基于Shared Memory的多核算法处理系统及实现  

Multi-core algorithm processing system based on Shared Memory and its realization

作　　者：吕涛涛[1] 邓正宏[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《现代电子技术》2013年第6期10-14,共5页Modern Electronics Technique

基　　金：西北工业大学研究生创业种子基金(Z2012109)

摘　　要：随着电子产业、工商业以及军事化产业的迅猛发展,越来越多的复杂运算已经无法单纯利用算法上的优化处理来大幅提升执行速度。为了解决日益复杂的计算问题,利用多核处理势必成为一种大势所趋。所以着手从多核和单核对比出发,利用Xilinx的XUP Virtex-2 Pro建立多核处理平台(基于Shared Memory通信机制)以及单核平台,并将相同的图像处理算法DCT分别运行在所构造的基于FPGA的单核和多核硬件平台上,观察实验结果,比较多核和单核运行所需的时间以及资源的消耗,最终的结果有力的说明多核在图像处理方面的绝对优势。Along with the rapid development of electronic industry, industry and commerce, and military industry, more and more complex operation has been unable to use simple optimization processing of algorithm design to improve execution speed. The multi-core processing will inevitably become a development trend to solve the increasingly complex calculation diffi- culties. Setting about from contrast between multi-core and mono-coreprocessing effects, the multi-core （based on the Shared Memory communication mechanism） and mono-core processing platforms are established with XUP Virtex-2 Pro of Xilinx, and the same image processing algorithm DCT is respectively run on the established multi-core and mono-core platforms based on FP- GA. The observation of the experimental results, and the comparison of the operation time and resource consumption required by the two algorithms indicate that the multi-core processing platform has absolute predominance in image processing.

关 键 词：多核处理平台 SHARED MEMORY 离散余弦变换 FPGA 图像处理 

分 类 号：TN919-34[电子电信—通信与信息系统] TP274.2[电子电信—信息与通信工程]