• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>IEEE International Conference in MOOCs, Innovation and Technology in Education >RMI approach to cluster based Winograd's variant of Strassen's method
【24h】

RMI approach to cluster based Winograd's variant of Strassen's method

机译：基于Class的Winograd的斯特索方法变种的RMI方法

获取原文
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

Present day's amount of computational requirements has shifted the processing of data from the regular way to parallel way of computation. Pipelined processors, array processors can be employed to construct design of parallel hardware. These systems constructed can be further extended with the help of scalar and super scalar systems. We provide an efficient way of implementing Winograd's variant of Strassen's matrix multiplication on parallel systems by making use of RMI(Remote Method Invocation ) which provides us distributed object oriented programming, multithreading programming. Multithreading approach helps a lot in concurrent, dynamic and asynchronous programming. In this paper clusters have been made from very basic hardware systems which are connected on local area network using switch and a master system connected to this switch uses three NIC. The facility of Winograd's variant of Strassen's method to run in divide and conquer approach is the basic fact around which whole analysis has been performed. Actual partitioning (AP) of a given matrix into sub matrix according to cluster size and at each system logical partitioning (LP) of each sub matrix into size 2×2 has been performed which on a whole covers the aspect of SIMD based model. The parallel computing metrics like speed up, efficiency, excessive parallel overhead proves RMI as a better approach for distributed computing as RMI provides look and feel like a user is working on local machine as not much overheads are being there in order to convert a non-distributed application to the distributed one.
机译：的计算需求现今的量将偏离常规的方式来计算的并行方式进行数据的处理。流水线处理器，阵列处理器可被用来构建并行硬件的设计。这些构造的系统可以与标量和超标量系统的帮助下，进一步延长。我们提供通过使用RMI（远程方法调用）的执行并行系统施特拉森的矩阵乘法威诺格拉德的变型的一种有效的方式为我们提供面向对象的分布式编程，多线程编程。多线程的方式帮助了很多并发，动态和异步编程。在本文中的簇已经被从被连接使用开关和连接到该开关的主系统局域网上非常基本的硬件系统由使用三个NIC。施特拉森的方法威诺格拉德的变型的设施在运行分而治之的办法是各地已经进行了整体分析的基本事实。根据簇尺寸和在每个子矩阵的每个系统逻辑分区（LP）成大小为2×2的给定矩阵的实际分区（AP）为子矩阵已经执行其上的整体盖基于SIMD模式的方面。并行计算般的速度了，效率指标，过多的并行开销证明RMI作为一个更好的方法用于分布式计算为RMI提供的外观和感觉像一个用户正在使用本地机器上没有太大的开销在那里，以转化非分布式应用分布式之一。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号