检索结果-内蒙古大学图书馆

RaftLib: A C plus plus template library for high performance stream parallel processing

INTERNATIONAL JOURNAL OF HIGH PERFORMANCE COMPUTING APPLICATIONS 2017年第5期31卷 391-404页

作者： Beard, Jonathan C. Li, Peng Chamberlain, Roger D. ARM Res 5707 Southwest Pkwy Suite 100 Austin Austin TX 78735 USA Amazon Inc Seattle WA USA Washington Univ Dept Comp Sci & Engn St Louis MO 63130 USA

Stream processing is a compute paradigm that has been around for decades, yet until recently has failed to garner the same attention as other mainstream languages and libraries (e.g. C++, OpenMP, MPI). Stream processing has great promise: the ability to safely exploit extreme levels of parallelism to process huge volumes of streaming data. there have been many implementations, both libraries and full languages. the full languages implicitly assume that the streaming paradigm cannot be fully exploited in legacy languages, while library approaches are often preferred for being integrable with the vast expanse of extant legacy code. Libraries, however are often criticized for yielding to the shape of their respective languages. RaftLib aims to fully exploit the stream processing paradigm, enabling a full spectrum of streaming graph optimizations, while providing a platform for the exploration of integrability with legacy C/C++ code. RaftLib is built as a C++ template library, enabling programmers to utilize the robust C++ standard library, and other legacy code, along with RaftLib's parallelization framework. RaftLib supports several online optimization techniques: dynamic queue optimization, automatic parallelization, and real-time low overhead performance monitoring.

关键词： Stream processing big-data C plus plus template library high performance computing RaftLib performance monitoring parallel processing

来源：评论

学校读者我要写书评

暂无评论

Lifeline-based Global Load Balancing 11

Lifeline-based Global Load Balancing

引用

16th acm symposium on principles and practice of parallel programming

作者： Saraswat, Vijay Kambadur, Prabhanjan Kodali, Sreedhar Grove, David Krishnamoorthy, Sriram Pacific NW Natl Lab Richland WA 99352 USA

ISBN: (纸本)9781450301190

On shared-memory systems, Cilk-style work-stealing [5] has been used to effectively parallelize irregular task-graph based applications such as Unbalanced Tree Search (UTS) [24, 28]. there are two main difficulties in extending this approach to distributed memory. In the shared memory approach, thieves (nodes without work) constantly attempt to asynchronously steal work from randomly chosen victims until they find work. In distributed memory, thieves cannot autonomously steal work from a victim without disrupting its execution. When work is sparse, this results in performance degradation. In essence, a direct extension of traditional work-stealing to distributed memory violates the work-first principle underlying work-stealing. Further, thieves spend useless CPU cycles attacking victims that have no work, resulting in system inefficiencies in multi-programmed contexts. Second, it is non-trivial to detect active distributed termination (detect that programs at all nodes are looking for work, hence there is no work). this problem is well-studied and requires careful design for good performance. Unfortunately, in most existing languages/frameworks, application developers are forced to implement their own distributed termination detection. In this paper, we develop a simple set of ideas that allow work-stealing to be efficiently extended to distributed memory. First, we introduce lifeline graphs: low-degree, low-diameter, fully-connected directed graphs. Such graphs can be constructed from k-dimensional hypercubes. When a node is unable to find work after w unsuccessful steals, it quiesces after informing the outgoing edges in its lifeline graph. Quiescent nodes do not disturb other nodes. A quiesced node is reactivated when work arrives from a lifeline, and itself shares this work with those of its incoming lifelines that are activated. Termination occurs precisely when computation at all nodes has quiesced. In a language such as X10, such passive distributed terminati

关键词： UTS global load balancing distributed computing X10 work-stealing parallel programming Algorithms Design

来源：评论

学校读者我要写书评

暂无评论

Resilient X10: Efficient failure-aware programming 14

Resilient X10: Efficient failure-aware programming

引用

2014 19th acm sigplan symposium on principles and practice of parallel programming, PPoPP 2014

作者： Cunningham, David Grove, David Herta, Benjamin Iyengar, Arun Kawachiya, Kiyokuni Murata, Hiroki Saraswat, Vijay Takeuchi, Mikio Tardieu, Olivier IBM T. J. Watson Research Center Japan Google Inc. Japan IBM Research Tokyo Japan

ISBN: (纸本)9781450326568

Scale-out programs run on multiple processes in a cluster. In scale-out systems, processes can fail. Computations using traditional libraries such as MPI fail when any component process fails. the advent of Map Reduce, Resilient Data Sets and MillWheel has shown dramatic improvements in productivity are possible when a high-level programming framework handles scale-out and resilience automatically. We are concerned with the development of generalpurpose languages that support resilient programming. In this paper we show how the X10 language and implementation can be extended to support resilience. In Resilient X10, places may fail asynchronously, causing loss of the data and tasks at the failed place. Failure is exposed through exceptions. We identify a Happens Before Invariance Principle and require the runtime to automatically repair the global control structure of the program to maintain this principle. We show this reduces much of the burden of resilient programming. the programmer is only responsible for continuing execution with fewer computational resources and the loss of part of the heap, and can do so while taking advantage of domain knowledge. We build a complete implementation of the language, capable of executing benchmark applications on hundreds of nodes. We describe the algorithms required to make the language runtime resilient. We then give three applications, each with a different approach to fault tolerance (replay, decimation, and domain-level checkpointing). these can be executed at scale and survive node failure. We show that for these programs the overhead of resilience is a small fraction of overall runtime by comparing to equivalent non-resilient X10 programs. On one program we show end-to-end performance of Resilient X10 is ∼100x faster than Hadoop. Copyright © 2014 acm.

关键词： Fault tolerance

来源：评论

学校读者我要写书评

暂无评论

Teaching parallel computing to science faculty: Best practices and common pitfalls

Teaching parallel computing to science faculty: Best practic...

引用

2006 acm sigplan symposium on principles and practice of parallel programming, PPOPP'06

作者： Joiner, David A. Gray, Paul Murphy, thomas Peck, Charles Kean University 1000 Morris Ave Union NJ 07083 University of Northern Iowa 1227 West 27th St Cedar Falls IA 50614 Contra Costa College 2600 Mission Bell Dr San Pablo CA 94806 Earlham College 801 National Rd West Richmond IN 47374

ISBN: (纸本)1595931899

In 2002, we first brought High Performance Computing (HPC) methods to the college classroom as a way to enrich Computational Science education. through the years, we have continued to facilitate college faculty in science, technology, engineering, and mathematics (STEM) disciplines to stay current with HPC methodologies. We have accomplished this by designing and delivering faculty workshops, hosted in a variety of lab settings, as well as by developing tools supporting the technical infrastructure necessary for HPC education, all this without requiring access to traditional HPC computing platforms. In all, we have so far presented 16 professional development workshops for close to 400 predominantly undergraduate STEM faculty. this paper presents the result of internal formative evaluation by workshop instructors and the materials and tools developed during that process. We did this work as part of the National Computational Science Institute (NCSI) and in collaboration with the following groups: the Minority Serving Institutions - High Performance Computing (MSI-HPC) program of the National Computational Science Alliance the Consortium for Computing Sciences in Colleges (CCSC) the Center for Excellence in High Performance Computing the Oklahoma University Supercomputing symposium series the Super Computing (SC) conference series education program We presented learners with a sequence of interactive, "run it, modify it, build it" open-ended lab exercises drawn from a variety of disciplines. Interactivity means having the ability to change parallel and algorithmic parameters, e.g. running software on more than one machine, using different models, refining the model, changing the problem scale, using different parallel algorithms. there is a lack of scientific parallel curricula suitable for illustrating Computational Science principles in the classroom. We addressed this need by locating, and where necessary creating, suitable open source software, data-sets, and cur

关键词： parallel processing systems

来源：评论

学校读者我要写书评

暂无评论

A programming Model for GPU Load Balancing 23

A Programming Model for GPU Load Balancing

引用

28th acm sigplan Annual symposium on principles and practice of parallel programming, PPoPP 2023

作者： Osama, Muhammad Porumbescu, Serban D. Owens, John D. University of California DavisCA United States

ISBN: (纸本)9798400700156

We propose a GPU fine-grained load-balancing abstraction that decouples load balancing from work processing and aims to support both static and dynamic schedules with a programmable interface to implement new load-balancing schedules. Prior to our work, the only way to unleash the GPU's potential on irregular problems has been to workload-balance through application-specific, tightly coupled load-balancing techniques. With our open-source framework for load-balancing, we hope to improve programmers' productivity when developing irregular-parallel algorithms on the GPU, and also improve the overall performance characteristics for such applications by allowing a quick path to experimentation with a variety of existing load-balancing techniques. Consequently, we also hope that by separating the concerns of load-balancing from work processing within our abstraction, managing and extending existing code to future architectures becomes easier. © 2023 Owner/Author.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Stack-Based parallel Recursion on Graphics Processors

引用

acm sigplan NOTICES 2009年第4期44卷 299-300页

作者： Yang, Ke He, Bingsheng Luo, Qiong Sander, Pedro V. Shi, Jiaoying Zhejiang Univ Hangzhou Zhejiang Peoples R China

Recent research has shown promising results on using graphics processing units (GPUs) to accelerate general-purpose computation. However, today's GPUs do not support recursive functions. As a result, for inherently recursive algorithms such as tree traversal, GPU programmers need to explicitly use stacks to emulate the recursion. parallelizing such stack-based implementation on the GPU increases the programming difficulty;moreover, it is unclear how to improve the efficiency of such parallel implementations. As a first step to address both ease of programming and efficiency issues, we propose three parallel stack implementation alternatives that differ in the granularity of stack sharing. Taking tree traversals as an example, we study the performance tradeoffs between these alternatives and analyze their behaviors in various situations. Our results could be useful to both GPU programmers and GPU compiler writers.

关键词： Algorithms Languages Stack parallel Recursion Graphics Processors

来源：评论

学校读者我要写书评

暂无评论

Reaching bandwidth saturation using transparent injection parallelization

引用

INTERNATIONAL JOURNAL OF HIGH PERFORMANCE COMPUTING APPLICATIONS 2017年第5期31卷 405-421页

作者： Chaimov, Nicholas Ibrahim, Khaled Z. Williams, Samuel Iancu, Costin Univ Oregon Eugene OR 97403 USA Lawrence Berkeley Natl Lab Computat Res Div Berkeley CA USA Lawrence Berkeley Natl Lab Performance & Algorithms Res Grp Berkeley CA USA

Although logically available, applications may not exploit enough instantaneous communication concurrency to maximize network utilization on HPC systems. this is exacerbated in hybrid programming models that combine single program multiple data with OpenMP or CUDA. We present the design of a multi-threaded runtime able to transparently increase the instantaneous network concurrency and to provide near saturation bandwidth, independent of the application configuration and dynamic behavior. the runtime offloads communication requests from application level tasks to multiple communication servers. the servers use system specific performance models to attain network saturation. Our techniques alleviate the need for spatial and temporal application level message concurrency optimizations. Experimental results show improved message throughput and bandwidth by as much as 150% for 4 KB messages on InfiniBand and by as much as 120% for 4 KB messages on Cray Aries. For more complex operations such as all-to-all collectives, we observe as much as 30% speedup. this translates into 23% speedup on 12,288 cores for a NAS FT implemented using FFTW. We observe as much as 76% speedup on 1500 cores for an already optimized UPC+OpenMP geometric multigrid application using hybrid parallelism. For the geometric multigrid GPU implementation, we observe as much as 44% speedup on 512 GPUs.

关键词： Communication concurrency multi-threaded runtime one sided communication parallel injection UPC

来源：评论

学校读者我要写书评

暂无评论

Certified Compilation of Financial Contracts 18

Certified Compilation of Financial Contracts

引用

20th International symposium on principles and practice of Declarative programming (PPDP)

作者： Annenkov, Danil Elsman, Martin INRIA GALLINETTE Res Ream Nantes France Univ Copenhagen Dept Comp Sci DIKU Copenhagen Denmark

ISBN: (纸本)9781450364416

We present an extension to a certified financial contract management system that allows for templated declarative financial contracts and for integration with financial stochastic models through verified compilation into so-called payoff-expressions. Such expressions readily allow for determining the value of a contract in a given evaluation context, such as contexts created for stochastic simulations. the templating mechanism is useful both at the contract specification level, for writing generic reusable contracts, and for reuse of code that, without the templating mechanism, needs to be recompiled for different evaluation contexts. We report on the effect of using the certified system in the context of a GPGPU-based Monte Carlo simulation engine for pricing various over-the-counter (OTC) financial contracts. the full contract-management system, including the payoff-language compilation, is verified in the Coq proof assistant and certified Haskell code is extracted from our Coq development along with Futhark code for use in a data-parallel pricing engine.

关键词： financial contracts contract languages domain-specific languages certified programming software correctness Coq

来源：评论

学校读者我要写书评

暂无评论

Decentralized execution of constraint handling rules for ensembles 13

Decentralized execution of constraint handling rules for ens...

引用

15th symposium on principles and practice of Declarative programming, PPDP 2013

作者： Lam, Edmund S. L. Cervesato, Iliano Carnegie Mellon University United States

ISBN: (纸本)9781450321549

CHR is a declarative, concurrent and committed choice rule-based constraint programming language. In this paper, we adapt CHR to provide a decentralized execution model for parallel and distributed programs. Specifically, we consider an execution model consisting of an ensemble of computing entities, each with its own constraint store and each capable of communicating with its neighbors. We extend CHR into CHRe, in which rules are executed at a location and are allowed to access the constraint store of its immediate neighbors. We give an operational semantics for CHRe, denoted ω 0e, that defines incremental and asynchronous decentralized rewriting for the class of CHRe rules characterized by purely local matching (0-neighbor restricted rules). We show the soundness of the ω0e semantics with respect to the abstract CHR semantics. We then give a safe encoding of the more general 1-neighbor restricted rules as 0-neighbor restricted rules, and discuss how this encoding can be generalized to all CHRe programs. © 2013 acm.

关键词： Constraint programming

来源：评论

学校读者我要写书评

暂无评论

Prolog programming with a map-reduce parallel construct 13

Prolog programming with a map-reduce parallel construct

引用

15th symposium on principles and practice of Declarative programming, PPDP 2013

作者： Côrte-Real, Joana Dutra, Inês Rocha, Ricardo CRACS INESC TEC University of Porto Rua Campo Alegre 1021/1055 4169-007 Porto Portugal

ISBN: (纸本)9781450321549

Map-Reduce is a programming model that has its roots in early functional programming. In addition to producing short and elegant code for problems involving lists or collections, this model has proven very useful for large-scale highly parallel data processing. In this work, we present the design and implementation of a high-level parallel construct that makes the Map-Reduce programming model available for Prolog programmers. To the best of our knowledge, there is no Map-Reduce framework native to Prolog, and so the aim of this work is to offer data processing features from which several applications can greatly benefit;the Inductive Logic programming field, for instance, can take advantage of a Map-Reduce predicate when proving newly created rules against sets of examples. Our Map-Reduce model was comprehensively tested with different applications. Our experiments, using the Yap Prolog system, show that: (i) the model scales linearly up to 24 processors;(ii) a dynamic distributed scheduling strategy performs better than centralized or static scheduling strategies;and (iii) the performance varies significantly with the number of items being sent to each processor at a time. Overall, our Map-Reduce framework presents as a good alternative for both taking advantage of the currently available low cost multi-core architectures and developing scalable data processing applications, native to the Prolog programming language. © 2013 acm.

关键词： Computer circuits

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：