Design And Implementation Of Turbo Parallel Decoding Algorithm Based On CUDA

Turbo-codes are an important forward error correcting (FEC) codes because of their Bit Error Rate (BER) are close to the SHANNON limit. Turbo-codes are widely used in3G and4G communication standards such as UMTS,3GPP LTE and LTE-A systems. But Turbo-codes have a serious of problems which cannot be ignored:complex computation procedure and low-speed decoding. These are all bottlenecks for the application and development of Turbo-codes.Graphic Processing Unit was applied in image rendering in the beginning, as general computing software platform CUDA was released, GPU was applied in many general computing fields and exhibits great potential.To resolve the complex decoding algorithm and long-time decoding delay, this paper make a deep inquiry of the MAP decoding algorithm and divides the MAP procedure into three key modules:state-branch-metric module, forward and backward recursive-metric module and likelihood module. Then design corresponding parallel scheme for each module according to the correlations between datas and the hardware platform parameters of GPU.This paper not only design efficient parallel schemes for Turbo decoding but also put them into implementation on the GPU platform with CUDA. Hundreds of thousands of simulation experiments not only verify the accuracy of these methods, but also show that the decoding speed can accelerate by more than4.5times compared to the traditional serial decoding algorithm.