基于CUDA平台的规则LDPC码的译码实现研究  

ON DECODING IMPLEMENTATION OF REGULAR LDPC CODE BASED ON CUDA

作　　者：俞永盛[1] 陆佩忠[1] 

机构地区：[1]复旦大学计算机科学与工程系,上海200433

出　　处：《计算机应用与软件》2010年第4期230-232,266,共4页Computer Applications and Software

摘　　要：低密度奇偶校验(LDPC)码性能优越,允许全并行高速译码,已经在个人数字设备、移动无线通信等领域显示出了很大的应用价值,极可能取代Turbo码成为第四代移动通信的首选编码方案。NVIDIA公司的CUDA是一种新的用于GPU通用计算的软硬件架构。基于CUDA平台程序员可以写出C风格的代码来启动大量的GPU线程并行工作。基于CUDA平台提出和研究一种AWGN信道下的规则LDPC码的译码实现方案。仿真实验对LDPC码译码的CPU实现和CUDA实现的性能作了详细比较。研究表明CUDA能够带来明显的性能提升。Low Density Parity Check（LDPC） codes have good performance and their high-speed decoding can be implemented in full parallel,so they have shown large potential in many application fields such as personal digital devices,mobile wireless communication and so on,they are quite possible to supersede the Turbo codes as the preferred coding scheme for mobile communication of the 4th generation.NVIDIA＇s CUDA is a new hardware and software architecture for general-purpose computation on GPU.On CUDA the programmers can write C-like codes to run a lot of GPU thread parallel works simultaneously.In this study we investigated decoding implementation of regular LDPC codes on CUDA based on AWGN channel.The performances of decoding the LDPC code on CPU and on CUDA are compared in detail based on simulation results.It is shown that CUDA can bring drastic acceleration in performance.

关 键 词：CUDA平台 计算密集型任务 并行计算 LDPC码 迭代译码 

分 类 号：TP334.7[自动化与计算机技术—计算机系统结构]