Architecture and System-Level Solutions for Real-Time Decoding in Fault-Tolerant Quantum Computers

The error rates of quantum devices are orders of magnitude higher than what is needed to run most practical quantum applications. Quantum error correction (QEC) enables us to close this gap by encoding logical qubits using several physical qubits. By periodically identifying errors in real-time using a decoder, QEC prevents errors from accumulating and achieves a logical error rate lower than the error rate of the physical qubits. Unfortunately, software decoding algorithms are often too slow to enable real-time decoding (within a latency of a few micro-seconds). In this talk, I will discuss the role of micro-architecture and system-level optimizations to improve the performance and scalability of decoders. First, I will describe LILLIPUT, a lightweight reconfigurable practical lookup table decoder for small QEC codes that programs Look-Up Tables on FPGAs with the error assignments offline and decodes online during actual experiments. Second, I will discuss the AFS decoder that focuses on accelerating the Union-Find decoding algorithm using specialized hardware and leverages contention-aware resource sharing to design the system level architecture and organization of decoders in fault-tolerant quantum computers with multiple logical qubits for enhanced scalability.