High Level Synthesis Design for Soft-Input GRAND

Abstract

Guessing Random Additive Noise Decoding (GRAND) is a universal channel code decoder for short linear block codes. GRAND can be implemented efficiently on hardware and features both hard-input and soft-input variants. Hardware implementations based on GRAND range from those that employ sequential error pattern generation to reduce resource consumption at the expense of higher decoding latency to those that employ parallel error pattern generation to reduce decoding latency at the expense of higher resource utilization. In this work, we present a soft-input Ordered Reliability Bits GRAND (ORBGRAND) hardware implementation framework based on High-Level Synthesis (HLS). The proposed HLS-based methodology simplifies Design Space Exploration (DSE) for ORBGRAND hardware implementations targeted at applications with varying constraints on decoding latency and resource utilization. To the best of our knowledge, this is the first-of-its-kind HLS-based soft-input ORBGRAND hardware implementation framework, and the source code for the proposed framework is provided under a permissive open-source license.