arXiv:2610.36965
From QCCD Trapped-Ion Hardware to Quantum Compilers: Capabilities, Architectures, and Compilation
Ludwig Schmid, Jurek Eisinger, Daniel Schoenberger, Christian D. Marciniak, Janine Hilder, Ulrich Poschinger, Robert Wille, Ferdinand Schmidt-Kaler
Abstract
Shuttling-based trapped-ion quantum computers (QCCD) rank among the most mature hardware platforms. Existing and envisaged devices exhibit architectural differences in terms of register layout and connectivity. Platform-specific compilers are required to translate quantum circuits into resource-optimized hardware-level command chains. Hardware platforms and compiler software stacks are mostly developed independently of one another by different research communities. This work aims to bridge these communities by abstracting QCCD architectures and native operations, enabling computer scientists to address compilation problems using generic methods without requiring a deep background in the physical operating principles of the platform. The resulting design space is organized into a small set of architectural archetypes, each posing its own compilation problem. We then select and apply compilation methods to one- and two-dimensional QCCD architectures, finding that the routing overhead scales approximately cubically with the number of qubits for the one-dimensional architecture and approximately quadratically for the two-dimensional one. Extending this to compilation with a quantum error-correction layer reveals additional differences and similarities between the physical and logical levels. We conclude that co-design is a requirement rather than an option, and this work provides the vocabulary, archetypes, and compilation abstraction it needs.
AI Summary
The paper addresses how to express shuttling-based trapped-ion (QCCD) hardware capabilities as compiler constraints, and how architectural choices affect routing overhead. It organizes native gates, measurement, cooling, shuttling, timing, and error considerations into compiler-facing abstractions; classifies architectures from single-well chains to modular devices; and describes physical and fault-tolerant compilation tasks. Two case studies compare a newly developed linear-trap compiler, using gate aggregation followed by greedy/tabu-search movement optimization, with an existing graph-based 2D shuttling compiler. Random circuits contain $L\lfloor N/2\rfloor$ two-qubit gates, with $L\propto\log N$. Over the register sizes studied, routing counts grow approximately cubically in 1D and quadratically in 2D. These are finite-size numerical observations, not asymptotic bounds. The metrics differ: physical SWAP counts for linear traps and junction moves for grids. A Steane-code logical teleportation circuit uses 41 physical qubits and 100 two-qubit gates. Increasing linear processing-zone capacity from two unaddressed ions to ten individually addressed ions reduces SWAP counts roughly fourfold with one zone and nearly eightfold with two. Adding a second two-ion zone instead increases SWAP counts by about one third. On the grid, additional processing zones reduce routing counts with diminishing returns. The comparisons do not isolate architectural effects from compiler performance: each architecture uses one compiler, and some trends may reflect heuristic or partitioning choices. Linear-trap costs exclude splitting, merging, and transport; the differing routing metrics do not provide a direct comparison of execution time, fidelity, or overall architectural performance.
Error found: The approximate cubic and quadratic routing-count scalings are supported only by finite-size visual trends, without fitted exponents or uncertainty estimates, so their robustness beyond the evaluated configurations is not established.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 17 pages, 8 figures, 3 tabels