首页> 外文OA文献 >Approches d'optimisation et de personnalisation des réseaux sur puce (NoC : Networks on Chip)
【2h】

Approches d'optimisation et de personnalisation des réseaux sur puce (NoC : Networks on Chip)

机译:片上网络的优化和个性化方法(NoC:片上网络)

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

Systems-on-chip (SoC) have become more and more complex due to the development of integrated circuit technology.Recent studies have shown that in order to improve the performance of a specific SoC application domain, the on-chipinter-connects (OCI) architecture must be customized at design-time or at run-time. Related approaches generallyprovide application-specific SoCs tailored to specific applications. The aim of this thesis is to carry out new approachesfor Network-on-Chip (NoC) and study their performances, especially in terms of latency, throughput, energyconsumption and simplicity of implementation.We have proposed an approach to allow designers to customize a candidate OCI architecture by adding strategiclinks in order to match large application workload. The analytical evaluation focuses on improving the physicalparameters of the NoC topology regardless of the application that should run on. The evaluation by simulationfocuses to evaluate the communication performances of the NoC. Simulations results show the effectiveness ofthis approach to improve the NoC performances. We have also introduced a compartmental Fluid-flow basedmodeling approach to allocate required resource for each buffer based on the application traffic pattern. Simulationsare conducted and results show the efficiency of this modeling method for a buffer space optimized allocation.Finally, we proposed a joint approach based on a system dynamics theory for evaluating the performance of a flowcontrol algorithm in NoCs. This algorithm allows NoC elements to dynamically adjust their inflow by using afeedback control-based mechanism. Analytical and simulation results showed the viability of this mechanism forcongestion avoidance in NoCs.
机译:由于集成电路技术的发展,片上系统(SoC)变得越来越复杂。最近的研究表明,为了提高特定SoC应用领域的性能,片上互连(OCI)必须在设计时或运行时自定义体系结构。相关方法通常提供针对特定应用量身定制的特定应用SoC。本文的目的是为片上网络(NoC)实施新方法并研究其性能,特别是在时延,吞吐量,能耗和实现简单性方面。我们提出了一种方法,使设计人员可以定制候选人OCI体系结构通过添加策略链接来匹配大型应用程序工作负载。无论应在何种应用程序上运行,分析评估均着重于改善NoC拓扑的物理参数。仿真评估的重点是评估NoC的通信性能。仿真结果表明,该方法可有效提高NoC性能。我们还介绍了一种基于分区流体流的建模方法,可以根据应用程序流量模式为每个缓冲区分配所需的资源。进行了仿真,结果表明了该建模方法对缓冲区空间优化分配的有效性。最后,我们提出了一种基于系统动力学理论的联合方法,用于评估NoC中流控制算法的性能。该算法允许NoC元素通过使用基于反馈控制的机制来动态调整其流入量。分析和模拟结果表明,该机制在NoC中避免拥塞的可行性。

著录项

  • 作者

    Chariete Abderrahim;

  • 作者单位
  • 年度 2014
  • 总页数
  • 原文格式 PDF
  • 正文语种 fr
  • 中图分类

相似文献

  • 外文文献
  • 中文文献
  • 专利

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