检索结果-南通市图书馆

A new parallel lattice reduction algorithm for BKZ reduced bases

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2014年第9期57卷 136-145页

作者： LIU xianghui FANG Xing WANG Zheng xie xianghui Department of Applied Mathematics Zhengzhou Information Science and Technology Institute

In order to implement the original BKZ algorithm in parallel,we describe it in terms of parallelism and give its parallel implementation *** we analyze the efficiency of algorithm’s parallel implementation and show that the speedup factor of BKZ algorithm in parallel is extremely *** we present a new parallel lattice reduction algorithm suitable for multiprocessor computer *** new algorithm can obtain a BKZ reduced basis and the parallel speedup is *** with the practical results,although the computational complexity increases compared with the original BKZ algorithm,we still indicate that the new algorithm performs well in parallel and the time cost in parallel is *** the same time,we show that the length of the shortest vector is smaller.

关键词： lattice reduction BKZ algorithm Goldstein-type lattices parallel algorithm multiprocessor computer architecture

Deep Learning Blockchain Integration Framework for Ureteropelvic Junction Obstruction Diagnosis Using Ultrasound Images

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Tsinghua Science and Technology 2024年第1期29卷 1-12页

作者： Yu Guan Pengceng Wen Jianqiang Li Jinli Zhang xianghui xie Faculty of Information Technology Beijing University of TechnologyBeijing 100124China Beijing Children’s Hospital Affiliated to Capital Medical University Beijing 100020China

UreteroPelvic Junction Obstruction(UPJO)is a common hydronephrosis disease in children that can result in an even progressive loss of renal *** is an economical,radiationless,noninvasive,and high noise preliminary diagnostic step for *** intelligence has been widely applied to medical fields and can greatly assist doctors'diagnostic *** demand for a highly secure network environment in transferring electronic medical data online,therefore,has led to the development of blockchain *** this study,we built and tested a framework that integrates a deep learning diagnosis model with blockchain *** diagnosis model is a combination of an attention-based pyramid semantic segmentation network and a discrete wavelet transformation-processed residual classification *** also compared the performance between benchmark models and our *** diagnosis model outperformed benchmarks on the segmentation task and classification task with MloU=87.93,MPA=93.52,and accuracy=91.77%.For the blockchain system,we applied the InterPlanetary File System protocol to build a secure and private sharing *** framework can automatically grade the severity of UPJO using ultrasound images,guarantee secure medical data sharing,assist in doctors'diagnostic ability,relieve patients'burden,and provide technical support for future federated learning and linkage of the Internet of Medical Things(loMT).

关键词： data mining image processing and computer vision machine learning medical information systems

On-Chip Generating FPGA Test Configuration Bitstreams to Reduce Manufacturing Test Time

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Chinese Journal of Electronics 2016年第1期25卷 64-70页

作者： WANG Fei WANG Da YANG Haigang xie xianghui FAN Dongrui System on Programmable Chip Research Department IE CAS State Key Laboratory of Computer Architecture ICT CAS

Statistics shows that over 95% of FPGA manufacturing test time is spent on loading test configuration bitstreams. Reducing the test time that spent on loading test configuration bitstreams could significantly reduce FPGA test time. A new approach which can significantly reduce the FPGA test time is presented. Experimental results show that the proposed technique can at least reduce the configuration loading time by 96%, while getting 100%test coverage with less than 1.2% hardware overhead.

关键词： FPGA test Test configuration bitstream Design-for-testability

Analysis of diagnosis and treatment of lipoblastomatosis

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Life Sciences) 2017年第7期60卷 778-780页

作者： Zhiqiang Mo xianghui xie Huanmin Wang Hong Qin Wei Han Xiaosong Li Department of Pediatric Surgery Beijing Children's Hospital Capital Medical University National Center for Children's Health Beijing 100045 China

Dear Editor,Lipoblastomatosis is a rare benign soft tissue tumor arising from the white fetal-embryonal adipose tissue（Choi et al.,2013）.Since the distinction of lipoblastomatosis from lipoblastma was reported by Ve... 详细信息

关键词：母细胞治疗诊断脂肪

Limited Multiple-Writer: An Approach to Dealing withFalse Sharing in Software DSMs

在线全文

学校读者我要写书评

暂无评论

Journal of Computer Science & Technology 2000年第5期15卷 453-460页

作者：谢向辉韩承德 Institute of Computing Technology Chinese Academy of Sciences Beijing P.R. China

False sharing is one of the most important factors impacting the performance of DSM (distributed shared memory) systems. The single-writer approach is simple, but it cannot avoid the ping-pong effect of the data page thrashing, while the multiple-writer approach is effective for false sharing but with high cost. This paper proposes a new approach, called limited multiple-writer (LMW) to handling multiple writers in software DSM. It distinguishes two kinds of multiple-writer as lock-based form and barrier-based form, and handles them with different policies. It discards the Twin and Diffin traditional multiple-writer approach, and simplifies the implementation of multiple-writer in software DSM systems. The implementation of LMW in a CVM (Coherent Virtual Machine) software DSM system, which is based on a network of workstations, is introduced. Evaluation results show that for some applications such as SOR (Successive Over-Relaxation), LU (Lower triangular and Upper triangular), FFT (Fast Fourier Transformation), and IS (Integer Sorting), LMW provides a significant reduction in'execution time (11%, 16%, 33% and 46%) compared with the traditional multiple-writer approach on the platform.

关键词： memory consistency false sharing single-writer multiple-writer multiple protocols migration and replication

在线全文

学校读者我要写书评

暂无评论

众核处理器的流水线紧耦合指令循环缓存设计

计算机研究与发展 2017年第4期54卷 813-820页

作者：张昆过锋郑方谢向辉数学工程与先进计算国家重点实验室江苏无锡214125

能效比是未来高性能计算机需要解决的重要问题.众核处理器作为高性能计算机的重要实现手段,其微结构的优化设计对能效比提升尤为关键.提出了1种面向众核处理器的流水线紧耦合的指令循环缓存设计,以较小的L0指令缓存提供更加高能效的指... 详细信息

能效比是未来高性能计算机需要解决的重要问题.众核处理器作为高性能计算机的重要实现手段,其微结构的优化设计对能效比提升尤为关键.提出了1种面向众核处理器的流水线紧耦合的指令循环缓存设计,以较小的L0指令缓存提供更加高能效的指令取指.作为体系结构研究同硬件可实现性紧密结合的1次尝试,设计始终考虑了硬件实现代价这一关键约束.为了控制L0指令缓存对流水线性能的影响,指令缓存采用了循环出口预取技术,以此保证指令缓存提供的低功耗的指令取指能够最终转化为流水线能效比的提升.在gem5模拟器上实现了对指令循环缓存的模拟.对SPEC2006的测试结果表明,在不影响流水线性能的前提下,设计的典型配置可以减少27%的指令取指功耗以及31.5%的流水线前段部件动态功耗.

关键词：循环缓存众核处理器能效比旨令缓存结构优化

在线全文

学校读者我要写书评

暂无评论

BFS算法与众核处理器的适应性研究

计算机研究与发展 2015年第5期52卷 1187-1197页

作者：叶楠郝子宇郑方谢向辉数学工程与先进计算国家重点实验室江苏无锡214215

以图计算为代表的数据密集型应用获得越来越广泛的关注,而传统的高性能计算机处理这类应用的效率较低.面向未来高性能计算机体系结构要有效支持数据密集型计算,深入研究以广度优先搜索(breadth-first search,BFS)算法为代表的图计算的... 详细信息

以图计算为代表的数据密集型应用获得越来越广泛的关注,而传统的高性能计算机处理这类应用的效率较低.面向未来高性能计算机体系结构要有效支持数据密集型计算,深入研究以广度优先搜索(breadth-first search,BFS)算法为代表的图计算的典型特征,设计实现轻量级启发式切换BFS算法,该算法通过基本搜索方式的自动切换,避免冗余内存访问,提高搜索效率;针对BFS算法的离散随机数据访问特征以及众核处理器执行机制,建立面向BFS算法的众核处理器体系结构分析模型;全面、深入研究了BFS算法在典型众核处理器上的运行特征和性能变化趋势.测试结果表明:Cache命中率、内存带宽、流水线利用效率等相关参数均处于较低水平,无法完全满足BFS算法的需求,因此需要能够支持大量离散随机访问和简单执行机制的新型众核处理器体系结构.

关键词：广度优先搜索算法众核处理器体系结构分析模型协同研究

在线全文

学校读者我要写书评

暂无评论

双精度浮点矩阵乘协处理器研究

计算机研究与发展 2019年第2期56卷 410-420页

作者：贾迅邬贵明谢向辉吴东数学工程与先进计算国家重点实验室江苏无锡214125

矩阵乘运算在多个应用领域特别是数值计算领域被广泛使用,但双精度浮点矩阵乘在CPU,GPGPU,FPGA等现有计算平台上的性能和效率受限,其往往成为大规模数值计算应用的性能瓶颈.针对该问题,以线性阵列计算结构为基础,研究了双精度浮点矩阵... 详细信息

矩阵乘运算在多个应用领域特别是数值计算领域被广泛使用,但双精度浮点矩阵乘在CPU,GPGPU,FPGA等现有计算平台上的性能和效率受限,其往往成为大规模数值计算应用的性能瓶颈.针对该问题,以线性阵列计算结构为基础,研究了双精度浮点矩阵乘的定制加速.首先,对线性阵列计算结构进行了双缓冲优化并设计了针对双缓冲的存储访问调度,以提高结构的计算效率.其次,提出了矩阵乘协处理器和加速计算系统的结构,构建了协处理器的性能模型并对其结构设计空间进行了探索.最后,验证了协处理器的功能正确性并在某主流工艺下评估了其硬件开销.实验结果表明,设计的双精度浮点矩阵乘协处理器可以达到3 TFLOPS的计算性能和99%的计算效率.与NVIDIA K40 GPGPU相比,协处理器执行双精度浮点矩阵乘的性能是K40的1.95倍,而面积开销仅为K40的21.05%.探索了定制加速结构设计在高性能计算中的应用,对现有计算系统的性能提升具有一定的参考价值.

关键词：矩阵乘协处理器加速浮点硬件定制

在线全文

学校读者我要写书评

暂无评论

蚁群:一种新型的高效多用计算平台

计算机研究与发展 2015年第6期52卷 1341-1350页

作者：谢向辉钱磊吴东原昊李祥数学工程与先进计算国家重点实验室江苏无锡214125

在科学计算和大数据处理应用需求的推动下,高性能计算机的性能不断提升、系统规模日益扩大,系统功耗越来越成为制约能力提升的重要瓶颈.在深入分析现有4类高性能计算机的基础上探讨了2项关键技术:1)可重构微服务器(reconfigurable micro... 详细信息

在科学计算和大数据处理应用需求的推动下,高性能计算机的性能不断提升、系统规模日益扩大,系统功耗越来越成为制约能力提升的重要瓶颈.在深入分析现有4类高性能计算机的基础上探讨了2项关键技术:1)可重构微服务器(reconfigurable micro server,RMS)技术.解决单个计算节点在领域应用加速能力、系统功耗和体积间的平衡兼顾问题.2)自治与分治相结合的集群构造技术.解决基于微小型化计算节点的大规模计算平台构造与扩展性问题.在此基础上,提出了一种新型的高效多用计算平台架构——"蚁群",构建了包含2 048个低功耗、微小型化RMS计算节点的蚁群平台原型系统,并实现了大规模指纹实时比对和多RMS节点协同排序2个典型应用.测试表明,单个RMS节点的指纹比对性能是Xeon单核的34倍,功耗仅5W,整个原型系统可实现千万量级指纹库的数百并发实时查询;蚁群平台的数据排序性能功耗比是GPU平台的10倍以上,有效提升数据排序的效率.

关键词：高性能计算集群可重构计算微服务器计算平台