Instruction-Directed MAC for Efficient Classical Communication in Scalable Multi-Chip Quantum Systems

Maurizio Palesi, Enrico Di Russo, Hamaad Rafique, Giuseppe Ascia, Davide Patti, Abhijit Das, Sergi AbadalarXiv 2025 · DOI · arXiv

TL;DR

As quantum computers transition from monolithic chips to modular multi-chip architectures, the classical communication used for control and synchronization is becoming a major bottleneck. This paper introduces ID-MAC, a networking protocol that uses the deterministic nature of quantum circuits to pre-schedule communication, reducing classical latency by up to 70% and significantly extending the "usable" life of qubits.

What Problem Does This Paper Address?

To reach the millions of qubits needed for practical applications, we can't just keep making bigger chips; we need to connect multiple "Quantum Cores" (QCs) together. This modular approach requires two types of connections: quantum links (to move entanglement) and classical links (to move control data).

In a cryogenic environment, wireless networks-on-chip (WiNoC) are a promising way to handle classical data. However, standard networking protocols are "blind." For example, a traditional Circulating Token MAC passes a "permission to speak" token from node to node in a fixed circle. If node A needs to talk to node D, the token might have to pass through nodes B and C first, even if they have nothing to say. In the quantum world, where qubits "die" (decohere) in microseconds, these wasted clock cycles are catastrophic.

Methods

The authors propose Instruction-Directed Token MAC (ID-MAC). The key insight is that, unlike classical software with its unpredictable "if-then" branches, a quantum circuit is a fixed, deterministic sequence of operations known at compile time.

  1. Compile-Time Scheduling: The compiler identifies every instance where a core needs to send classical data (e.g., during a teleportation protocol). It assigns a specific "token order" to these events.
  2. The Dispatch Phase: A global Control Unit sends out instruction bundles. These bundles include "token order" fields that tell each core exactly when it will be allowed to access the wireless channel.
  3. The Execution Phase: Instead of the token circulating blindly, it is passed directly from one active transmitter to the next scheduled transmitter. Idle nodes are skipped entirely.
  4. Hardware Integration: The protocol defines specific packet formats (like the TPS Instruction Packet) that embed this networking logic directly into the quantum control flow.

Datasets

The researchers evaluated ID-MAC using the QCOMM simulation framework. They tested the protocol against two types of workloads:

  • Synthetic Random Circuits: Used to stress-test the system as it scales from 1 to 100 quantum cores.
  • Real-World Benchmarks: Circuits pulled from the Qiskit and MQTBench libraries, including Amplitude Estimation (ae), GHZ state preparation, Quantum Fourier Transform (qft), and Quantum Neural Networks (qnn).

Key Insights and Ideas

  • Determinism as a Feature: The most elegant part of this work is turning a limitation of quantum programming (lack of dynamic branching) into a networking advantage. By "spoiling" the communication schedule ahead of time, the network removes all arbitration overhead.
  • The "Classical Bottleneck" is Real: The paper demonstrates that as quantum gates get faster (thanks to better physics), the relative time spent on classical communication increases. Without networking optimizations like ID-MAC, the classical control plane eventually becomes the primary reason quantum programs fail.
  • Effective Coherence: By finishing the classical "chatter" faster, the system completes the entire quantum program sooner. This effectively "extends" the coherence time of the qubits because they spend less time sitting idle waiting for control signals.

Key Results

  • Communication Speedup: ID-MAC reduces classical communication time by up to 70% compared to traditional token-passing.
  • Total Execution Time: Depending on the system size and the speed of the quantum gates, total program execution time was reduced by 30% to 70%.
  • Scalability: The benefits of ID-MAC grow as the system scales. In a 100-core system, the protocol is essential for maintaining any semblance of efficiency.
  • Benchmark Performance: Even on small 4-core systems running real benchmarks, ID-MAC provided a 4-7% total execution time improvement by nearly eliminating classical delays.

Limitations and Future Work

  • Single Token Constraint: The current version of ID-MAC uses a single token, meaning only one core can talk at a time. The authors suggest that future work could explore multi-token operations to allow parallel classical communication in different parts of the chip.
  • Dynamic Scenarios: While most quantum circuits are deterministic, some future error-correction schemes might involve conditional logic based on measurement results. The paper does not deeply explore how ID-MAC would handle highly non-deterministic, branchy quantum code.
  • Hardware Overhead: While the protocol simplifies the MAC layer, it requires the compiler and the global Control Unit to handle more complex instruction dispatching, the overhead of which would need to be managed in physical hardware implementations.

Summary updated

Abstract

Scalable quantum computing requires modular multi-chip architectures integrating multiple quantum cores interconnected through quantum-coherent and classical links. The classical communication subsystem is critical for coordinating distributed control operations and supporting quantum protocols such as teleportation. In this work, we consider a realization based on a wireless network-on-chip for implementing classical communication within cryogenic environments. Traditional token-based medium access control (MAC) protocols, however, incur latency penalties due to inefficient token circulation among inactive nodes. We propose the instruction-directed token MAC (ID-MAC), a protocol that leverages the deterministic nature of quantum circuit execution to predefine transmission schedules at compile time. By embedding instruction-level information into the MAC layer, ID-MAC restricts token circulation to active transmitters, thereby improving channel utilization and reducing communication latency. Simulations show that ID-MAC reduces classical communication time by up to 70% and total execution time by up to 30-70%, while also extending effective system coherence. These results highlight ID-MAC as a scalable and efficient MAC solution for future multi-chip quantum architectures.
BibTeX
@misc{palesi2025instruction,
  author = {Maurizio Palesi and Enrico Di Russo and Hamaad Rafique and Giuseppe Ascia and Davide Patti and Abhijit Das and Sergi Abadal},
  title = {{Instruction-Directed MAC for Efficient Classical Communication in Scalable Multi-Chip Quantum Systems}},
  year = {2025},
  eprint = {2510.27273},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2510.27273},
  howpublished = {arXiv},
}

Loading related content...

Figure