Study Of Parrallel ADMM-PD Decoding Algorithms

Low-Density Parity Check(LDPC)codes have high reliability in channel coding,and its performance of the decoding can be well achieved by the linear programming(LP)decoding algorithm.However,when the signal-noise ratio(SNR)is low,the decoding performance is poor.The parallel decoding(PD)algorithm based on the Alternating Direction Method of Multipliers(ADMM)can effectively compensate for the lack of decoding performance of the LP algorithm,but its decoding speed is low.Therefore,it is the main research work of this paper to consider designing and implementing the parallelization scheme of PD algorithms to improve the decoding speed.Not only does the Graphic Processing Unit(GPU)have huge advantages in terms of floating-point operations and memory bandwidth,the NVIDIA Corporation's Compute Unified Device Architecture(CUDA)model also greatly promotes the highly parallelized GPU's applications.Therefore,considering the parallel decoding process of the ADMM-PD algorithm combined with the GPU platform can not only guarantee a very high decoding rate,but also apply to regular LDPC codes and irregular LDPC codes.The main work of this paper is summarized as follows:1.The algorithm for iterative update of variable nodes and check nodes with graphics processor(GPU)are designed,and the ADMM-PD algorithms based on GPU platform are implemented.This method can be applied to both regular LDPC codes and irregular LDPC codes.2.The parallelized decoding structure of the algorithm under the GPU platform is divided in detail,and the algorithms are compiled according to the CUDA model.The experimental results show its decoding performance and decoding speedup-ratio on the GPU platform.