检索结果-内蒙古大学图书馆

out-of-core computation of the QR Factorization on Multi-core Processors

15th International Euro-Par Conference on Parallel Computing

作者： Marques, Mercedes Quintana-Orti, Gregorio Quintana-Orti, Enrique S. van de Geijn, Robert Univ Jaume 1 Depto Ingn & Ciencia Comp Castellon de La Plana 12071 Spain Univ Texas Austin Dept Comp Sci Austin TX 78712 USA

ISBN: (纸本)9783642038686

We target the development of high-performance algorithms for dense matrix operations where data resides on disk and bias to be explicitly moved in and out of the main memory. We provide strong evidence that, even for a complex operation like the QR, factorization, the use of a run-time system creates a separation of concerns between the matrix computations and I/O operations with the result that no significant changes need to be introduced to existing in-core algorithms. The library developer can thus focus on the design of algorithms-by-blocks, addressing disk memory as just;another level of the memory hierarchy. Experimental results for the out-of-core computation of the QR factorization on a multi-core processor reveal the potential of this approach.

关键词： Dense linear algebra out-of-core computation QR factorization multi-core processors high performance

来源：评论

学校读者我要写书评

暂无评论

A Data-Centric Directive-Based Framework to Accelerate out-of-core Stencil computation on a GPU

引用

IEICE TRANSACTIONS ON INFORMATION AND SYSTEMS 2020年第12期E103D卷 2421-2434页

作者： Shen, Jingcheng Ino, Fumihiko Farres, Albert Hanzich, Mauricio Osaka Univ Grad Sch Informat Sci & Technol Suita Osaka 5650871 Japan Barcelona Supercomp Ctr Barcelona 08034 Spain

Graphics processing units (GPUs) are highly efficient architectures for parallel stencil code;however, the small device (i.e., GPU) memory capacity (several tens of GBs) necessitates the use of out-of-core computation to process excess data. Great programming effort is needed to manually implement efficient out-of-core stencil code. To relieve such programming burdens, directive-based frameworks emerged, such as the pipelined accelerator (PACC);however, they usually lack specific optimizations to reduce data transfer. In this paper, we extend PACC with two data-centric optimizations to address data transfer problems. The first is a direct-mapping scheme that eliminates host (i.e., CPU) buffers, which intermediate between the original data and device buffers. The second is a region-sharing scheme that significantly reduces host-to-device data transfer. The extended PACC was applied to an acoustic wave propagator, automatically extending the length of original serial code 2.3-fold to obtain the out-of-core code. Experimental results revealed that on a Tesla V100 GPU, the generated code ran 41.0, 22.1, and 3.6 times as fast as implementations based on Open Multi-Processing (OpenMP), Unified Memory, and the previous PACC, respectively. The generated code also demonstrated usefulness with small datasets that fit in the device capacity, running 1.3 times as fast as an in-core implementation.

关键词： stencil computation out-of-core computation data-centric optimizations GPU

来源：评论

学校读者我要写书评

暂无评论

The out-of-core KNN Awakens: The Light Side of computation Force on Large Datasets 4th

The Out-of-core KNN Awakens: The Light Side of Computation F...

引用

4th International Conference on Networked Systems (NETYS)

作者： Chiluka, Nitin Kermarrec, Anne-Marie Olivares, Javier Inria Rennes France

ISBN: (纸本)9783319461403;9783319461397

K-Nearest Neighbors (KNN) is a crucial tool for many applications, e.g. recommender systems, image classification and web-related applications. However, KNN is a resource greedy operation particularly for large datasets. We focus on the challenge of KNN computation over large datasets on a single commodity PC with limited memory. We propose a novel approach to compute KNN on large datasets by leveraging both disk and main memory efficiently. The main rationale of our approach is to minimize random accesses to disk, maximize sequential accesses to data and efficient usage of only the available memory. We evaluate our approach on large datasets, in terms of performance and memory consumption. The evaluation shows that our approach requires only 7% of the time needed by an in-memory baseline to compute a KNN graph.

关键词： K-nearest neighbors out-of-core computation Graph processing

来源：评论

学校读者我要写书评

暂无评论

Cache-Aware out-of-core Tensor Decomposition on GPUs

引用

JOURNAL OF INFORMATION SCIENCE AND ENGINEERING 2018年第6期34卷 1383-1403页

作者： Tsai, Yu-Ting Wang, Wei-Jhih Kao, Tzu-Yuan Yuan Ze Univ Dept Comp Sci & Engn Taoyuan 320 Taiwan

For compressing large-scale multidimensional datasets, out-of-core tensor decomposition often consumes a lot of time. This article particularly presents a method based on two key ideas to improve its performance. First, cache-aware static scheduling schemes are employed to reduce the total number of disk accesses. Second, we take advantage of the massively parallel computing power and large memory size of modern GPUs to accelerate linear algebra operations of tensor decomposition. Our experiments demonstrate that the proposed method can achieve speedups of 11-16 over a naive implementation and 2.5-5.3 over previous work [43] for practical data-driven rendering applications.

关键词： data-driven photorealistic rendering multidimensional data analysis tensor decomposition out-of-core computation general-purpose GPU computing

来源：评论

学校读者我要写书评

暂无评论

Reducing the amount of out-of-core data access for GPU-accelerated randomized SVD

引用

CONCURRENCY AND computation-PRACTICE & EXPERIENCE 2020年第19期32卷 e5754-e5754页

作者： Lu, Yuechao Yamazaki, Ichitaro Ino, Fumihiko Matsushita, Yasuyuki Tomov, Stanimire Dongarra, Jack Osaka Univ Grad Sch Informat Sci & Technol Osaka Japan Univ Tennessee Innovat Comp Lab Knoxville TN USA

We propose two acceleration methods, namely, Fused and Gram, for reducing out-of-core data access when performing randomized singular value decomposition (RSVD) on graphics processing units (GPUs). out-of-core data here are data that are too large to fit into the GPU memory at once. Both methods accelerate GPU-enabled RSVD using the following three schemes: (1) a highly tuned general matrix-matrix multiplication (GEMM) scheme for processing out-of-core data on GPUs;(2) a data-access reduction scheme based on one-dimensional data partition;and (3) a first-in, first-out scheme that reduces CPU-GPU data transfer using the reverse iteration. The Fused method further reduces the amount of out-of-core data access by merging two GEMM operations into a single operation. By contrast, the Gram method reduces both in-core and out-of-core data access by explicitly forming the Gram matrix. According to our experimental results, the Fused and Gram methods improved the RSVD performance up to 1.7x and 5.2x, respectively, compared with a straightforward method that deploys schemes (1) and (2) on the GPU. In addition, we present a case study of deploying the Gram method for accelerating robust principal component analysis, a convex optimization problem in machine learning.

关键词： divide and conquer GPU out-of-core computation singular value decomposition

来源：评论

学校读者我要写书评

暂无评论

MPI windows on storage for HPC applications

引用

PARALLEL COMPUTING 2018年 77卷 38-56页

作者： Rivas-Gomez, Sergio Gioiosa, Roberto Peng, Ivy Bo Kestor, Gokcen Narasimhamurthy, Sai Laure, Erwin Markidis, Stefano KTH Royal Inst Technol S-10044 Stockholm Sweden Oak Ridge Natl Lab Oak Ridge TN 37830 USA Seagate Syst UK Havant PO9 1SA England

Upcoming HPC clusters will feature hybrid memories and storage devices per compute node. In this work, we propose to use the MPI one-sided communication model and MPI windows as unique interface for programming memory and storage. We describe the design and implementation of MPI storage windows, and present its benefits for out-of-core execution, parallel I/O and fault-tolerance. In addition, we explore the integration of heterogeneous window allocations, where memory and storage share a unified virtual address space. When performing large, irregular memory operations, we verify that MPI windows on local storage incurs a 55% performance penalty on average. When using a Lustre parallel file system, "asymmetric" performance is observed with over 90% degradation in writing operations. Nonetheless, experimental results of a Distributed Hash Table, the HACC I/O kernel mini-application, and a novel MapReduce implementation based on the use of MPI one-sided communication, indicate that the overall penalty of MPI windows on storage can be negligible in most cases in real-world applications. (C) 2018 Elsevier B.V. All rights reserved.

关键词： MPI windows on storage out-of-core computation Parallel I/O

来源：评论

学校读者我要写书评

暂无评论

Key concepts for parallel out-of-core LU factorization

引用

PARALLEL COMPUTING 1997年第1-2期23卷 49-70页

作者： Dongarra, JJ Hammarling, S Walker, DW NAG LTD OXFORD OX2 8DRENGLAND OAK RIDGE NATL LAB MATH SCI SECTOAK RIDGETN 37831

This paper considers key ideas in the design of out-of-core dense LU factorization routines. A left-looking variant of the LU factorization algorithm is shown to require less I/O to disk than the right-looking variant, and is used to develop a parallel, out-of-core implementation, This implementation makes use of a small library of parallel I/O routines, together with ScaLAPACK and PBLAS routines, Results for runs on an Intel Paragon are presented and interpreted using a simple performance model.

关键词： out-of-core computation LU factorization parallel computing parallel I/O

来源：评论

学校读者我要写书评

暂无评论

A parallel programming interface for out-of-core cluster applications

引用

CLUSTER COMPUTING-THE JOURNAL OF NETWORKS SOFTWARE TOOLS AND APPLICATIONS 2006年第3期9卷 321-327页

作者： Tang, Jianqi Fang, Binxing Hu, Mingzeng Zhang, Hongli Harbin Inst Technol Dept Comp Sci & Engn Harbin 150006 Peoples R China

Clusters of workstations are a practical approach to parallel computing that provide high performance at a low cost for many scientific and engineering applications. In order to handle problems with increasing data sets, methods supporting parallel out-of-core computations must be investigated. Since writing an out-of-core version of a program is a difficult task and virtual memory systems do not perform well in some cases, we have developed a parallel programming interface and the support library to provide efficient and convenient access to the out-of-core data. This paper focuses on how these components extend the range of problem sizes that can be solved on the cluster of workstations. Execution time of Jacobi iteration when using our interface, virtual memory and PVFS are compared to characterize the performance for various problem sizes, and it is concluded that our new interface significantly increases the sizes of problems that can be efficiently solved.

关键词： out-of-core computation cluster of workstations global out-of-core array local array file in-core data

来源：评论

学校读者我要写书评

暂无评论

An object-oriented method for out-of-core parallel computations on cluster of workstations

An object-oriented method for out-of-core parallel computati...

引用

4th International Conference on Parallel and Distributed Computing, Applications and Technologies (PDCAT 2003)

作者： Tang, JQ Fang, BX Hu, MZ Zhang, HL Harbin Inst Technol Sch Comp Sci & Technol Harbin 150001 Peoples R China

ISBN: (纸本)0780378407

Many Applications in different fields need to access large amounts of data. The difficulty of handling out-of-core data limits the performance of supercomputers. This paper first describes an out-of-core computation model on cluster of workstations. Then an object-oriented method is offered, which makes the out-of-core programs easier to write. Some optimal strategies, such as distribution of Local Array Files, data sieving and prefetching, also make an improvement in the performance of out-of-core parallel computations.

关键词： out-of-core computation out-of-core array cluster of workstations (COW)

来源：评论

学校读者我要写书评

暂无评论

An out-of-core Branch and Bound Method for Solving the 0-1 Knapsack Problem on a GPU 17th

An Out-of-Core Branch and Bound Method for Solving the 0-1 K...

引用

17th International Conference on Algorithms and Architectures for Parallel Processing (ICA3PP)

作者： Shen, Jingcheng Shigeoka, Kentaro Ino, Fumihiko Hagihara, Kenichi Osaka Univ Grad Sch Informat Sci & Technol 1-5 Yamadaoka Suita Osaka 5650871 Japan Hitachi High Technol Corp Minato Ku 14-24-1 Nishishinjuku Tokyo 1058717 Japan

ISBN: (纸本)9783319654829;9783319654812

In this paper, we propose an out-of-core branch and bound (B&B) method for solving the 0-1 knapsack problem on a graphics processing unit (GPU). Given a large problem that produces many subproblems, the proposed method dynamically swaps subproblems out to CPU memory. We adopt two strategies to realize this swapping-out procedure with minimum amount of CPU-GPU data transfer. The first strategy is a GPU-based stream compaction strategy that reduces the sparseness of arrays. The second strategy is a double buffering strategy that hides the data transfer overhead by overlapping data transfer with GPU-based B&B operations. Experimental results show that the proposed method can store 33.7 times more subproblems than the previous method, solving twice more instances on the GPU. As for the stream compaction strategy, an input-output separated scheme runs 13.1% faster than an input-output unified scheme.

关键词： out-of-core computation Branch and bound Knapsack GPU

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：