Convolutional code high-speed parallel decoding method and decoder based on GPU

TLDR: In this article, a convolutional code high-speed parallel decoding method based on a GPU is presented. But the decoding method comprises the steps of firstly segmenting code stream information and sending to the GPU in an asynchronous transmission manner; performing forward path measure computation and path backtracking on segments in parallel according to a viterbi algorithm on the GPU, achieving full parallel decoding, and at last combining the decoding result of each segment and outputting decoding stream information

Abstract: The invention discloses a convolutional code high-speed parallel decoding method and a convolutional code high-speed parallel decoder based on a GPU, and belongs to the field of communication The decoding method comprises the steps of firstly segmenting code stream information and sending to the GPU in an asynchronous transmission manner; performing forward path measure computation and path backtracking on segments in parallel according to a viterbi algorithm on the GPU, achieving full parallel decoding, and at last combining the decoding result of each segment and outputting decoding stream information The decoder of the invention comprises an initialization module positioned on a host and the GPU device, a serial parallel converter positioned at the host, and a forward computation core and a backtracking core positioned at the GPU device Data transmission between the host and the GPU device is achieved by an asynchronous transmission channel According to the method and the decoder, in the case of ensuring that the error resistance is invariable, the degree of parallelism of decoding is enhanced, and the character of the butterfly structure of the convolutional code grid chart is excavated, the computation complexity is reduced, the applicable convolutional code range is wide, the implementation difficulty is low, and the flexibility and compatibility are strong