NAQsim: Full-Stack Architecture Simulation Framework for Fast and Space-Efficient Neutral Atom Quantum Computing

arXiv:2609.14478 · quant-ph, cs.AR · Submitted 2026-09-13 · Read on arXiv

Listen

Radio episode about this paper

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: We're looking at the paper: "NAQsim: Full-Stack Architecture Simulation Framework for Fast and Space-Efficient Neutral Atom Quantum Computing".

Paper summary: Kai: What’s interesting is that this framework is tailored specifically for neutral-atombased fault tolerant quantum computing architectures that rely on transversal gates. This is crucial for exploring fast and space efficient designs because those are the real goals in this field.

Mira: The paper claims NAQsim enables detailed full stack architecture evaluation, which opens up doors to explore previously overlooked performance bottlenecks. It’s about having a holistic tool for evaluating architectural ideas where performance, space overhead, and logical error rate all interact.

Lev: I think the real problem they’re tackling is that each physical operation on neutral atoms is still orders of magnitude slower than it is on superconducting qubits, so architects need a comprehensive modeling tool for this platform.

Kai: And the paper addresses this by proposing NAQsim, which they present as a novel modeling tool specifically for neutral atom fault tolerant quantum computing architectures. It’s the first full stack simulation framework they describe for this type of system.

Mira: They are highlighting that existing modeling tools often focus on things like lattice surgery, but those don't capture the unique capabilities of neutral atoms, like atom shuttling or the constraints of control hardware such as AOD route conflicts.

Lev: So what makes it different is that NAQsim handles those unique constraints and allows for more complex FTQC protocols that introduce much greater modeling complexity than what’s been done before.

Conclusion: Kai: So, looking at NAQsim, the authors are really pushing for a comprehensive modeling tool because they feel the community is lacking something that evaluates architectures holistically. They want people to have a reliable way to evaluate ideas considering all three factors we mentioned: performance, space overhead, and logical error rate.

Mira: I think the main implication here is that this framework gives architects a way to test their designs before they commit significant time and resources to building physical hardware on neutral atom platforms. It moves things from just theoretical models to something more grounded in actual system constraints.

Lev: And if NAQsim works as they claim, it means we can start looking at how to design protocols that are genuinely fast and space efficient for neutral atoms, rather than just trying to make slow physical operations run faster without changing the architecture fundamentally.

Kai: The paper’s focus on identifying and mitigating performance bottlenecks in a very specific way—like the patch rotation issue—shows they’re not just building a simulator; they are actively driving architectural optimization right through their tool.

Mira: It sounds like NAQsim is important because it provides the necessary environment for architects to find those specific trade-offs that are currently hidden, especially when dealing with things like logical qubit overlap in the qubit plane.

Yosuke Ueno, Shinichi Sunami, Toshihide Hinokuma, Yasunari Suzuki, Akihisa Goban, Hayata Yamasaki, Teruo Tanimoto, Ilkwon Byun

RIKEN Center for Quantum Computing · University of Oxford

quant-ph, cs.AR

Submitted: 2026-09-13

Updated: 2026-11-18

Comments: 18 pages, 20 figures, 4 tables. Accepted to the 59th IEEE/ACM International Symposium on Microarchitecture (MICRO 2026). v2: Fixed minor issues in Fig. 7

Code: https://github.com/naqsim/naqsim

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 90/100

The gist: NAQsim proposes an open-source simulation framework for neutral-atom fault-tolerant quantum computing architectures based on transversal gates…

Key concepts

NAQsim
A full-stack simulation framework designed specifically for neutral-atom fault-tolerant quantum computing. It simulates the entire process from compiling instructions to estimating final execution time, qubit overhead, and footprint.
Patch Rotation Bottleneck
A performance issue in neutral-atom systems where rotating logical qubits (patches) consumes a large portion of the total execution time, sometimes up to 77.6%. This slow rotation limits the overall speed of quantum operations.
D3-ROT Architecture
An optimized architecture derived through co-optimization that addresses the rotation bottleneck. It reduces average exposed rotation latency significantly, achieving a 2.27x speedup over the baseline by implementing direct rotations and dedicated rotation patches.

Terminology

Summary

NAQsim proposes an open-source simulation framework for neutral-atom fault-tolerant quantum computing architectures based on transversal gates, which is crucial for exploring fast and space-efficient designs The key feature of NAQsim enables detailed full-stack architecture evaluation that opens opportunities to explore previously overlooked performance bottlenecks As a first use case for NAQsim, the authors identify one such bottleneck, patch rotations, perform full-stack co-optimization, and finally derive a near-rotation-free architecture (D3-ROT) D3-ROT achieves a 2. p. 027× speedup from this single bottleneck alone, with minimal footprint overhead p. 0. p. 0

Modeling the Neutral Atom System

The paper details the neutral-atom quantum computer system components, including atom trapping and transport using optical tweezers generated by a spatial light modulator (SLM) Neutral-atom systems offer the ability to physically move atoms at runtime using acousto-optic deflectors (AODs) Single-qubit gates are performed by driving, for example, Raman transitions between the two qubit states (0⟩ and 1⟩) using focused laser beams Two-qubit entangling gates utilize the Rydberg blockade Readout uses state-dependent fluorescence, where an imaging laser excites one qubit state, and the emitted photons are collected by a camera to distinguish 0⟩ from 1⟩.

Developing the Simulation Framework

NAQsim is described as a full-stack simulation framework tailored for neutral-atombased FTQC, evaluating performance, space overhead, and success probability in three stages: compilation, execution, and estimation The compilation stage generates instruction schedules for the given logical instruction set The execution stage iteratively simulates the scheduled instructions on the target qubit plane, accounting for the control hardware configuration The estimation stage finally outputs execution time, qubit overhead and footprint, and logical error rate.

Identifying Performance Bottlenecks

The authors perform full-stack co-optimization to resolve the previously overlooked performance bottleneck, which is identified as logical-qubit (or patch) rotation Their performance analysis across various practical FTQC workloads shows that patch rotation accounts for 55.3%–77.6% of the total execution time. p. 1 To resolve this bottleneck, three architectural solutions are proposed: (1) direct rotation with spatial light modulator (SLM) laser toggling in the control hardware, (2) dedicated rotation patches in the qubit plane, and (3) distribution of rotations in the compiler. p. 1

Architectural Optimization Strategies

The paper explores optimization opportunities at the compilation stage, including defining separate logical instructions for transversal H (TR H) and patch rotation (ROT) This separation exposes two novel compiler optimization opportunities: unnecessary patch rotations and flexible rotation scheduling. The authors propose two extreme rotation scheduling policies: aggregation, which maximizes the number of rotations within one layer, and distribution, which minimizes the number of rotations within one layer. p. 9

Proposing the D3-ROT Architecture

The D3-ROT architecture is proposed through full-stack co-optimization with (1) direct rotation, (2) dedicated rotation patches enabled by a heterogeneous qubit plane, and (3) distribution of rotations. This approach reduces the average exposed rotation latency from 520 µs to 49. p. 28 µs D3-ROT successfully realizes a near-rotation-free architecture by reducing the average exposed rotation latency from 520 µs (Baseline) to 49. p. 18 µs.

Comparative Performance Analysis

The evaluation compares D3-ROT against intermediate implementations and lattice surgery architectures, showing that D3-ROT achieves an average speedup of 2.27× and up to 2.57× (Hei) The baseline NA architecture requires 58× the execution time of the superconducting LS architecture D3-ROT reduces the execution time to 25. p. 85× that of the superconducting LS architecture. p. 3

Resource Overhead Analysis

The paper analyzes resource-state (e.g., T⟩, Y⟩) distillation overhead for QPE workloads, showing that resource-state factories increase the space-time costs of D3-ROT and the LS architectures by factors of 1.81 and 1.14, respectively The authors attribute the higher space-time overhead of D3-ROT mainly to two factors: MSC is more costly in transversal architectures because its throughput is low relative to the rate at which logical operations are executed, and MSD in D3-ROT requires 15 distilled Y⟩ states.

Conclusion and Future Work

The work makes the following contributions: (1) Neutral-atom FTQC modeling framework, which is the first full-stack simulation framework for modeling and evaluating neutral-atom FTQC architectures (2) Novel full-stack co-optimization, providing cooptimization of the compiler, control hardware, and qubit plane for fast and space-efficient neutral-atom FTQC (3) Open-source tool release, releasing NAQsim to the community to help explore future NA FTQC architectures. The work is the first study to develop a modeling tool for FTQC on a neutral-atom platform and conduct detailed architecture-level analysis and optimization to address the performance bottleneck. p. 3 p. 3

Related Work Context

The paper contrasts its work with prior resource analysis studies, noting that previous analyses assume equal atomshuttling times for transversal H and S gates and transversal entangling gates, which can lead to inaccurate resource estimates The authors emphasize that in the Baseline, H and S gates account for as much as 73% of the total latency. Their work focuses on the transversal surface code as the first target for architecture-level modeling and optimization The methodology can be extended to qLDPC codes through appropriate modifications to corresponding modules. p. 1 The primary architectural contribution is identifying and mitigating the patch rotation bottleneck in NA-based FTQC systems, which is orthogonal to existing compiler optimizations. p. 9

Experimental Setup and Evaluation

The evaluation methodology uses eight practical FTQC workloads spanning four representative algorithm classes with various logical-qubit scales and program sizes The control hardware setup assumes state-of-the-art selective laser operations and their latency values for shuttling-free ESM The authors assume a physical error rate of 0.1% for CZ gates, measurement, reset, and AOD–SLM transfers, and 0. p. 8 p. 801% for single-qubit gates. p. 1 The end-to-end simulation takes approximately 30 hours on the tested AMD EPYC 9684X server. p. 6

Artifact Availability

The artifact evaluation package is available at https://zenodo.org/records/21542759, and the end-to-end simulation takes approximately 30 hours on a server with an AMD EPYC 9684X processor The fast path takes less than one minute on a laptop-class CPU. The code licenses are MIT and data licenses are MIT. p. 2

Summary of Results

The results show that D3-ROT achieves an average speedup of 2.27× and up to 2.57× (Hei), with a modest 22. p. 14% footprint overhead on average D3-ROT successfully realizes a near-rotation-free architecture by reducing the average exposed rotation latency from 520 µs (Baseline) to 49. p. 18 µs. The performance gap compared to an ideal, zero-overhead direct rotation (DIR IDEAL) is a mere 13. p. 98% (842.3 µs vs. 740.1 µs overall). p. 1 D3-ROT limits the footprint overhead to just 22. p. 04% while maintaining a low LER. p. 3 The performance gap compared to lattice surgery architectures narrows this gap to 3. p. 09× by applying architectural solutions in a space-time-efficient manner. p. 1 The work is the first study to develop a modeling tool for FTQC on a neutral-atom platform and conduct detailed architecture-level analysis and optimization to address the performance bottleneck. The single most informative sentence about the paper's subject and finding is D3-ROT achieves a 2.27× speedup from this single bottleneck alone, with minimal footprint overhead

--- Page 1 ---

D3-ROT achieves a 2.27× speedup from this single bottleneck alone, with minimal footprint overhead

Pruning redundant reasoning tokens

The paper proposes NAQsim, an open-source simulation framework for neutral-atom FTQC architectures based on transversal gates As its key feature, NAQsim enables detailed full-stack architecture evaluation that opens opportunities to explore previously overlooked performance bottlenecks

How the training signal is built

The compilation stage generates instruction schedules for the given logical instruction set The execution stage iteratively simulates the scheduled instructions on the target qubit plane, accounting for the control hardware configuration The estimation stage finally outputs execution time, qubit overhead and footprint, and logical error rate

Where the gains come from

The authors perform full-stack co-optimization to resolve the previously overlooked performance bottleneck, which is identified as logical-qubit (or patch) rotation Their performance analysis across various practical FTQC workloads shows that patch rotation accounts for 55.3%–77.6% of the total execution time

The paper explores optimization opportunities at the compilation stage, including defining separate logical instructions for transversal H (TR H) and patch rotation (ROT) The authors propose two extreme rotation scheduling policies: aggregation, which maximizes the number of rotations within one layer, and distribution, which minimizes the number of rotations within one layer.

The D3-ROT architecture is proposed through full-stack co-optimization with (1) direct rotation, (2) dedicated rotation patches enabled by a heterogeneous qubit plane, and (3) distribution of rotations This approach reduces the average exposed rotation latency from 520 µs to 49.8 µs.

The work makes the following contributions: (1) Neutral-atom FTQC modeling framework, which is the first full-stack simulation framework for modeling and evaluating neutral-atom FTQC architectures (2) Novel full-stack co-optimization, providing cooptimization of the compiler, control hardware, and qubit plane for fast and space-efficient neutral-atom FTQC (3) Open-source tool release, releasing NAQsim to the community to help explore future NA FTQC architectures. The work is the first study to develop a modeling tool for FTQC on a neutral-atom platform and conduct detailed architecture-level analysis and optimization to address the performance bottleneck

The paper contrasts its work with prior resource analysis studies, noting that previous analyses assume equal atomshuttling times for transversal H and S gates and transversal entangling gates, which can lead to inaccurate resource estimates The authors emphasize that in the Baseline, H and S gates account for as much as 73% of the total latency Their work focuses on the transversal surface code as the first target for architecture-level modeling and optimization The methodology can be extended to qLDPC codes through appropriate modifications to corresponding modules. The primary architectural contribution is identifying and mitigating the patch rotation bottleneck in NA-based FTQC systems, which is orthogonal to existing compiler optimizations

The evaluation methodology uses eight practical FTQC workloads spanning four representative algorithm classes with various logical-qubit scales and program sizes The control hardware setup assumes state-of-the-art selective laser operations and their latency values for shuttling-free ESM The authors assume a physical error rate of 0.1% for CZ gates, measurement, reset, and AOD–SLM transfers, and 0.01% for single-qubit gates The end-to-end simulation takes approximately 30 hours on the tested AMD EPYC 9684X server. p. 6

The artifact evaluation package is available at https://zenodo.org/records/21542759, and the end-to-end simulation takes approximately 30 hours on a server with an AMD EPYC 9684X processor The fast path takes less than one minute on a laptop-class CPU The code licenses are MIT and data licenses are MIT.

D3-ROT achieves a 2.27× speedup from this single bottleneck alone, with minimal footprint overhead

Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.

Improvements for AI systems

  1. Full-stack architecture simulation framework NAQsim can be used to explore various design choices across the software, classical hardware, and quantum device stacks of neutral-atom platforms by deriving execution time, qubit overhead, and footprint from a single program input.

  2. The system can perform full-stack co-optimization to resolve performance bottlenecks like patch rotations by applying novel optimization ideas in compiler, control hardware, and qubit plane stacks to derive an architecture like D3-ROT which achieves a 2.27x speedup from this single bottleneck alone.

  3. The AI system can generate optimized instruction schedules through the compilation stage using techniques such as the rotation skip algorithm to reduce total rotations by up to 10.2% and improve execution time by up to 35.3%.

  4. The system can implement flexible rotation scheduling policies, including aggregation and distribution, allowing architects to maximize the number of rotation-free instruction layers or minimize the number of rotations within one layer.

  5. The AI system can model and analyze route conflicts during AOD micro-operation execution by introducing a novel concept to accurately determine the minimum delay required to avoid a route conflict, preventing physical atom collisions.

  6. The system can estimate program success rate by incorporating AOD–SLM transfer noise as depolarizing noise, allowing for a more accurate LER evaluation in neutral-atom platforms.

  7. The system can compare the performance of D3-ROT against lattice surgery architectures, showing that D3-ROT reduces the execution time to 25.5x that of the superconducting LS architecture.

Sources

Related papers