Parallel and low-latency streaming decoding of quantum error correction codes

Real-time decoding is necessary for fault-tolerant quantum computations, which must adapt in response to certain logical measurement results that must be decoded to be reliable. Real-time decoding in hardware was recently demonstrated by integrating Riverlane's decoding chip into the control system of Rigetti's superconducting quantum processor, demonstrating the viability of low-latency quantum error correction (QEC). To perform QEC in real-time, the decoding graph can be broken up into windows, which must be decoded fast enough to keep up with syndrome generation and overcome the backlog problem. We present a new windowing technique that reduces data transfer overheads, increasing the throughput in comparison to other parallel techniques whilst simultaneously reducing the latency. This approach has the potential to support more complex decoding graphs and is easily extensible to different types of decoders, making it a strong candidate for low-latency decoding of a variety of surface codes.