Parallell implementering av en QC-LDPC-kodare på en NVIDIA-GPU : En utvärdering av LDPC-koder i enlighet med 5G-standarden

With the exponential growth of data transmission in 5G networks, efficient channel coding has become essential [1]. QC-LDPC codes provide strong error correction and a hardware-friendly structure, but their high computational demands make real-time encoding challenging on standard CPUs.This thesis investigates the use of Graphics Processing Units (GPUs) and CUDA-based parallel programming to accelerate QC-LDPC encoding. The core contributionlies in demonstrating how GPU parallelism can dramatically reduce encoding latency and scale more efficiently than CPU-based solutions. Experimental results show that GPU acceleration achieved speedups exceeding 80× for large blocksizes—reducing encoding times from over 14 seconds on a CPU to less than 200 milliseconds on a GPU. Although performance improvements were limited for smaller messages, the GPU demonstrated a clear and growing advantage as message sizes grew linearly.The findings highlight the strengths of GPUs in accelerating QC-LDPC encodingand point toward more efficient, scalable, and energy-aware solutions for 5Gsystems, meeting the growing demands of next-generation wireless networks.