Advantage with slow interconnects versus disconnected devices

Determine whether quantum processing units (QPUs) connected by slow quantum interconnects can perform any computational task significantly better than fully disconnected devices, beyond the specific cases analyzed for distributed CliNR implementations of random Clifford circuits.

Background

The paper studies distributed quantum computing where inter-QPU entanglement generation is much slower than local gate operations. While circuit knitting can mitigate sparse remote operations, its cost grows quickly with the number of remote gates, raising concerns about the usefulness of slow interconnects.

The authors develop a distributed version of the CliNR scheme and present simulations and asymptotic analysis showing advantages under certain conditions. However, the broader question of whether slow-interconnect architectures can outperform disconnected devices in general remains explicitly stated as unclear.