A Multi-Core Processor Platform for Open Embedded Systems
Hiroaki Inoue
备注:
A dissertation presented to the graduate school of science and technology of Keio University in candidacy for the degree of Doctor of Philosophy in Engineering, September 2009
Recent proliferation of embedded systems has generated a bold new paradigm, known as open embedded systems. While traditional embedded systems provide only closed base applications (natively-installed software) to users, open embedded systems allow the users to freely execute open applications (additionally-installed software) in order to meet various user requirements, such as user personalization and device coordination. Key to the success of platforms required for open embedded systems is the achievement of both the scalable extension of base applications and the secure execution of open applications. Most existing platforms, however, have focused on either scalable or secure execution, limiting their applicability. This dissertation presents a new secure platform using multi-core processors, which achieves both scalability and security. Four techniques feature the new platform: (1) seamless communication, by which legacy applications designed for a single processor make it possible to be executed on multiple processors without any software modifications; (2) secure processor partitioning with hardware support, by which Operating Systems (OSs) required for base and open applications are securely executed on separate processors; (3) asymmetric virtualization, by which many OSs over the number of processors are securely executed under secure processor partitioning; and (4) secure dynamic partitioning, by which the number of processors allocated to individual OSs makes it possible to be dynamically changed under secure processor partitioning.
Generation of the Single Precision BLAS library for the Parallella platform, with Epiphany co-processor acceleration, using the BLIS framework
The Apriella是一个混合计算平台,是Adapteva的Kickstarter项目的结果。它由高性能、高能效、多核架构、Epiphany芯片(用作协处理器)和一个Zynq-7000系列芯片组成,该芯片通常运行常规Linux操作系统版本,作为主处理器,并在其内部FPGA中实现“胶水逻辑”,以与P2P中的许多接口进行通信。在这篇论文中,我们创建了一个Epiphany加速的BLAS库,它适用于Epiphany平台(也适用于类似的混合平台,包括Epiphany芯片作为协处理器)。在实际实施双语法例及会计系统时,我们采用了双语法例及会计系统的架构。在这个平台上,以前已经有过矩阵-矩阵乘法的实现,在Epiphany芯片内实现了非常好的性能(高达峰值的85%),但对于完整的Sparkella平台来说并不是那么好(由于芯片间数据传输带宽限制)。这项工作的主要目的是为了更接近实际的线性代数应用程序的整个Atelella平台,与科学计算的观点。
Power, Energy and Speed of Embedded and Server Multi-Cores applied to Distributed Simulation of Spiking Neural Networks: ARM in NVIDIA Tegra vs Intel Xeon quad-cores