial searches are computationally hard both in theory and practice, and efficiently exploring the huge number of combinations is a real challenge, often addressed using approximate search algorithms. Alternatively, exact search can be parallelised to reduce execution time. However, parallel search is challenging due to both highly irregular search trees and sensitivity to search order, leading to anomalies that can cause unexpected speedups and slowdowns.
    [Show full text]