LAUSR

In terms of overall latency, one of the main bottlenecks of successive cancellation list (SCL) decoding of polar codes is the path pruning unit (PPU), whose goal is to prune $2L$ paths to $L$ paths. To remedy this shortcoming, we propose a two-step metric sorter adapted to a parallel implementation of SCL decoding. We describe our sorter architecture and we show that our proposed solution outperforms the state-of-the-art PPUs in terms of area and latency reduction.