Evaluations of Distinct Bitonic Sorters for ORBGRAND
Szu-Hao Huang, Cheng-Hung Dixson Lin · 2024
Guessing Random Additive Noise Decoding (GRAND) is a recently proposed decoding algorithm for forward error correction codes in fifth-generation (5G) consumer communications. Order Reliability Bits GRAND (ORBGRAND) was proposed to utilize received channel values and sort all of them to get more probable error patterns. Therefore, choosing an efficient algorithm to sort the channel values significantly influences the hardware design of the ORBGRAND. In this study, different bitonic sorters are evaluated in terms of frame error rate, synthesized silicon area, and power consumption. To achieve a low-cost ORBGRAND, the half-output bitonic sorter with segmentation not only reduces the number of the compare and swap unit, but also reduces number of the pipeline register. The evaluation results show that the half-output bitonic sorter with two segmentation can achieve 38% area reduction and 39% power reduction, respectively.