Message Passing Iterative Decoders based on Imprecise Arithmetic for Multi-Objective Power-Area-Delay Optimization – DIAMOND

Throughput and power consumption are key challenges for next-generation communication systems. High throughput is required to support the continuously increasing demand of traffic volume, while power consumption incorporates sustainability concerns and is a major constraint in mobile devices. Due to the erroneous nature of communication channels, error correcting codes are essential to any communication system, no matter the way the information is transported, for instance using a point-to-point, multicast, or broadcast technology, a wired or wireless link, and a terrestrial, satellite-based or hybrid infrastructure. However, these error-correction mechanisms require intensive computing tasks that turn into a critical bottleneck both in terms of throughput and power consumption. The DIAMOND project proposes to exploit the robustness of modern decoders to arithmetic inaccuracies, for improving their latency and power consumption. The project focuses on graph-based codes that are widely used in modern communication systems, as Low-Density Parity-Check (LDPC) and Turbo codes, and targets the design of message-passing iterative decoders using imprecise arithmetic units. We aim at harnessing the inaccuracies produced by imprecise arithmetic, or more generally imprecise computational units, while benefiting of their significant reductions in area, latency and power consumption. The DIAMOND project will first provide the scientific foundation for both the theoretical analysis and practical design of imprecise iterative decoders. Then, imprecise computational units will be tailored to suit different requirements arising from the multi-objective optimization of the design with respect to energy consumption, latency and area. The proposed solutions will be assessed and validated through proof-of concept models and implementations. In the context of next-generation communication systems, DIAMOND presents a novel approach that tackles the increasingly stringent requirements for high data rates and low power consumption.