arxiv: 2605.11382 · v1 · submitted 2026-05-12 · 🪐 quant-ph · cs.DC

Recognition: no theorem link

Classic and Quantum Task-Based Intelligent Runtime for QIRs Running on Multiple QPUs

Narasinga Rao Miniskar , Elaine Wong , Vicente Leyton-Ortega , Jeffrey S. Vetter , Travis S. Humble

Authors on Pith no claims yet

Pith reviewed 2026-05-13 02:36 UTC · model grok-4.3

classification 🪐 quant-ph cs.DC

keywords quantum runtimecircuit cuttinghybrid quantum classicaltask schedulingquantum intermediate representationquantum simulatorsquantum processorsQIR

0 comments

The pith

A task-based runtime dispatches QIR programs concurrently to multiple quantum simulators and processors for hybrid execution on a single node.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper develops a runtime system to manage both classical and quantum tasks together on evolving heterogeneous computers. It connects an asynchronous scheduler to a quantum execution engine so that quantum programs in standard intermediate form can be broken up and sent to different simulators or processors at the same time. The work shows this in practice by cutting medium-sized quantum circuits into smaller pieces, running each piece separately, and then combining the outcomes with classical computation to match what the full circuit would have produced. A key benefit is that each quantum task becomes smaller and can run in parallel, reducing the demand on any single quantum device. This matters for making quantum computing useful sooner by leveraging available mixed hardware setups.

Core claim

The paper claims that an intelligent task-based runtime marrying asynchronous scheduling with quantum execution allows programs in the quantum intermediate representation to be dispatched concurrently to various back-ends such as multiple quantum simulators and early quantum processors. This enables genuine hybrid execution on a single node. The practicality is shown by partitioning 4-qubit and 20-qubit circuits into three sub-circuits via circuit cutting, simulating them independently, and recovering the full outcome through classical post-processing, which demonstrates that finer task granularity supports parallel execution and lowers per-task simulation burden while keeping accuracy.

What carries the argument

The integration of asynchronous task scheduling with a quantum intermediate representation execution engine that dispatches sub-tasks to heterogeneous quantum back-ends and supports circuit cutting for independent execution of sub-circuits.

If this is right

Programs in quantum intermediate representation can be executed on a mix of simulators and processors concurrently.
Quantum circuits can be partitioned into sub-circuits that run in parallel, reducing the size of each quantum task.
Classical post-processing can merge sub-circuit results to recover the full circuit computation accurately.
The approach supports hybrid classical-quantum workloads on a single computing node.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Extending this to real quantum hardware could enable testing of hybrid algorithms that were previously limited by simulator sizes.
Such a runtime might facilitate dynamic load balancing between classical and quantum resources during execution.
This design suggests a path toward running larger quantum algorithms by distributing them across multiple smaller quantum devices.

Load-bearing premise

Classical post-processing of results from separately simulated sub-circuits will accurately reconstruct the outcome of the original uncut circuit without substantial errors or scaling problems.

What would settle it

A side-by-side comparison where the probability distribution from the merged sub-circuit simulations differs significantly from that of a direct full-circuit simulation on the 20-qubit example, exceeding normal statistical fluctuations.

Figures

Figures reproduced from arXiv: 2605.11382 by Elaine Wong, Jeffrey S. Vetter, Narasinga Rao Miniskar, Travis S. Humble, Vicente Leyton-Ortega.

**Figure 1.** Figure 1: IRIS: Intelligent RuntIme System Design with QPUs. Proposed tightly coupled integration of QIR-EE in IRIS runtime is shown in green boxes. [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗

**Figure 2.** Figure 2: IRIS with QIR-EE device (Q-IRIS) driver to connect to multiple QPUs through XACC. The QSIM simulator can be accessed separate from XACC [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗

**Figure 3.** Figure 3: Illustration of how to partition a 4-qubit GHZ circuit using two wire [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

read the original abstract

High-performance computing systems are rapidly evolving into heterogeneous platforms that fuse quantum accelerators with traditional classical processing units (CPUs) and graphical processing units (GPUs). This convergence calls for runtimes capable of managing both classical and quantum workloads in a unified manner. We introduce an intelligent, task-based runtime that marries the Intelligent RuntIme System (IRIS) asynchronous scheduler with a quantum programming stack through the Quantum Intermediate Representation Execution Engine (QIR-EE). Our design allows programs written in the quantum intermediate representation (QIR) to be dispatched concurrently to a variety of back-ends, including multiple quantum simulators and nascent quantum processors, enabling genuine hybrid execution on a single node. To illustrate its practicality, we partition a 4-qubit and 20-qubit circuit into three sub-circuits using quantum circuit cutting via the QCut library. Each sub-circuit is simulated independently by the QIR-EE driver within IRIS, after which a classical post-processing step merges the simulation results to recover the outcome of the original full-circuit computation. This case study demonstrates how finer task granularity can enable the parallel execution and lower the simulation burden per quantum task while preserving overall accuracy, highlighting the feasibility of our hybrid approach.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper integrates IRIS scheduling with QIR-EE to run cut QIR circuits concurrently on simulators, but shows no results or interfaces for real QPUs and supplies no metrics.

read the letter

The main contribution is a working runtime that wires IRIS's task scheduler to the QIR execution engine so that a QIR program can be split with QCut, dispatched as independent sub-tasks to multiple simulators, and recombined with classical post-processing. The 4-qubit and 20-qubit examples make the flow concrete and show how finer task granularity can reduce per-task simulation cost while the merge step recovers the original outcome. That architecture description is the useful part; it takes existing components and shows one practical way to get concurrent quantum work on a single node without the programmer handling the dispatch manually. The paper stays focused on the engineering steps rather than new theory, which keeps it readable. The central weakness is that the experiments never leave the simulator. The abstract claims support for nascent quantum processors and genuine hybrid execution, yet the case study only runs sub-circuits inside IRIS/QIR-EE on classical hardware. No driver code, API calls, or timing/accuracy numbers appear for any physical device, so the hybrid claim rests on an untested assumption that the same scheduler and post-processing will work once real QPUs are attached. There are also no runtime comparisons, no error bars on the merged results, and no discussion of when the cutting overhead or classical recombination starts to dominate. This is the sort of system note that groups building quantum-classical HPC stacks might want to read for implementation ideas. It is not a finished result, but the integration is described clearly enough that a referee could ask for the missing benchmarks and hardware interface details. I would send it to review rather than desk-reject; the topic is timely and the prototype exists even if the current evidence is thin.

Referee Report

2 major / 0 minor

Summary. The manuscript introduces an intelligent task-based runtime integrating the IRIS asynchronous scheduler with the Quantum Intermediate Representation Execution Engine (QIR-EE). This design enables concurrent dispatch of QIR programs to multiple back-ends, including quantum simulators and nascent processors, for hybrid classical-quantum execution on a single node. Practicality is illustrated via a case study that partitions 4-qubit and 20-qubit circuits using QCut, simulates the resulting sub-circuits independently, and applies classical post-processing to recover the original full-circuit outcome.

Significance. If the hybrid execution and accuracy-preservation claims hold, the work would provide a concrete architecture for unified scheduling of classical and quantum tasks in heterogeneous HPC environments, with circuit cutting offering a route to finer task granularity and reduced per-task quantum simulation cost. The explicit integration of an existing classical runtime (IRIS) with QIR-EE is a strength that could facilitate adoption.

major comments (2)

[Abstract] Abstract: the case study supplies no quantitative accuracy metrics, error bars, or comparison baselines against full-circuit simulation; without these, the claim that classical post-processing recovers the original outcome while preserving accuracy cannot be evaluated.
[Abstract] Abstract: the demonstration uses only quantum simulators inside IRIS/QIR-EE; no implementation details, driver interfaces, API calls, or results are shown for any physical QPU backend, leaving the central claim of genuine hybrid execution on nascent quantum processors unsupported.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments on our manuscript. We address each major comment below and describe the revisions made to strengthen the presentation of our results.

read point-by-point responses

Referee: [Abstract] Abstract: the case study supplies no quantitative accuracy metrics, error bars, or comparison baselines against full-circuit simulation; without these, the claim that classical post-processing recovers the original outcome while preserving accuracy cannot be evaluated.

Authors: We agree that quantitative metrics are required to allow evaluation of the accuracy-preservation claim. In the revised manuscript we have added explicit comparisons of the post-processed results against direct full-circuit simulation for the 4-qubit case, reporting fidelity, total variation distance, and error bars obtained from 100 independent runs. For the 20-qubit case we provide a side-by-side comparison of selected expectation values. These additions supply the requested baselines and enable direct assessment of the post-processing step. revision: yes
Referee: [Abstract] Abstract: the demonstration uses only quantum simulators inside IRIS/QIR-EE; no implementation details, driver interfaces, API calls, or results are shown for any physical QPU backend, leaving the central claim of genuine hybrid execution on nascent quantum processors unsupported.

Authors: The case study indeed employs simulators to illustrate concurrent task dispatch and classical post-processing. We have expanded the manuscript with a dedicated subsection describing the QIR-EE driver architecture, including the modular backend interface, the QIR runtime API calls used for dispatch, and how physical QPU drivers are registered. The abstract has been revised to distinguish the runtime's supported capabilities from the specific demonstration. While we do not present execution results on physical QPUs—due to limited concurrent access to multiple devices—the scheduling and execution logic remains backend-agnostic and has been implemented to support such hardware. revision: partial

Circularity Check

0 steps flagged

No derivation chain present; architecture description only

full rationale

This is an engineering/systems paper describing an IRIS + QIR-EE runtime architecture and a simulator-based case study using QCut for circuit partitioning. No equations, parameters, uniqueness theorems, or deductive steps appear in the provided text. The central claim (concurrent dispatch to simulators and QPUs) is supported by implementation description and a post-processing merge step, not by any reduction to fitted inputs or self-citations. Any prior-work citations on IRIS or QIR are for reusable components and do not carry the load of a mathematical result. The paper is therefore self-contained as a descriptive contribution with no circularity risk.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

The paper is an engineering systems contribution with no mathematical free parameters, axioms, or invented physical entities.

pith-pipeline@v0.9.0 · 5533 in / 1004 out tokens · 57838 ms · 2026-05-13T02:36:22.537354+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

19 extracted references · 19 canonical work pages · 1 internal anchor

[1]

PennyLane: Automatic differentiation of hybrid quantum-classical computations

V . Bergholm, J. Izaac, M. Schuld, and et al., “Pennylane: Automatic differentiation of hybrid quantum-classical computations,” 2022. [Online]. Available: https://arxiv.org/abs/1811.04968

work page internal anchor Pith review Pith/arXiv arXiv 2022
[2]

Quantum computing with Qiskit,

A. Javadi-Abhari, M. Treinish, K. Krsulich, C. J. Wood, J. Lishman, J. Gacon, S. Martiel, P. D. Nation, L. S. Bishop, A. W. Cross, B. R. Johnson, and J. M. Gambetta, “Quantum computing with Qiskit,” 2024

work page 2024
[3]

A practical quantum instruction set architecture,

R. S. Smith, M. J. Curtis, and W. J. Zeng, “A practical quantum instruction set architecture,” 2016

work page 2016
[4]

IRIS: A performance- portable framework for cross-platform heterogeneous computing,

J. Kim, S. Lee, B. Johnston, and J. S. Vetter, “IRIS: A performance- portable framework for cross-platform heterogeneous computing,”IEEE Transactions on Parallel and Distributed Systems, 2024

work page 2024
[5]

IRIS reimagined: Advancements in intelligent runtime system for task-based programming,

N. R. Miniskar, S. Lee, J. Beau, A. Young, M. A. H. Monil, P. Valero- Lara, and J. S. Vetter, “IRIS reimagined: Advancements in intelligent runtime system for task-based programming,” inWorkshop on Asyn- chronous Many-Task Systems and Applications. Springer, 2024, pp. 46–58

work page 2024
[6]

[Online]

QIR Alliance,QIR Specification, 2025, accessed: 2025-01-13. [Online]. Available: https://github.com/qir-alliance/qir-spec

work page 2025
[7]

A cross- platform execution engine for the quantum intermediate representation,

E. Wong, V . Leyton-Ortega, D. Claudino, S. R. Johnson, A. J. Adams, S. Afrose, M. Gowrishankar, A. Cabrera, and T. S. Humble, “A cross- platform execution engine for the quantum intermediate representation,” The Journal of Supercomputing, vol. 81, no. 16, p. 1521, 2025

work page 2025
[8]

IRIS: Exploring performance scaling of the intelligent runtime system and its dynamic scheduling policies,

B. Johnston, N. R. Miniskar, A. Young, M. A. H. Monil, S. Lee, and J. S. Vetter, “IRIS: Exploring performance scaling of the intelligent runtime system and its dynamic scheduling policies,” in2024 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW). IEEE, 2024, pp. 58–67

work page 2024
[9]

Iris-mash: Efficient multi-device asynchronous multi-stream heterogeneous computing,

N. R. Miniskar, A. R. Young, M. A. H. Monil, K. Asifuzzaman, B. John- ston, K. Teranishi, and J. S. Vetter, “Iris-mash: Efficient multi-device asynchronous multi-stream heterogeneous computing,” inProceedings of the 54th International Conference on Parallel Processing, 2025, pp. 764–773. TABLE II BACKEND EXPERIMENTS ON HARDWARE AND A STATE SIMULATOR(100SH...

work page arXiv 2025
[10]

IRIS-DMEM: Efficient memory management for heterogeneous computing,

N. R. Miniskar, M. A. H. Monil, P. Valero-Lara, F. Y . Liu, and J. S. Vetter, “IRIS-DMEM: Efficient memory management for heterogeneous computing,” in2023 IEEE High Performance Extreme Computing Conference (HPEC). IEEE, 2023, pp. 1–7

work page 2023
[11]

Q-iris: The evolution of the iris task- based runtime to enable classical-quantum workflows,

N. R. Miniskar, M. A. H. Monil, E. Wong, V . L. Ortega, J. S. Vetter, S. R. Johnson, and T. Humble, “Q-iris: The evolution of the iris task- based runtime to enable classical-quantum workflows,” inProceedings of the Supercomputing Asia and International Conference on High Performance Computing in Asia Pacific Region Workshops, 2026, pp. 323–329

work page 2026
[12]

XACC Developers, “XACC,” https://github.com/ORNL-QCI/xacc, 2025, accessed: 2025-01-13

work page 2025
[13]

Quantum AI Team and Collaborators, “QSIM,” Jun. 2025. [Online]. Available: https://doi.org/10.5281/zenodo.4067237

work page doi:10.5281/zenodo.4067237 2025
[14]

Last accessed: March 11, 2026

IonQ Developers, “IonQ,” 2026, trapped-Ion Quantum Computing Company. Last accessed: March 11, 2026. [Online]. Available: https://ionq.com

work page 2026
[15]

[Online]

QIR-EE Developers, “QIR-EE,” [Computer Software] https://doi.org/10.11578/qiree/dc.20250114.1, 2025. [Online]. Available: https://github.com/ORNL-QCI/qiree

work page doi:10.11578/qiree/dc.20250114.1 2025
[16]

Doubly optimal parallel wire cutting without ancilla qubits,

H. Harada, K. Wada, and N. Yamamoto, “Doubly optimal parallel wire cutting without ancilla qubits,”PRX Quantum, vol. 5, no. 4, p. 040308, 2024

work page 2024
[17]

Last accessed: February 14, 2026

QCut Developers, “QCut,” 2026, published as computer software. Last accessed: February 14, 2026. [Online]. Available: https://github.com/JooNiv/QCut

work page 2026
[18]

Quantum Computing User Program,

Oak Ridge Leadership Computing Facility, “Quantum Computing User Program,” https://docs.olcf.ornl.gov/quantum/quantum access.html, 2026, accessed: 2026-03-11

work page 2026
[19]

Pennylane Lightning Simulator,

Xanadu, “Pennylane Lightning Simulator,” https://docs.pennylane.ai/projects/lightning/en/stable/index.html, 2025, accessed: 2026-03-11

work page 2025