High-throughput LDPC decoding using the RHS algorithm

The relaxed half-stochastic (RHS) algorithm is a recently proposed binary message-passing decoding algorithm for low-density parity check codes that can reach the same error rate performance as belief propagation algorithms that exchange LLR messages. Because of its low-complexity interleaver, the RHS algorithm makes it possible to achieve a fully-parallel implementation that can converge to a codeword in only a few clock cycles on average, enabling high throughput and power efficiency. To demonstrate the practicality of the RHS algorithm, we implement a decoder for the popular IEEE 802.3an 10GBASE-T standard. The paper presents details of the hardware implementation, as well as post-layout results for an ASIC implementation in 65nm CMOS technology, which indicate that the decoder can operate at 448 MHz and occupies an area of 4.41 mm
²
. The results obtained from bit-accurate software simulations show that the decoder meets the latency requirement prescribed by the standard and provides an average throughput of 160 Gbps.