arXiv:2610.26099 · International Conference on Quantum Communications, Networking, and Computing (QCNC)
Distributing Quantum Circuits Using Pre-Distributed Entanglement Pairs over Quantum Networks
Yukun Yang, Ranjani G. Sundaram, Himanshu Gupta
Abstract
Distributing quantum circuits across a network of quantum devices is a recent approach to improve the scalability of quantum computation. Entangled pairs of qubits distributed across network nodes are used as a communication resource for executing inter-node (remote) gates. However, establishing these entanglement pairs incurs significant latency, and doing so during the circuit's execution may lead to the decoherence of logical qubits. Instead, it may be beneficial to distribute certain entanglement pairs before the circuit is executed. Thus, we consider the problem of distributing a quantum circuit across a quantum network containing some pre-distributed entanglement pairs and seek to minimize the execution time of the circuit. We address this problem in three steps. First, given the quantum network and (optionally) a probability distribution over training circuits, we determine a set of pre-distributed entanglement pairs under a specified budget. Second, we compute a qubit-to-memory allocation that minimizes the expected execution time given these pre-distributed resources. Finally, we identify any additional entanglement pairs required during execution, generate them concurrently, and execute the circuit gates. We evaluate our methods on a broad range of benchmark circuits.
AI Summary
The paper studies how to minimize distributed quantum-circuit execution latency when a network can prepare a limited stock of entangled pairs (EPs) before execution. These pairs are consumed once, with additional pairs generated on demand. The approach jointly considers EP provisioning and a static assignment of circuit qubits to network memories; remote CNOTs use telegates, while local routing swaps are reversed to restore the assignment. The optimization uses a weighted sum of pairwise gate demand remaining after subtracting pre-distributed EP supply as a latency surrogate. It proposes three provisioning heuristics—topology-based allocation, workload-aware greedy allocation, and projected subgradient updates—and three placement methods based on greedy search, Hungarian assignment refinement, and simulated annealing. Execution scheduling reuses a prior batching procedure to accommodate decoherence. A further, unevaluated extension adds charges for consuming pre-distributed EPs. NetSquid simulations cover QPE, QFT, Deutsch–Jozsa, and GHZ circuits on randomized networks with 10 nodes and 10 memories per node, using a 1-second decoherence threshold. Each size bin supplies training circuits and 100 test samples from that same bin. Workload-aware provisioning outperforms topology-only provisioning; greedy placement performs best among the tested placement methods. The preferred provisioning method varies by circuit family, and some small instances have essentially no remaining online EP-generation latency. No numerical speedup factors are reported in the text. The scope excludes dynamic qubit mappings, teleportation-based routing, and continuous EP replenishment during execution. The optimization methods are heuristic; the pricing extension likewise provides no constrained-optimality guarantee.
Central claim may not hold: The optimized objective sums independent entanglement-generation latencies, which generally differs from elapsed execution time under concurrent generation and resource constraints. The reported near-zero execution times establish negligible additional entanglement-generation cost under the approximation, not negligible total circuit execution time. Refreshing pre-distributed entanglement only before execution does not ensure that pairs remain valid until consumption, so the claimed decoherence guarantee is unsupported. The projected-subgradient justification incorrectly treats a minimum over placement-dependent convex functions as admitting the asserted convex subgradient, and its placement oracle may not even return an exact minimizer. The local-routing cost charges 2d swaps for memories at distance d, although bringing the operands together and restoring their placement requires only 2(d-1) swaps along that path.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Journal reference
- Proceedings - 2026 International Conference on Quantum Communications, Networking, and Computing (QCNC), pp. 252-258 (2026)
- DOI
- 10.1109/QCNC69040.2026.00043
- Comments
- 9 pages, 4 figures. Full author version revised after acceptance; includes the cost-aware entanglement consumption section omitted from the QCNC 2026 version to meet the conference page limit