Research On Turbo Parallel Decoding Simulations Based On Cuda

Because of the excellent decoding performance, Turbo code is widely used in modern communication. However, with the complicated decoding algorithm, the decoding delay of Turbo code is large on the CPU simulation. To address this issue, different scenarios will be discussed.This paper firstly introduces the basic theoretical knowledge of Turbo code. Then, a brief introduction of Turbo coding structure and a detailed description of Turbo coding of the3rd generation partnership project long term evolution (3GPP LTE) standard are given. Finally, Turbo decoding schemes are simulated and studied decoding performance on different parameters and different decoding algorithms. For the serial decoding delay, the most efficient sub-block parallel decoding is described. That means each frame is divided into sub-frames of the same length, i.e. sub-blocks, and with different decoders for decoding respectively. So, sub-blocks can be decoded simultaneously which is greatly alleviated the problem of decoding delay. However, the bit error rate will be loss. Because of divided into sub-blocks, the boundary value cannot be determined. This article has introduced three programs which also adopted the reasonable boundary recursive initialization schemes, And decoding simulations are given respectively.For decoding simulation on the CPU is not really parallel decoding, we introduce the Compute Unified Device Architecture (CUDA). This paper firstly introduces the basic knowledge of CUDA architecture, including the organization and management of multi-threading, memory architecture, programming model. Then a simple parallel decoding theme is given, including a detailed description of the kernel functions, the parallel mode of recursive variables, memory capacity, and finally analysis of simulation performance. For the disadvantages of simple parallel decoding scheme, simulation analysis of Turbo sub-blocks parallel decoding based on CUDA is discussed. In terms of memory, it has been greatly improved. And a comparative analysis is given with Turbo decoding implemented on the CPU. The feasibility and acceleration for Turbo decoding are improved by the performance on GPU.