An Efficient Parallel Successive Cancellation List Polar Decoder Based on GPUs

Polar codes are a class of codes that can achieve the symmetric capacity. They are adopted to be control code for the enhanced mobile broadband (eMBB) for the fifth generation(5G) standard. Although Polar codes can be efficiently decoded by successive cancellation algorithm with complexity O(NlogN), decoding performance of this algorithm is not good enough for short codewords. The successive cancellation list(SCL) decoder is recently investigated in most studies. It has better frame error rate(FER) performance but poor latency and throughput. In this study, a parallel SCL decoder based on the graphic processing unit(GPU) is designed to reduce the latency and improve the decoding throughput. An efficient approach for sharing the intermediate values among different decoding paths is introduced. This method reduces the computing complexity and decoding latency. The implementation of parallel non-recursive decoding algorithm also increases the throughput significantly. For the typical case of code length N=1024 and list size L=4 with code rate R = 0.5, the parallel decoder based on GPU achieves throughput of 49 Mbps on Nvidia GTX 980 and 79 Mbps on Nvidia Titan X. The throughputs are 240 and 392 times higher than the decoder based on the CPU.