Toward Performance Visualization for TAU Instrumented Exascale Scientific Workflows

Toward Performance Visualization for TAU Instrumented Exascale Scientific Workflows
复制标题

实现 TAU 仪表化百亿亿次科学工作流程的性能可视化

DOI:
--
复制
发表时间:
2017
期刊:
影响因子:
--
通讯作者:
W. Xu
W. Xu
中科院分区:
--
文献类型:
--
作者:
W. Xu

文献摘要

被引文献

相似文献

在亿级科学计算中,有效地监测、评估和改进性能是至关重要的。可视化,尤其是可视化分析,在亿级计算时代是有用和必要的技术,以实现和增强这种以人为本的体验。在这项工作中,我们设计了一个可视化平台的性能评估的科学工作对exascale的情况。我们首先改进了Tau仪器工具集以适应工作流测量。并设计了新的可视化方法来显示工作流的轨迹和过程信息。为了支持可扩展性,提出了几个层次的细节可视化方法与用户交互。最后,通过一个NWChem的应用实例对所提出的方法进行了验证.
In exascale scientic computing, it is essential to eciently monitor, evaluate and improve performance. Visualization and especially visual analytics are useful and necessary techniques in the exascale computing era to enable and enhance such a human-centered experience. In this work, we devise a visualization platform for performance evaluation of scientic workows toward the exascale scenario. We rst improve Tau instrumentation toolset to accommodate workow measurements. en we design new visualization methods to show trace and prole information of the workow. In order to support the scalability, a few level-of-detail visual methods are proposed with user interactions. Finally, an NWChem use case is adopted to verify our methods.