首页> 外文OA文献 >Exploring performance and power properties of modern multicore chips via simple machine models
【2h】

Exploring performance and power properties of modern multicore chips via simple machine models

机译:探索现代多核芯片的性能和功率特性   简单的机器模型

代理获取
本网站仅为用户提供外文OA文献查询和代理获取服务,本网站没有原文。下单后我们将采用程序或人工为您竭诚获取高质量的原文,但由于OA文献来源多样且变更频繁,仍可能出现获取不到、文献不完整或与标题不符等情况,如果获取不到我们将提供退款服务。请知悉。

摘要

Modern multicore chips show complex behavior with respect to performance andpower. Starting with the Intel Sandy Bridge processor, it has become possibleto directly measure the power dissipation of a CPU chip and correlate this datawith the performance properties of the running code. Going beyond a simplebottleneck analysis, we employ the recently published Execution-Cache-Memory(ECM) model to describe the single- and multi-core performance of streamingkernels. The model refines the well-known roofline model, since it can predictthe scaling and the saturation behavior of bandwidth-limited loop kernels on amulticore chip. The saturation point is especially relevant for considerationsof energy consumption. From power dissipation measurements of benchmarkprograms with vastly different requirements to the hardware, we derive asimple, phenomenological power model for the Sandy Bridge processor. Togetherwith the ECM model, we are able to explain many peculiarities in theperformance and power behavior of multicore processors, and derive guidelinesfor energy-efficient execution of parallel programs. Finally, we show that theECM and power models can be successfully used to describe the scaling and powerbehavior of a lattice-Boltzmann flow solver code.
机译:现代多核芯片在性能和功耗方面表现出复杂的行为。从Intel Sandy Bridge处理器开始,直接测量CPU芯片的功耗并将此数据与正在运行的代码的性能属性相关联成为可能。除了简单的瓶颈分析之外,我们还使用最近发布的Execution-Cache-Memory(ECM)模型来描述流内核的单核和多核性能。该模型完善了众所周知的Roofline模型,因为它可以预测多核芯片上带宽受限的环路内核的缩放比例和饱和行为。饱和点对于能耗的考虑尤其重要。从对硬件有很大不同要求的基准程序的功耗测量中,我们得出了Sandy Bridge处理器的简单现象学功率模型。与ECM模型一起,我们能够解释多核处理器在性能和功率行为方面的许多特殊性,并为并行程序的节能执行提供指导。最后,我们证明了ECM和功率模型可以成功地描述晶格-玻尔兹曼流求解器代码的缩放和功率行为。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
代理获取

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