首页> 外文会议>IEEE International Parallel and Distributed Processing Symposium Workshops and PhD Forum >OOGen: An Automated Generation Tool for Custom MPSoC Architectures Based on Object-Oriented Programming Methods
【24h】

OOGen: An Automated Generation Tool for Custom MPSoC Architectures Based on Object-Oriented Programming Methods

机译:OOGen:基于面向对象编程方法的自定义MPSoC架构的自动生成工具

获取原文

摘要

As multiprocessor systems are continuing to be adopted in academic laboratories and industry, researchers and application developers are routinely designing multiprocessor systems from the low level. Before they can validate their ideas, most of their time is spent on handling massive engineering details. It includes building a multiprocessor system from scratch by handling details of IP components and their topologies. A new method to design multiprocessor systems is necessary to improve the productivity and the accessibility. In this work, we present OOGen, an automatic generation tool for multiprocessor architectures by using object-oriented programming (OOP) methods. A system topology can be described in an OOP language (i.e., Java) without indicating massive low level details. OOGen provides libraries and automation framework that compiles design programs to generate corresponding hardware architectures. In order to demonstrate the flexibility of OOGen, we extend the non-uniform memory access (NUMA) template architecture to automatically generate two existing hardware platforms to support programming models as Hthreads and HOpenCL. We present evaluation on resource utilization, performance and energy efficiency by examining 5 typical benchmarks.
机译:随着多处理器系统继续在学术实验室和行业中采用,研究人员和应用程序开发人员通常从低端开始常规设计多处理器系统。在他们验证想法之前,他们的大部分时间都花在处理大量的工程细节上。它包括通过处理IP组件及其拓扑的详细信息从头开始构建多处理器系统。设计多处理器系统的新方法对于提高生产率和可访问性是必要的。在这项工作中,我们介绍OOGen,这是一种通过使用面向对象的编程(OOP)方法为多处理器体系结构自动生成的工具。可以用OOP语言(即Java)描述系统拓扑,而无需指示大量的底层细节。 OOGen提供的库和自动化框架可以编译设计程序以生成相应的硬件体系结构。为了证明OOGen的灵活性,我们扩展了非均匀内存访问(NUMA)模板体系结构,以自动生成两个现有的硬件平台,以支持Hthreads和HOpenCL等编程模型。通过检查5个典型基准,我们对资源利用,性能和能源效率进行了评估。

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