When designing a quantum algorithm, we typically operate in an idealized, abstract world. We assume a collection of logical qubits where any qubit can instantly undergo an entangling operation with any other qubit.
In real quantum hardware, this all-to-all connectivity is an illusion. Physical quantum processing units (QPUs) are highly constrained networks. Qubits are arranged in rigid physical geometries—such as 2D grids, heavy-hex patterns, or modular multi-core clusters—and can typically only interact with their immediate physical neighbors.
To bridge this gap, compilers must map logical qubits to physical locations. If an algorithm dictates a gate between two distant qubits, the compiler must insert a sequence of SWAP gates to physically route the quantum information across the chip until the qubits are adjacent. In the current Noisy Intermediate-Scale Quantum (NISQ) era, these routing steps act as a massive "tax" on performance: every additional SWAP gate increases circuit depth, introduces operational noise, and risks decoherence.
To solve this problem, researchers have shifted away from generic compilation toward topology-aware qubit mapping and hardware-algorithm co-design. Instead of treating the hardware layout as an afterthought, this paradigm structures the algorithm and its physical placement simultaneously to minimize data movement and maximize fidelity.
The Cost of Topological Mismatch
Why does physical layout matter so much? It is not merely a matter of counting gates; the specific geometry of a chip dictates how noise and errors propagate through a computation.
Crickmore et al. demonstrate in Effect of component variations on the gate fidelity in linear optical networks that even when two circuit designs perform the exact same mathematical operation, their physical arrangement determines their resilience to manufacturing imperfections. By choosing specific topologies, sensitivity to hardware errors can be concentrated into fewer components, making the overall system more robust.
On superconducting chips, Nannapaneni shows in Experimental evaluation of quantum Bayesian networks on IBM QX hardware that different hardware topologies (such as bow-tie versus linear layouts) inherently yield different baseline error rates. Mapping logical qubits to physical locations based on the device's specific connectivity graph is therefore a primary factor in maintaining state fidelity.
Initial Layout Optimization
The first line of defense against the SWAP gate tax is choosing the optimal starting positions for logical qubits on the physical grid.
Zhou et al. show in Quantum Circuit Transformation Based on Simulated Annealing and Heuristic Search that optimizing this initial layout can reduce total communication overhead by over 30% compared to naive placement. Rasconi et al. confirm in An Innovative Genetic Algorithm for the Quantum Circuit Compilation Problem that providing a compiler with the flexibility to optimize the initial layout significantly improves final execution times, even though it expands the compilation search space.
For structured mathematical operations, this spatial optimization can be baked directly into the algorithm's design. For example, Saeedi et al. demonstrate in Constant-Factor Optimization of Quantum Adders on 2D Quantum Architectures that strategically placing auxiliary qubits within a 2D grid facilitates faster carry propagation and aligns output positions with the input requirements of subsequent blocks, eliminating downstream communication overhead.
Similarly, when solving real-world optimization problems, aligning the logical structure of the problem graph (such as a 2D image grid) directly with the physical topology of a quantum annealer minimizes the need for qubit cloning, as shown by Venkatesh et al. in Q-Seg: Quantum Annealing-based Unsupervised Image Segmentation.
Hardware-Algorithm Co-Design
While optimizing the layout of a fixed circuit is beneficial, a more powerful approach is to design the circuit for the target topology from the very beginning. As Rieffel et al. argue in Assessing and advancing the potential of quantum computing: A NASA case study, simultaneously optimizing algorithms alongside physical connectivity and noise characteristics is essential for practical NISQ applications.
This co-design philosophy manifests in several ways:
- SWAP-Free Circuits: Li et al. demonstrate in A Novel Spatial-Temporal Variational Quantum Circuit to Enable Deep Learning on NISQ Devices that structuring variational circuits so that entangling operations occur only between adjacent physical qubits completely eliminates the need for SWAP gates. This hardware-aware approach preserves state fidelity and drastically improves classification accuracy on real-world noisy processors.
- Hardware-Efficient Ansatze: In quantum chemistry and materials science, Shee et al. (Quantum simulation of preferred tautomeric state prediction) and Choudhary et al. (JARVIS-Leaderboard: a large scale benchmark of materials design methods) show that tailoring the gate configuration to the physical constraints of the processor is just as influential to calculation accuracy as the choice of the underlying algorithm itself.
- Topology-Aligned Parametric Circuits: For generative models, Balauca et al. demonstrate in Practical Hybrid Quantum Language Models with Observable Readout on Real Hardware that aligning parametric quantum circuits with specific hardware layouts, such as IBM's heavy-hex topology, minimizes gate overhead and enables successful end-to-end training on physical hardware.
[Abstract Circuit] ──┐
├──► [Co-Search Engine] ──► [Optimized Hardware Execution]
[Hardware Topology] ─┘
Rather than relying on manual design, researchers use automated search frameworks to find these optimal configurations. Wang et al. introduce this concept via QuantumNAS in QuantumNAS: Noise-Adaptive Search for Robust Quantum Circuits, demonstrating that a simultaneous "co-search" of both the circuit architecture and its physical qubit mapping yields vastly superior results compared to optimizing them independently.
Similarly, Hsieh et al. show in Differentiable quantum architecture search that custom-designed shallow circuit topologies optimized for specific connectivity achieve higher success probabilities than deeper, rigid algorithmic templates.
Scaling to Modular and Multi-Core Topologies
As quantum computers scale up, hardware manufacturers are moving away from single monolithic chips toward modular, multi-core architectures. This introduces a hierarchical routing problem: compilers must manage fast intra-core communication alongside slower inter-core interconnects.
To address this, Ben Rached et al. (Characterizing the Inter-Core Qubit Traffic in Large-Scale Quantum Modular Architectures) and Rodrigo et al. (Characterizing the spatio-temporal qubit traffic of a quantum intranet aiming at modular quantum computer architectures) introduce traffic-aware qubit mapping. By analyzing the communication intensity and interaction patterns of an algorithm over time, they map high-traffic virtual qubits to physical locations that minimize long-distance inter-core transfers, mitigating communication bottlenecks.
Core A (High Local Traffic) Core B (High Local Traffic)
┌───────────────┐ ┌───────────────┐
│ Q1 ─── Q2 │ │ Q4 ─── Q5 │
│ │ │ │ │ │ │ │
│ Q3 ───[P1]───┼─── Inter-Core ───┼──[P2]─── Q6 │
└───────────────┘ Interconnect └───────────────┘
(Teleportation Port)
To streamline this distributed execution, Liu et al. propose a compilation method in ECDQC: Efficient Compilation for Distributed Quantum Computing with Linear Layout that restructures complex multi-core topologies into a Linear Nearest Neighbor layout. They strategically designate "dangling qubits" at the boundaries of these linear chains as dedicated communication ports for quantum teleportation, minimizing the internal swap gates required to move data to an inter-chip link.
Finally, to accelerate the compilation process itself at scale, Escofet et al. introduce Route-Forcing in Route-Forcing: Scalable Quantum Circuit Mapping for Scalable Quantum Computing Architectures. This routing mechanism treats qubits as physical entities governed by attraction forces derived from the circuit's gate queue, dynamically pulling interacting qubits toward each other across the hardware topology to achieve massive speedups in compilation time and reductions in circuit depth.
Key Foundational Work
Constant-Factor Optimization of Quantum Adders on 2D Quantum Architectures (2013) — contributed 2 concept elements to this direction.
Effect of component variations on the gate fidelity in linear optical networks (2016) — contributed 1 concept elements to this direction.