Research On Iterative Decoding Algorithms Of LDPC Codes On Multi-Core Platform

Low-density parity-check(LDPC)codes are a class of linear block codes with good decoding performance.With their strong error-correcting ability,parallelizable decoding algorithms and Shannon-limit approaching performance,LDPC codes have been widely used in deep space communication,optical fiber communication and mobile wireless communication.In recent years,as the development of communication industry,people have increasingly high requirements on data transmission rate.The efficient decoding of LDPC codes have become an attractive research topic.By using the Min-Sum(MS)algorithm based on flooding scheduling,layered scheduling and shuffled scheduling,LDPC codes are usually decoded in a serial way.In order to improve the decoding speed,the parallel decoding algorithms for the MS decoding algorithm based on the above three scheduling modes are investigated in this thesis.The main research contents are summarized as follows:1.The digital communication system,the classical channel models,the basic principle of LDPC codes and their two iterative decoding algorithms are outlined.Then the architecture and the programming technology for the multi-core CPU are summarized.Moreover,the Open MP and the instruction set of SSE are analyzed in detail.2.Three parallel strategies such as message parallelism,multiple code words parallelism and SSE instruction parallelism for MS decoding algorithm are analyzed.By combining the message parallel strategy with the SSE instruction parallel strategy,a parallel decoding algorithm for MS decoding of LDPC codes under the flood scheduling mode is designed.Compared to the serial MS decoding algorithm,the simulation results show that the proposed parallel decoding algorithm can improve the decoding speed significantly without performance degradation.3.Since the message parallel strategy exhibits a low acceleration rate for LDPC codes with small column and row weights,a second parallel decoding algorithm for the MS decoding algorithm under the flood scheduling mode is designed by combining the strategy of multiple code words parallel and the strategy of SSE instruction parallel.Simulation results show that the speedup ratio of the second parallel decoding algorithm is superior to the parallel decoding algorithm by combining message parallelism and SSE instruction parallelism.4.Compared with the flooding scheduling mode,the serial scheduling mode for the iterative decoding algorithms for LDPC codes shows faster convergence speed.Therefore,the parallel MS decoding algorithms based on layered and shuffled scheduling mode are designed respectively by combining the strategy of multiple code words parallelism and the strategy of SSE instruction parallelism.Simulation results show that the proposed parallel decoding algorithms can improve the decoding speed of LDPC codes without decoding performance loss.