1 / 9
文档名称:

00632030Image processing PCI-based shared memory architecture design.pdf

格式:pdf   页数:9
下载后只包含 1 个 PDF 格式的文档,没有任何的图纸或源代码,查看文件列表

如果您已付费下载过本站文档,您可以点这里二次下载

00632030Image processing PCI-based shared memory architecture design.pdf

上传人:kuo08091 2012/11/29 文件大小:0 KB

下载得到文件列表

00632030Image processing PCI-based shared memory architecture design.pdf

文档介绍

文档介绍:Image Processing BCI-based Shared Memory Architecture Design
Dominique Houzet, Abdelkrim Fatni, Jean-Luc Basille
IRIT-ENSEEIHT-INP
2, rue Camichel
3 107 1 Toulouse cedex, France
houzet @
Abstract Interface) should be available on a parallel machine. This
is the aim of the presented architecture. The GFLOPS
The object of the GFLOPS project was to study all puter provides an elaborate shell munication
concerning the design of a low cost parallel architecture. The and synchronization logic to support global-memory
project's aim was to develop a parallel architecture as well as access, prefetch, atomic operations, barriers, and block
its software environment to implement image processing transfers [l]. Based on this design, a simple but powerful
applications efficiently. The proposed architecture supports up parallel extension to the C language has been designed
to 512 processor nodes, which are PC mother boards, with the goal of extracting the full performance capability
connected over a scaleable and cost-work, via the out of this machine. The basic approach is to provide a
PCI-bus, at a constant cost per node. The aim is to obtain at
low cost a scaleable virtually shared memory machine with a full C language on every node, with a rich set of
physically distributed architecture. In this paper we present the assignment operatio