Low-Latency CUDA LDPC Decoder for SDR Solutions

Abstract

This paper outlines a method for low-latency CUDA LDPC decoder implementation applicable for software-defined radio (SDR) solutions. The traditional approach to CUDA LDPC decoder implementation allows to get very high decoder throughput but has high decoding latency. It is shown that the proposed implementation of the CUDA LDPC decoder allows to significantly reduce the decoding latency. NVIDIA RTX 2080 was used as a simulation platform. Obtained latency and throughput values satisfy technical requirements for ultra low latency 5G services in NGMN specifications.