• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>International Scientific-Technical Conference "Dynamic of Technical Systems" >Application of the Methodology of Creating Parallel-Pipeline Programs for Reconfigurable Computer Systems on the Example of Implementation of Surface-Related Multiple Prediction Problem in Real Time
【24h】

Application of the Methodology of Creating Parallel-Pipeline Programs for Reconfigurable Computer Systems on the Example of Implementation of Surface-Related Multiple Prediction Problem in Real Time

机译：在实时与曲面相关多预测问题的实现示例中创建并行管道程序的应用方法

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

摘要

Application of the existing methodology of creating parallel-pipeline programs for reconfigurable computer systems, based on field programmable gate arrays, allows reaching the highest possible computer system performance, when using an existing hardware resource. When solving real-time tasks, this approach often leads to a significant decrease in the specific performance of reconfigurable systems. The decrease is associated with redundant hardware costs. We have suggested a new methodology of creating parallel-pipeline programs for solution of real-time computationally intensive tasks. The methodology allows you to synthesize a balanced task computing structure, which requires the minimum hardware resource for the specified rate. The main idea of the new methodology is to bring the performance of the computing structure to the required level by implementation an operation reduction of performance. Unlike the existing methods, the reduction coefficient does not depend on the available hardware resource of the system, but on the specified time to solve the problem. In this case, the order of execution of various reduction methods depends on the critical resource of the task. The application of the methodology is illustrated by the example of solving the problem of surface-related multiple prediction. This example belongs to the class of tightly coupled computationally laborious tasks. Currently, the solution to this problem is obtained within a few days on cluster computing systems; therefore, it is reasonable to use reconfigurable computing systems to predict multiple waves in real time. The developed methodology allowed us to create a parallel-pipeline program solving the problem of surface-related multiple prediction in an hour and a half. The order of reduction transformations was chosen, in which the pipeline computational structure of the problem was obtained. The article presents a comparison of the hardware resource obtained by applying the developed methodol
机译：在使用现有硬件资源时，基于现场可编程门阵列的创建可重新配置计算机系统的并行管道程序的现有方法的应用允许达到最高可能的计算机系统性能。在解决实时任务时，这种方法通常导致可重新配置系统的特定性能的显着降低。减少与冗余硬件成本相关联。我们已经建议创建并行管道程序的新方法，以解决实时计算密集型任务。该方法允许您综合平衡任务计算结构，该结构需要最小的硬件资源以获得指定的速率。新方法的主要思想是通过实现降低性能的操作来将计算结构的性能带到所需的水平。与现有方法不同，减少系数不依赖于系统的可用硬件资源，但在指定的时间内解决问题。在这种情况下，各种减少方法的执行顺序取决于任务的关键资源。通过解决与表面相关的多次预测问题的示例来说明方法的应用。此示例属于紧密耦合的计算费力的任务类别。目前，在群集计算系统的几天内获得该问题的解决方案;因此，使用可重新配置的计算系统是合理的，以实时预测多个波。开发的方法使我们能够在一个小时半的时间和一半的表面相关的多个预测问题创建并行管道程序。选择了减少变换的顺序，其中获得了问题的流水线计算结构。该物品介绍了通过应用发达的方法获得的硬件资源的比较

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