arXiv:2609.20573
FT-Weave: Real-Time Compilation Framework for Reconfigurable Fault-Tolerant Quantum Architectures
Wan-Hsuan Lin, Milan Kornjača, Chen Zhao, Sheng-Tao Wang, Jason Cong
Abstract
Fault-tolerant quantum computing (FTQC) is essential for large-scale quantum computation, but realizing useful application throughput requires coordinating resource preparation, assignment, routing, and logical execution under strict hardware and timing constraints. Many FTQC compilation approaches construct offline schedules using nominal or fixed magic-state factory throughput. Such schedules cannot respond to stochastic resource-preparation and teleportation outcomes, leading to execution stalls and hardware underutilization. In this work, we introduce FT-Weave, a stage-aware real-time FTQC compilation framework that jointly coordinates resource preparation, resource assignment, teleportation routing, and correction handling. By adapting to runtime resource availability and hardware constraints, FT-Weave allows preparation, communication, and logical execution to overlap. We instantiate FT-Weave on two representative neutral-atom, early FTQC architectures: transversal STAR and a T-state cultivation architecture. Under the evaluated hardware and latency model, FT-Weave achieves a speedup of up to 3X over a baseline compilation flow for simulations of the two-dimensional transverse-field Ising model. In the case study, we further find that maximizing exposed concurrency does not necessarily minimize execution time. Although fine-grained asynchronous execution can reduce local idle time, its smaller optimization windows and increased routing contention can outweigh these gains. Together, these results show that effective runtime coordination, rather then exposed parallelism alone, determines how efficiently FTQC resources translate into application throughput:FT-Weave provides a blueprint for solving this real-time orchestration problem across resource protocols and architectures.
AI Summary
FT-Weave addresses how to coordinate stochastic resource-state preparation, assignment, routing, and teleportation corrections during fault-tolerant execution, rather than scheduling against fixed factory throughput. It separates offline circuit scheduling, layout design, and predictable routing from runtime decisions driven by preparation and measurement outcomes. The framework is instantiated for neutral-atom transversal STAR and $T$-state cultivation architectures. Its policies include speculative preparation of STAR correction angles, minimum-cost matching for resource assignment and cultivation-stage redistribution, AOD-compatible movement batching, and resource rematerialization when preparing a replacement is estimated to cost less than delivery. It also implements synchronous and event-driven asynchronous execution. Evaluation uses simulated single-step, second-order Trotter circuits for the periodic two-dimensional transverse-field Ising model, with 16–100 logical qubits, distance-$9$ surface codes, typically one factory per logical qubit, and one to five AODs. Results averaged over ten trials show up to $3\times$ speedup over a synchronous baseline with separated data and factory regions. The largest improvements accompany interleaving factories and data; several STAR runtime optimizations are enabled simultaneously, limiting attribution. Coordinated synchronous STAR execution is approximately 10% faster than asynchronous execution with one AOD and 1% faster with five, attributed to larger assignment and movement-batching windows. Cultivation-stage redistribution provides a further 8% improvement. The stated limitations include model-dependent operation and movement latencies, a homogeneous workload, and an architecture without explicit specialized zones. Broader architectures, heterogeneous applications, and dynamically switching execution policies remain future work; the reported speedup is not presented as universal.
Central claim may not hold: The evaluated baseline already responds to stochastic outcomes, so the experiments do not establish an advantage over the nonadaptive offline scheduling used to motivate FT-Weave. The approximately 3× speedup includes changing the factory/data layout and enabling multiple optimizations together, preventing attribution of that gain specifically to runtime orchestration. The evaluation does not quantify or incorporate the complete classical feedback-loop latency, leaving the real-time feasibility and achievable end-to-end speedup unestablished.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 28 pages, 10 figures