| 摘要: |
| 智能化赋予了物联网更深刻的实用价值，但是在计算能力强与功耗低的之间寻求性能最优是目前物联网设备极难解决的问题.异构多处理器结构与单一或者同构的多处理器相比可以结合不同处理器的优势，同时满足高计算能力与低功耗的系统需求，但异构多处理器结构下软件编程难度大的问题以及如何优化顶层应用在多处理器设备上的运行性能都是目前亟待解决的技术难题.针对以上问题，设计并实现了一个面向异构多处理器设备的自适应命令解释系统.首先，该系统允许用户将物联网应用安装到设备上，应用程序以命令脚本形式呈现；其次，系统设计了命令在异构多处理器设备上的自动分发算法，该算法考虑性能和功耗的多维参数，在满足时间上限的条件下最优化应用执行能耗.最后，提出了针对同时满足不同用户应用需求的解决方案，在物联网设备的资源受限的条件下，根据具体用户使用习惯，提出了一种基于用户使用历史的命令解释系统自适应方案，可以根据用户个性化习惯自动完成命令解释系统的自适应部署和运行时优化. |
| 关键词: 异构多处理器设备 命令解释器 用户习惯 智能化 物联网 |
| DOI： |
| 分类号: |
| 基金项目:国家自然科学基金（61672498，61502461） |
|
| Self-Adaptive Command Interpretation System for Heterogeneous Multiprocessor Devices |
|
LIU Wen-Qing, LI Dong, CUI Li
|
|
Institute of Computing Technology Chinese Academy of Sciences, Beijing 100190, China
|
| Abstract: |
| Intelligence has given the Internet more practical value, but the dual requirements for both the computing power and low power consumption in the current single-processor networking equipment have not been met. The heterogeneous multiprocessor architecture can combine the advantages of different processors with a single or isomorphic multiprocessor to select the right processor to meet the requirements for both high-performance and low-power intelligent device. At present, the software programming under heterogeneous multiprocessor structure and how to optimize the performance of top-level applications on multiprocessor devices are the technical problems that need to be solved. In view of the problems above, this paper designs and implements an adaptive command interpretation system for heterogeneous multiprocessor devices. First, the system allows users to install the Internet of Things application to the device in the form of URL access, and the application is presented in the form of a command script contained in the URL. Secondly, the system designs an automatic distribution algorithm for commands on heterogeneous multiprocessor devices, which adds the energy consumption parameters on the basis of the DAG graph model, and designs a command distribution scheme with the best energy consumption under the condition of satisfying the time limit. Finally, the system is faced with the problem of satisfying the needs of different users at the same time, but the resources of the equipment in the Internet of Things are limited. So the system needs to be customized according to the specific users. This paper presents a self-adaptive scheme of command interpretation system based on users' habits, which can automatically complete the system adapting based on users' specific habits. |
| Key words: heterogeneous multiprocessor devices command interpreter user habits intelligence internet of things |
