检索结果-内蒙古大学图书馆

A Repair of Workflow Models Based on Mirroring Matrices

INTERNATIONAL JOURNAL OF parallel PROGRAMMING 2017年第4期45卷 1001-1020页

作者： Sun, Yanan Du, Yuyue Li, Maozhen Shandong Univ Sci & Technol Coll Informat Sci & Engn Qingdao 266590 Peoples R China Brunel Univ London Dept Elect & Comp Engn Uxbridge UB8 3PH Middx England

Process models can be discovered from event logs and conformance checking techniques can be used to detect and diagnose the differences between observed and modeled behaviors. Some existing conformance checking techniques have been used for repairing a process model, but there are also many techniques can uncover these differences efficiently. So we can use these techniques to repair a process model. In this paper we research the problem of repairing the workflow models w.r.t. the corresponding real process has been changed and the change is adding new activities. To solve this problem, a mirroring matrix is presented based on the variant of footprints and parallel programs. The mirroring matrix is a uniform standard to compare an original model with a complete event log gained from the corresponding real process. Based on this information, the differences between a model and a log can be found. We suggest four algorithms that make model repair realized. Finally, the effectiveness and correctness of the proposed techniques are illustrated via contrastive experiments and analysis.

关键词： Conformance checking Model repair Mirroring matrix Workflow net Event log parallel programs

来源：评论

学校读者我要写书评

暂无评论

SWattention: designing fast and memory-efficient attention for a new Sunway Supercomputer

引用

JOURNAL OF SUPERCOMPUTING 2024年第10期80卷 13657-13680页

作者： Wu, Ruohan Zhu, Xianyu Chen, Junshi Liu, Sha Zheng, Tianyu Liu, Xin An, Hong Univ Sci & Technol China Sch Comp Sci & Technol Hefei Peoples R China Zhejiang Lab Hangzhou Peoples R China Natl Supercomp Ctr Wuxi Wuxi Peoples R China

In the past few years, Transformer-based large language models (LLM) have become the dominant technology in a series of applications. To scale up the sequence length of the Transformer, FlashAttention is proposed to compute exact attention with reduced memory requirements and faster execution. However, implementing the FlashAttention algorithm on the new generation Sunway Supercomputer faces many constraints such as the unique heterogeneous architecture and the limited memory bandwidth. This work proposes SWattention, a highly efficient method for computing the exact attention on the SW26010pro processor. To fully utilize the 6 core groups (CG) and 64 cores per CG on the processor, we design a two-level parallel task partition strategy. Asynchronous memory access is employed to ensure that memory access overlaps with computation. Additionally, a tiling strategy is introduced to determine optimal SRAM block sizes. Compared with the standard attention, SWattention achieves around 2.0x speedup for FP32 training and 2.5x speedup for mixed-precision training. The sequence lengths range from 1k to 8k and scale up to 16k without being out of memory. As for the end-to-end performance, SWattention achieves up to 1.26x speedup for training GPT-style models, which demonstrates that SWattention enables longer sequence length for LLM training.

关键词： Artificial intelligence Attention optimization parallel programs Sunway architecture

来源：评论

学校读者我要写书评

暂无评论

Energy consumption model in multicore architectures with variable frequency

引用

JOURNAL OF SUPERCOMPUTING 2021年第3期77卷 2458-2485页

作者： Meneses-Viveros, Amilcar Paredes-Lopez, Mireya Hernandez-Rubio, Erika Gitler, Isidoro Ctr Res & Adv Studies Dept Comp Sci Mexico City DF Mexico Univ Americas Puebla Cholula Mexico Inst Politecn Nacl SEPI ESCOM Mexico City DF Mexico Ctr Res & Adv Studies Dept Math Mexico City DF Mexico

Models extending Amdahl's law have been developed to study the behavior of parallel programs energy consumption. In addition, it has been shown that energy consumption of those programs also relies on the layout of the resources on the chip, such as power supply. Other extensions over Amdahl's law have been conducted to study the behavior of parallel programs speedup for frequency variable processors. Previous models have focused on the use of Turbo Boost in the parallel regions of a program, without considering that Turbo Boost also affects the sequential regions. Hence, we present a model to analyze energy consumption of parallel programs executed on Intel multicore processors with Turbo Boost frequencies to cover this gap. The model is an extension to Amdahl's law, and it is validated with a double-precision matrix multiplication running on Intel multicore processors that enable Turbo Boost technology.

关键词： Amdahl's law Energy model Multicore processors parallel programs Turbo Boost Variable frequency

来源：评论

学校读者我要写书评

暂无评论

A test suite for high-performance parallel Java

引用

ADVANCES IN ENGINEERING SOFTWARE 2000年第8-9期31卷 687-696页

作者： Häuser, J Ludewig, T Williams, RD Winkelmann, R Gollnick, T Brunett, S Muylaert, J Ctr Logist & Expertsyst GmbH Dept High Performance Comp Salzgitter Germany CALTECH Ctr Adv Comp Res Pasadena CA 91125 USA European Space Agcy Estec NL-2200 AG Noordwijk Netherlands

The Java programming language has a number of features that make it attractive for writing high-quality, portable parallel programs. Aobject formulation, strong typing and the exception model make programs easier to create, debug and maintain. The elegant threading provides a simple route to parallelism on shared-memory machines. Anticipating great improvements in numerical performance, this paper presents a suite of simple programs that indicate how a pure Java Navier-Stokes solver might perform. The suite includes a parallel Euler solver. We present results from a 32-processor Hewlett-Packard machine and a 4-processor Sun server. While speedup is excellent on both machines, indicating a high-quality thread scheduler, the single-processor performance needs much improvement. (C) 2000 Published by Elsevier Science Ltd. Ail lights reserved.

关键词： Java parallel programs Navier-Stokes

来源：评论

学校读者我要写书评

暂无评论

SYNTHETIC-PERTURBATION TECHNIQUES FOR SCREENING SHARED-MEMORY programs

引用

SOFTWARE-PRACTICE & EXPERIENCE 1994年第8期24卷 679-701页

作者： SNELICK, R JAJA, J KACKER, R LYON, G UNIV MARYLAND COLLEGE PK MD 20742 USA NATL INST STAND & TECHNOL DIV STAT ENGN COMP & APPL MATH LAB GAITHERSBURG MD 20899 USA

The synthetic-perturbation screening (SPS) methodology is based on an empirical approach;SPS introduces artificial perturbations into the MIMD program and captures the effects of such perturbations by using the modern branch of statistics called design of experiments. SPS can provide the basis of a powerful tool for screening MIMD programs for performance bottlenecks. This technique is portable across machines and architectures, and scales extremely well on massively parallel processors. The purpose of this paper is to explain the general approach and to extend it to address specific features that are the main source of poor performance on the shared memory programming model. These include performance degradation due to load imbalance and insufficient parallelism, and overhead introduced by synchronizations and by accessing shared data structures. We illustrate the practicality of SPS by demonstrating its use on two very different case studies: a large image understanding benchmark and a parallel quicksort.

关键词： DESIGN OF EXPERIMENTS parallel programs PERFORMANCE SHARED MEMORY PROGRAMMING MODEL SYNTHETIC PERTURBATION

来源：评论

学校读者我要写书评

暂无评论

RANK ORDER FILTERS AND PRIORITY-QUEUES

引用

DISTRIBUTED COMPUTING 1992年第2期6卷 99-105页

作者： KALDEWAIJ, A UDDING, JT WASHINGTON UNIV DEPT COMP SCIST LOUISMO 63130

A derivation of a parallel algorithm for rank order filtering is presented. Both derivation and result differ from earlier designs: the derivations are less complicated and the result allows a number of different implementations. The same derivation is used to design a collection of priority queues. Both filters and priority queues are highly efficient: they have constant response time and small latency.

关键词： PROGRAM DESIGN VLSI SYSTOLIC ARRAYS parallel programs PRIORITY QUEUES

来源：评论

学校读者我要写书评

暂无评论

Formal Sequentialization of Distributed Systems via Program Rewriting

引用

ELECTRONIC NOTES IN THEORETICAL COMPUTER SCIENCE 2007年 188卷 53-75页

作者： Bertran, Miquel Babot, Francesc Climent, August Univ Ramon Llull Informatica Salle Barcelona Spain

Formal sequentialization is introduced as a rewriting process for the reduction of parallelism and internal communication statements of distributed imperative programs. It constructs an equivalence proof in an implicit way, via the application of equivalence laws as rewrite rules, thus generating a chain of equivalent programs. The variety of the possible sequentialization degrees which are attainable is illustrated with an example. The approach is static, thus avoiding the state explosion problem, has an impressive state-vector reduction in many cases, and could be combined, as a model simplification step, with model checking and interactive theorem proving in system verification. Prior grounding results needed in formal sequentialization are overviewed;more specifically, an algorithm for the automatic elimination of communications under the scope of sequential and parallel compositions, elimination laws which the algorithm applies, and a suitable equivalence criterion for the sequentialization process. The main contribution of this work is the extension of these results to encompass the formal elimination of both synchronous communications embedded within a subclass of selection statements, and of non-disjoint synchronous communication pairs. None of these cases has been treated in the literature before, and their solution considerably widens the application domain of formal sequentialization.

关键词： Formal communication elimination formal sequentialization distributed programs parallel programs formal verification laws of distributed programs rewriting

来源：评论

学校读者我要写书评

暂无评论

Static Prediction of parallel Computation Graphs

引用

PROCEEDINGS OF THE ACM ON PROGRAMMING LANGUAGES-PACMPL 2022年第POPL期6卷 1–31页

作者： Muller, Stefan K. IIT Chicago IL 60616 USA

Many algorithms for analyzing parallel programs, for example to detect deadlocks or data races or to calculate the execution cost, are based on a model variously known as a cost graph, computation graph or dependency graph, which captures the parallel structure of threads in a program. In modern parallel programs, computation graphs are highly dynamic and depend greatly on the program inputs and execution details. As such, most analyses that use these graphs are either dynamic analyses or are specialized static analyses that gather a subset of dependency information for a specific purpose. This paper introduces graph types, which compactly represent all of the graphs that could arise from program execution. Graph types are inferred from a parallel program using a graph type system and inference algorithm, which we present drawing on ideas from Hindley-Milner type inference, affine logic and region type systems. We have implemented the inference algorithm over a subset of OCaml, extended with parallelism primitives, and we demonstrate how graph types can be used to accelerate the development of new graph-based static analyses by presenting proof-of-concept analyses for deadlock detection and cost analysis.

关键词： parallel programs graph types cost graphs computation graphs type systems type inference

来源：评论

学校读者我要写书评

暂无评论

parallel SAT Solving in Bounded Model Checking

引用

JOURNAL OF LOGIC AND COMPUTATION 2011年第1期21卷 5-21页

作者： Abraham, Erika Schubert, Tobias Becker, Bernd Fraenzle, Martin Herde, Christian Rhein Westfal TH Aachen Aachen Germany Univ Freiburg Freiburg Germany Carl von Ossietzky Univ Oldenburg D-26111 Oldenburg Germany

Bounded model checking (BMC) is an incremental refutation technique to search for counterexamples of increasing length. The existence of a counterexample of a fixed length is expressed by a first-order logic formula that is checked for satisfiability using a suitable solver. We apply communicating parallel solvers to check satisfiability of the BMC formulae. In contrast to other parallel solving techniques, our method does not parallelize the satisfiability check of a single formula, but the parallel solvers work on formulae for different counterexample lengths. We adapt the method of constraint sharing and replication of Shtrichman, originally developed for sequential BMC, to the parallel setting. Since the learning mechanism is now parallelized, it is not obvious whether there is a benefit from the concepts of Shtrichman in the parallel setting. We demonstrate on a number of benchmarks that adequate communication between the parallel solvers yields the desired results.

关键词： parallel programs bounded model checking SAT solving linear programming hybrid systems

来源：评论

学校读者我要写书评

暂无评论

Analysis and measurement of the effect of kernel locks in SMP systems

引用

CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE 2001年第2期13卷 141-152页

作者： Kaieda, A Nakayama, Y Tanaka, A Horikawa, T Kurasugi, T Kino, I Univ Electrocommun Dept Comp Sci Tokyo 1828585 Japan NEC Corp Ltd Kawasaki Kanagawa 2168555 Japan

This article reports the use of case studies to evaluate the performance degradation caused by the kernel-level lock. We define the lock ratio as a ratio of the execution time for critical sections to the total execution time of a parallel program. The kernel-level lack ratio determines how effective programs work on symmetric multiprocessor (SMP) systems. me have measured the lock ratios and the performance of three types of parallel programs on SMP systems with Linux 2.0: matrix multiplication, parallel make, and WWW server programs. Experimental results show that the higher the lock ratio of parallel programs, the worse their performance becomes. Copyright (C) 2001 John Wiley & Sons, Ltd.

关键词： SMP systems operating systems parallel programs performance evaluation kernel lock

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：