Research On Parallel Decoding Methods Of LDPC Codes On GPU Platform

With the development of communication theory, the low density parity check code is a kind of linear block codes which is close to the Shannon limit. With its low decoding complexity and good error correction performance, it has become a hot research topic in the field of encoding, and has been widely used in digital communication system.For the large scale LDPC codes,the decoding time of the Sum Product, Min Sum and Normalized Min Sum is too long,which is not conducive to practical application.In order to shorten the decoding time,the multicore technology of the graphics processor platform is uesd and the above decoding algorithms are researched by the parallel method.The main works of this thesis are shown as follows:1. The decoding algorithms of LDPC codes are summarized and several common channel models are intorduced, Then the origin and development of channel coding are expounded. The hardware/software system and the communication mode are described.The storage structure and optimization technology of CUDA C are emphasizely analyzed.2. The iteration method of SP decoding algorithm is analyzed and the parallel decoding scheme is designed, then the simulation results and analysis of SP decoding algorithm are given. Through the simulation results of multiple code words, it can be concluded that the decoding performance with the multicore implementation is similar to that with the conventional CPU implementation, but the former has a faster running speed.3. The decoding algorithms of MS and NMS are analyzed, and parallel implementation schemes of the two algorithms in GPU are given according to the characteristics of the two algorithms. From the simulation results,the decoding time of MS and NMS in the GPU is shortened,which compares to the results of pure CPU implementation.