基于GPU的RS码并行译码算法研究

Research on GPU-Based RS-code Parallel Decoding Algorithm

针对RS码硬件译码器中常用的BMA译码算法并行性较差,导致译码延迟较大和译码速率较低的问题,研究了RS码的高速并行译码算法及其软件实现方式,设计和实现了一种基于GPU的RS码并行译码算法cu PGZ。实验结果表明,对CCSDS建议的RS(255,223)码,在128 bits错误的情况下,cu PGZ可以达到590 Mbps的最大译码速率以及0.5 ms的最小译码延迟,与BMA算法的GPU实现相比,译码速率提高1倍,译码延迟降低为1/60。实际工程应用表明,cu PGZ能够满足实际测控数传的信道译码要求。

The low parallel performance of BMA decoding algorithm in RS-code hardware decoder results in high decoding delay and low decoding speed.Aiming at this problem,this paper studies RS-code high-speed parallel decoding algorithm and software imple- ment method,designs and implements a GPU-based RS-eode parallel decoding algorithm cuPGZ.The experimental results show that by leveraging the massive cores inside a GPU card, 590 Mbps in decoding rate and 0.5 ms in decoding delay for RS cnde （255,223） recommended by CCSDS can be obtained with 128 bits error.Compared to GPU-based BMA,cuPGZ increases, one times in decoding rate and decreases 60x in deoding delay.The practical engineering application shows that the euPGZ can meet the actual engineering require- ments in high speed channel decoding.