DVB-S2中LDPC和积译码的GPU实现

GPU Implementation of LDPC Sum-product Decoding in DVB-S2

第二代卫星数字广播系统DVB-S2采用接近Shannon限的LDPC码作为内码。在LDPC译码方式中,软判决的和积译码性能最佳,但是由于其采用大量浮点数运算,使得译码器的软硬件实现都较困难。为此,提出一种采用图形处理器(GPU)编程进行译码的实现方式。GPU的并行处理功能使其可以同时满足高精度浮点运算和高速实时解码的要求,为DVB-S2的实际应用提供了新的思路。采用计算机上NVIDIA GeForce 9600显卡编程环境,实现了满足高清视频要求的信息吞吐速率。

Second generation of digital broadcasting system DVB-S2 uses Low Density Parity Check（LDPC） code as its inner error correction code.Among decoding methods of LDPC,sum-product decoding based on soft-decision has the best performance,but it is hard to implement on software or hardware,because it uses many float-point calculations.To solve this problem,a decoding method is promoted using Graphic Processing Units（GPU） programming.The parallel processing ability of GPU enables high-accuracy calculation and high-speed real time decoding,offers a new possibility in DVB-S2 application.NVIDIA GeForce 9600 graphics card is used for programming,which can achieve throughput requirement of DVB-S2.