Architectural Design and Orchestration of Heterogeneous
Quantum-Classical Computing Systems
J.A. Bravo-Montes
1,2 a
, Miriam Bastante
1 b
and Cyril Allouche
3
1
HPC & Quantum, Eviden Iberia, Madrid, Spain
2
Faculty of Informatics, Complutense University of Madrid, Madrid, Spain
3
Research Quantum, Eviden, France
Keywords:
Heterogeneous Architecture, HPC, Quantum Computing, Quantum Emulators, Qaptiva.
Abstract:
This study presents a comprehensive multilevel software architecture for integrating quantum and classical
computing systems in High-Performance Computing environments. The proposed software architecture im-
plements a hierarchical approach that seamlessly connects quantum circuit development with execution in both
emulated and physical quantum processors. The system is structured in distinct layers: a Circuit Generator in-
terface, HPC programming environment with specialized APIs and tools, NISQ calibration layer for managing
quantum noise, and an integration layer that orchestrates the hybrid workload distribution. We demonstrate the
implementation of this architecture in two leading European supercomputing facilities: the TGCC and the JSC.
These implementations showcase the capability of the architecture to manage hybrid quantum-classical work-
flows, incorporating both local and remote QPUs while maintaining security and efficiency. The effectiveness
of the architecture was validated through implementation of the VQE algorithm, demonstrating its practical
application. The results highlight the architecture’s ability to efficiently manage heterogeneous computational
resources, provide secure access to quantum hardware, and facilitate the development and execution of hybrid
quantum-classical algorithms. This study contributes to the advancement of quantum computing integration
in HPC environments, providing a scalable framework for future quantum applications.
1 INTRODUCTION
High-Performance Computing (HPC) systems are
nowadays at the forefront of modern computing tech-
nology by leveraging the power of parallel process-
ing. This paradigm allows solving complex scien-
tific and engineering problems that are fundamental to
the development of new advances in different realms,
such as computer science, geoscience, and physics
simulation (Park et al., 2012). HPC systems often de-
pend on deeply interconnected nodes that support a
large amount of data movement and shared informa-
tion tasks, allowing multithread processing through-
out the integration of conventional Central Processing
Units (CPUs) and the more recent Graphics Process-
ing Units (GPUs) (Humble et al., 2021).
However, HPC faces several setbacks that hinder
its application to specific problems of exponential na-
ture, which are intractable even for the most advanced
a
https://orcid.org/0000-0002-1935-0997
b
https://orcid.org/0009-0001-4076-3753
HPC systems. To address these problems, the emer-
gence of quantum computing in recent decades offers
a solution by harnessing quantum mechanics proper-
ties like superposition, entanglement or interference
(Forcer et al., 2002). These basic principles have been
demonstrated in small-scale systems, and efforts to
create large-scale platforms are being pursued glob-
ally, reaching $40 billion public investment in 2024
(Insider, 2025). Despite these efforts, the idea of a
fully quantum computer replacing classical ones is
far from reality and does not really hold up, which
is why there is increasing interest in the development
of Quantum Processing Units (QPUs) to complement
other processing units like CPUs and GPUs.
Historically, a technological revolution has al-
ready taken place with the arrival of GPUs in HPC
centers. This is why integrating QPUs into the HPC
model promises to elevate them to the same level of
impact that GPUs had in the past. Companies around
the world are in the race to develop and fully leverage
this technology, and although QPUs are still in the
early stages, various approaches are already under-
154
Bravo-Montes, J., Bastante, M. and Allouche, C.
Architectural Design and Orchestration of Heterogeneous Quantum-Classical Computing Systems.
DOI: 10.5220/0013653600004525
In Proceedings of the 1st International Conference on Quantum Software (IQSOFT 2025), pages 154-161
ISBN: 978-989-758-761-0
Copyright © 2025 by Paper published under CC license (CC BY-NC-ND 4.0)
way to implement them, aiming to mitigate the main
challenges arising from their quantum nature, such as
decoherence and scalability (Gill et al., 2024).
While fault-tolerant QPUs are still under devel-
opment, HPC centers are integrating quantum envi-
ronments that can leverage their HPC infrastructure.
These environments enable the simulation of quan-
tum circuit behavior using classical computing re-
sources for various purposes, such as error-correction
codes or quantum algorithm development. This ap-
proach makes use of HPC cluster infrastructure and
allows the execution of larger quantum circuits with
reduced execution times, simplifying the implemen-
tation of heterogeneous quantum-classical workflows
for large-scale industrial problems (Beck et al., 2024).
However, there are several challenges and limita-
tions to integrate QPUs into existing HPC architec-
tures. These challenges arise from the agnosticism of
current quantum paradigms, the lack of memory re-
sources, and the need for smooth communication be-
tween the numerous components of the systems.
Throughout this paper, we explore the different
aspects of integration and orchestration of quantum-
classical components in heterogeneous systems, par-
ticularly in HPC environments. This study has been
structured as follows: section 2 gives an overview
of existing processing options for executing jobs
and available examples; section 3 details Eviden’s
proposed heterogeneous ecosystem architecture and
other real-world implementations. Finally section 4
concludes the paper by discussing the scalability and
flexibility of these architectures, underscoring their
role in advancing quantum computing adoption in
HPC.
2 BACKGROUND
HPC centers are facilities that house supercomputers
and computer clusters designed to process large-scale
and complex computational tasks. These centers al-
low researchers and industries to explore and solve
problems in a wide range of applications, including
scientific simulations, Artificial Intelligence (AI), and
climate modeling (Park et al., 2012).
Although traditional HPC architectures rely on the
combination of CPUs and GPUs, the rise in quantum
computing has increased interest in integrating QPUs
into existing architectures (Beck et al., 2024).
To achieve this heterogeneous system, it is impor-
tant to orchestrate all its different components so that
they can work together. HPC-Quantum environments
require specialized middleware to enable interaction
between classical and quantum processors. An effec-
tive implementation and integration of the building
blocks of such architectures will optimize the work-
flow and lead to more efficient solutions in the in-
dustry. However, the agnostic nature of current QPU
technologies, as well as frameworks, compilers, and
schedulers, makes this task anything but trivial.
The current structure of typical HPC architectures
is based on clusters of different types of processing
units, including CPUs, GPUs or Parallel Processing
Units (PPUs) (Gao and Zhang, 2016). This heteroge-
neous platform requires several layers before it can be
used for real applications. In addition, integrating real
(or emulated) QPUs into the system adds complexity
to the already sophisticated workflow.
Upon initiation of a computational task by the
user, an analysis is performed to determine the most
suitable type(s) of processing units required for exe-
cution. Subsequently, a scheduler manages the tem-
poral coordination and sequencing of computational
processes. When considering computational environ-
ments, there are several options depending on the task
at hand. HPC clusters provide the power needed for
large-scale simulations, while quantum emulators en-
able the simulation of quantum algorithms on classi-
cal hardware. Finally, QPUs introduce a novel com-
putational layer based on a new paradigm.
2.1 HPC Clusters
HPC clusters operate with several processing units ca-
pable of working together to perform complex com-
putations at high speeds, leveraging the power of par-
allel processing and distributing tasks across nodes.
With the integration of GPUs into the HPC realm,
there has been significant improvement in perfor-
mance and speed. These systems have applications in
a wide range of fields, including AI, DNA sequenc-
ing, and stock trading.
The world’s fastest supercomputer is “El Capi-
tan” (Hewlett Packard Enterprise, 2025) located at
Lawrence Livermore National Laboratory (LLNL).
This supercomputer can operate at 1.74 EFlops/s
and has an optimized processor for nuclear sim-
ulations, AI, and scientific research. In Europe,
Italy’s “HPC6” supercomputer is positioned the 5th
most powerful supercomputer in the TOP500 list as
November 2024 (TOP500 Project, 2024). This cluster
has a peak performance of 606 PFloops/s with main
applications in optimization of industrial facilities and
accuracy of geological and fluid dynamic studies for
CO2 storage (Eni S.p.A., 2025). Another remark-
able mention in the 9th position is the “Leonardo”
supercomputer, which allows computations at 306.31
PFlops/s.
Architectural Design and Orchestration of Heterogeneous Quantum-Classical Computing Systems
155
Finally, in Spain the “Marenostrum 5 ACC”, is
the most powerful HPC cluster of the country, with
a peak of 249.44 PFlops/s (Barcelona Supercomput-
ing Center, 2025) and is 11th in the TOP500 list. It
is at the Barcelona Supercomputing Center and uses
an energy-efficient architecture for AI and language
model development.
2.2 Quantum Emulators
While some companies work on physical QPUs, an-
other line of research focuses on developing and op-
timizing a complete quantum software stack ready to
take advantage of the power of physical QPUs once
they become operational. Quantum computer emu-
lators enable researchers to test and optimize algo-
rithms, diagnose hardware performance, and validate
theoretical models without requiring access to phys-
ical quantum hardware, the resources of which are
usually scarce.
Although emulating a QPU with high hardware
resources is computationally expensive, particularly
in terms of memory, it allows us to model and con-
trol the noise that naturally arises in a quantum com-
puter (LaRose, 2018) (Georgopoulos et al., 2021). By
studying different types of noise, we can overcome
certain hardware limitations, leading to more power-
ful QPUs with a larger number of logical qubits.
Similarly to the physical case, the rapid advance-
ment of research has led to the development of numer-
ous models for simulating quantum computing. On
the one hand, we have quantum emulators based on
the digital, analog and annealing paradigm. Examples
include Cirq (Google) (AI, 2024), Qiskit Aer (IBM
Quantum) (Quantum, 2024), and EMU-TN (Pasqal)
(Bidzhiev et al., 2023); the latter also uses tensor net-
works to simulate larger-scale quantum systems. D-
Wave (Systems, 2024a) has an emulator for its quan-
tum annealing computer. Moreover, there are also
references capable of emulating the three comput-
ing paradigms: digital, analog and annealing, such as
Qaptiva (Eviden, 2024).
2.3 QPUs
Since any two-level quantum system can be used
to create a qubit, researchers are developing various
types of qubits, each with characteristics that make
them better suited for specific applications. For exam-
ple, superconducting qubits benefits of already exist-
ing methods and processes that have been improved
during last years. IBM (Condor) (AbuGhanem,
2024), Google (Willow) (Research, 2023) and Rigetti
(Ankaa-3) (Rigetti, 2024) are betting on this tech-
nology. Another advanced approach is the trapped-
ion qubits, which have longer decoherence times than
superconducting qubits (Linke et al., 2017). Com-
panies like Quantinuum (H2) (Quantinuum, 2024),
IonQ (Forte) (IonQ, 2022) and Alpine Quantum Tech-
nologies (Marmot) ((AQT), 2024) are focusing their
efforts on this technology.
Although these are currently the most advanced
technologies, many others are in development with
very promising characteristics. There is a line of re-
search based on photonic qubits, which could theoret-
ically create a universal quantum computer (Romero
and Milburn, 2024). Following these results, Xanadu
(X-Series) (Xanadu, 2024), Quantum Computing Inc
(Dirac-3) (Diraq, 2024), Alice&Bob (Alice & Bob
Quantum Computing, 2024) or PsiQuantum (Omega)
(PsiQuantum, 2025) are putting their efforts into this
research. Neutral atoms, which uses optically trapped
atoms and Rydberg interactions for quantum oper-
ations, allowing for scalability and connectivity be-
tween qubits. Pasqal (Orion-Beta) (Pasqal, 2025) and
QuEra (Aquila) (QuEra, 2022) are the leading com-
panies.
In addition, we can take a look at other tech-
nologies such as topoconductors, developed by Mi-
crosoft (Majorana1) (Aasen et al., 2025), quantum
dots qubits, where we find Quobly (Quobly, 2024)
and Diraq (Diraq, 2024), and the more different ap-
proach developed by D-Wave with its Advantage2
Prototype (Systems, 2024b).
3 PROPOSAL
The proposed software architecture in Figure 1 imple-
ments a multilevel hierarchical approach for efficient
integration of quantum and classical systems. In the
top layer, the Circuit Generator provides a specialized
interface for the generation and manipulation of quan-
tum circuits, establishing an entry point for the de-
velopment of quantum algorithms. The second layer
of the architecture implements three essential com-
ponents: specialized Application Programming Inter-
faces (APIs) for systematic interaction, mathematical
and scientific libraries that provide fundamental func-
tions for quantum computing, and a robust set of per-
formance and debugging tools.
The Noisy Intermediate-Scale Quantum (NISQ)
systems calibration layer is a mandatory component
that addresses the challenges inherent in NISQ. This
layer implements two essential components: “gate
noise” and “qubit noise”, facilitating the incorpora-
tion of specific noise models that allow accurate cali-
bration of the emulated hardware characteristics with
IQSOFT 2025 - 1st International Conference on Quantum Software
156
Tools (perf, debug...)
Math/Science libs
APIs
Circuit Generator
HPC Programming Environment
NISQ Calibra
on
Integrator
Hybrid quantum-classical workload/resource manager
Classical Systems
On-premise
Quantum Systems
Quantum
Emulator
CPU
GPU
Remote
Figure 1: Heterogeneous software architecture for the inte-
gration of HPC and Quantum Computing resources.
the target QPU (Bravo-Montes et al., 2024).
The integration layer forms the central compo-
nent of the system by implementing a dedicated hy-
bridization node that acts as the main orchestrator.
This component manages the efficient distribution of
tasks between classical and quantum systems, opti-
mizes the use of resources, and ensures optimal per-
formance in hybrid computing environments. The dy-
namic adaptive capability of this integrator enables
the efficient management of available computational
resources while implementing robust security proto-
cols that guarantee the integrity and confidentiality of
operations throughout the task distribution and execu-
tion process.
This software architecture culminates in a lower
layer that implements two complementary execution
systems that operate synergistically. The Classic Sys-
tems integrate CPUs for traditional processing, GPUs
for computational acceleration, and a quantum emula-
tor for algorithm emulation and optimization. In par-
allel, Quantum Systems incorporate both on-premises
and remote QPUs, thereby providing flexible access
to real quantum hardware.
The proposed software architecture facilitates
the adaptive and efficient management of classical-
quantum computational resources, allowing the sys-
tem to dynamically adjust to the specific requirements
of each quantum emulation or execution job.
3.1 Heterogeneous Ecosystem
Architectures
Eviden (Eviden, 2025a) has developed a heteroge-
neous software architecture specifically designed to
integrate quantum computing into HPC infrastruc-
tures. Its design enables the implementation of a com-
plete and transparent workflow, covering all stages
from algorithm development to execution on real
hardware, whether through emulators or QPUs.
The architecture is structured into interconnected
layers, each performing specific functions in man-
aging quantum jobs, including authentication, sub-
mission, scheduling, execution, and result retrieval.
Regardless of the specific infrastructure utilized, the
workflow in Eviden’s quantum environments follows
a general framework that ensures interoperability and
optimization in algorithm execution.
The process begins with user authentication in
myQLM (Eviden, 2025b), an open-source framework
designed for the development and optimization of
quantum algorithms in a flexible environment. Dur-
ing this stage, Qaptiva Access (Eviden, 2024) man-
ages authentication through a robust client-server ar-
chitecture, enabling access via the myQLM client or
through JupyterLab. This system ensures a secure
multi-user environment, capable of managing concur-
rent sessions and integrating external users via au-
thentication mechanisms that preserve the integrity
and confidentiality of operations.
Once authenticated, users can submit jobs from
myQLM to Qaptiva (Eviden, 2024) through Qaptiva
Access. During this stage, the system performs an
initial validation of the job to ensure its compatibil-
ity with the execution infrastructure and compliance
with computational requirements. Depending on the
complexity of the algorithm, the job may require op-
timization and adaptation to the underlying architec-
ture. In this context, Qaptiva Access provides com-
patibility with external compilers, allowing the com-
pilation process to be adapted according to the char-
acteristics of the target hardware.
The scheduler system, such as Simple Linux Util-
ity for Resource Management (SLURM) (SchedMD,
2025), allocates jobs to available computational re-
sources. This mechanism facilitates efficient work-
load distribution, optimizing resource utilization
across quantum emulators, hybrid infrastructures or
physical QPUs.
When a job requires execution on a real quantum
processor, the QPU integration layer is activated. This
component enables connections to quantum hard-
ware, whether on-premise or remote, through secure
communication channels. To connect to third-party
Architectural Design and Orchestration of Heterogeneous Quantum-Classical Computing Systems
157
QPUs, a dedicated front-end is necessary to act as a
translator between the user software and the physical
architecture of the quantum processor. During this
process, quantum circuits are optimized to adapt to
the hardware topology, adjusting quantum gates to the
set of native operations and applying specific calibra-
tion parameters if necessary.
The workflow concludes with result collection
and analysis. The data generated by the QPU un-
dergoes an initial processing stage by the front-end,
which verifies its integrity before sending it to Qaptiva
for post-processing. At this stage, advanced statisti-
cal analysis techniques and necessary transformations
can be applied to facilitate interpretation. Finally, the
results are delivered to myQLM, where users can per-
form detailed interpretations, evaluate performance
metrics and explore potential improvements to their
algorithms.
The Tr
`
es Grand Centre de Calcul (TGCC) (CEA
TGCC, 2025) and the J
¨
ulich Supercomputing Centre
(JSC) (J
¨
ulich, 2025) are two leading European infras-
tructures that have adopted a heterogeneous ecosys-
tem similar to the one under discussion for the in-
tegration of quantum computing with HPC centers.
A detailed description of these infrastructures is pro-
vided below.
3.1.1 Tr
`
es Grand Centre de Calcul (TGCC)
The TGCC is one of France’s leading supercomput-
ing infrastructures, capable of hosting petaflop-scale
supercomputers. Initially conceived to host Curie,
the first petaflop-scale machine in France, TGCC
has evolved with the incorporation of advanced sys-
tems such as Joliot-Curie (CEA, 2025b), Cobalt, and
Topaze, which support scientific and industrial appli-
cations through the Centre de Calcul Recherche et
Technologie (CCRT) of the Commissariat
`
a l’
´
Energie
Atomique (CEA) (CEA, 2025a).
In the field of quantum computing, TGCC has im-
plemented a preconfigured software ecosystem within
the “ccc-quantum container” image. This configu-
ration integrates myQLM, specialized libraries such
as Pulser and Perceval, Qaptiva Access, Qaptiva, and
a communication front-end for QPUs. Figure 2 il-
lustrates the complete workflow of this architecture,
highlighting the interactions between users, CEA, and
Pasqal’s quantum hardware.
From the user’s perspective, interaction with
TGCC’s quantum environment begins in myQLM or
through specialized libraries. Pulser (Pulser Devel-
opment Team, 2022), developed by Pasqal, allows
the design and simulation of laser pulse sequences
for controlling neutral-atoms in both analog and digi-
tal quantum computing. Perceval (Quandela, 2025),
login
myQLM
Pulser
User
CEA
Pasqal
Qap
va
Access
push
submit
Return Qap va result
Return Qap va result
Send to QPU
Return result
submit
job
submit job
Queue
Cast to
myQLM
Create
sequence
Translate
to Pulser result
Execute job
on QPU
(using Remote QPU)
Slurm
Qap
va
QPU
Frontend
qpu.server
Figure 2: Software architecture deployed at TGCC.
developed by Quandela, provides tools for emulat-
ing photonic quantum computing, enabling the mod-
eling of linear optical circuits with components such
as beam splitters and phase modulators.
Once the quantum algorithm has been designed,
the user can choose between different execution
strategies. For small-scale emulations, execution
can be performed on a TGCC computing node us-
ing myQLM. If greater computational capacity is re-
quired, the job is sent via Qaptiva Access to Qap-
tiva, which allows the emulation of a higher number
of qubits and the inclusion of quantum noise models.
If the user wishes to execute the algorithm on a real
QPU, the job is transferred to the Pasqal front-end,
which translates and optimizes the circuit for execu-
tion on a neutral-atom QPU.
Finally, the results are processed and analyzed in
Qaptiva before being sent to myQLM, where users
can assess algorithm performance and iteratively re-
fine their implementations based on the obtained data.
3.1.2 J
¨
ulich Supercomputing Centre (JSC)
The JSC has an advanced infrastructure for the inte-
gration of quantum and classical computing through
the J
¨
ulich UNified Infrastructure for Quantum Com-
puting (JUNIQ) (J
¨
ulich, 2025). This environment
provides access to various quantum technologies via
JupyterHub, facilitating the development and exper-
imentation of quantum algorithms within a flexible
and efficient framework. JSC hosts multiple quantum
systems, including JUQCS (De Raedt et al., 2019),
a gate-based quantum computing emulator capable
of simulating circuits with up to 43 qubits; Eviden
Qaptiva, which enables the emulation of up to 41
qubits and noise modeling; and specialized hardware
such as D-Wave Advantage JUPSI (J
¨
ulich, 2025) and
PASQAL Fresnel. Additionally, projects such as
QSolid (QSolid, 2022), focused on the construction
of high-fidelity superconducting qubits, and Open-
SuperQplus (OpenSuperQPlus, 2023), dedicated to
developing a superconducting qubit-based quantum
IQSOFT 2025 - 1st International Conference on Quantum Software
158
computer, are currently under development.
Figure 3 illustrates the workflow of a hybrid
job integrating several of the aforementioned sys-
tems within JSC, highlighting the interaction between
quantum systems and the classical HPC infrastruc-
ture.
Figure 3: Software architecture deployed at JSC.
The workflow at JSC begins with the development
of algorithms in JUNIQ through myQLM. Quantum
circuits are optimized using the external compiler Par-
ityQC (ParityQC, 2025), which generates circuits in
Qiskit language. These circuits are then converted
into myQLM format via an interoperability module
included within myQLM. Once transpiled, the cir-
cuits are sent to Qaptiva Access, where they are man-
aged by the ParTec Scheduler. This scheduler over-
sees the transfer of circuits to Qaptiva, optimizing
workload distribution across the heterogeneous in-
frastructure and ensuring that algorithms are executed
in the most suitable environment based on their com-
putational characteristics. These circuits may be ex-
ecuted on Qaptiva’s emulated QPUs or on a remote
QPU through Qruise (QRuise, 2025), which enables
the interconnection of Qaptiva with the QPUs pro-
vided by IQM.
After execution, the results are processed and vi-
sualized in myQLM, allowing for subsequent detailed
analysis and iterative optimization of the algorithms.
3.2 Application of Heterogeneous
Architecture for the VQE Algorithm
The Variational Quantum Eigensolver (VQE) algo-
rithm is a hybrid quantum-classical approach de-
signed to solve eigenvalue problems in quantum me-
chanics. Its primary application lies in material simu-
lations and quantum chemistry, where it approximates
the ground states of complex quantum systems. The
algorithm operates through a parameterized quantum
circuit, whose parameters are iteratively optimized
using a classical algorithm to minimize the expected
value of a given Hamiltonian, thereby estimating the
system’s energy efficiently.
QPU
Figure 4: Example for Heterogeneous Quantum-Classical
VQE Algorithm.
For efficient execution, it is essential to distribute
the computational workload effectively, particularly
in heterogeneous environments. Figure 4 illustrates
the workflow of a hybrid quantum-classical job within
an HPC system, demonstrating how classical and
quantum tasks seamlessly interleave to achieve an op-
timal balance of computational resources. This scal-
able architecture facilitates the execution of large-
scale hybrid jobs, driving advancements in the sim-
ulation of complex quantum systems with high com-
putational demands. The process follows three main
stages:
1. Connection to the HPC Cluster: The pro-
cess begins with this step, establishing a link between
computational resources and hybrid job management
tools. This connection is crucial for coordinating the
distributed execution of tasks and ensuring efficient
communication across different computational levels.
Resource allocation within the cluster allows both
classical and quantum computation to run in parallel,
optimizing the overall system performance.
2. Work Distribution Between CPU and QPU:
Once the connection is established, the workload is
distributed between the CPU and the QPU. The CPU
handles parameter optimization using numerical al-
gorithms, while the QPU performs quantum calcula-
tions by evaluating the expected value of the Hamil-
tonian for the current parameters. Depending on the
available infrastructure, the QPU may be a quantum
Architectural Design and Orchestration of Heterogeneous Quantum-Classical Computing Systems
159
emulator or a remote quantum device, introducing
challenges in latency management and response times
within the hybrid system.
3. Execution of the VQE Algorithm: The pro-
cess starts with an initial set of parameters, which
are iteratively adjusted in the CPU based on measure-
ments obtained from the QPU. In each iteration, the
QPU uses the updated parameters to compute the ex-
pected energy of the system. This feedback cycle con-
tinues until convergence is reached, meaning the vari-
ation in expected energy falls below a defined thresh-
old. Convergence ensures that a good approximation
of the quantum system’s ground state has been found.
To enhance efficiency, execution planning occurs
at two levels. A high-level scheduler oversees job exe-
cution within the HPC cluster, managing resource ac-
cess and workload distribution. Meanwhile, a low-
level scheduler optimizes communication between
the CPU and QPU, reducing data transfer latency and
improving computational efficiency.
4 CONCLUSIONS
The integration and orchestration of quantum-
classical components in heterogeneous systems rep-
resent a significant advancement in the development
of quantum computing, particularly in HPC environ-
ments. As analyzed throughout this article, numer-
ous companies and institutions have recognized the
potential of heterogeneous ecosystems, where CPUs,
GPUs, and QPUs work together to tackle complex
problems in optimization, material simulation, drug
development, cryptography, and more.
In this context, we have presented two software
architectures implemented in leading European super-
computing centers, TGCC and JSC, where quantum-
classical integration is based on the principles out-
lined in this study. The versatility of this architecture
allows compatibility with a wide range of technolo-
gies and the incorporation of various QPUs through
the creation of specific front-ends for their access.
Additionally, its hardware-agnostic approach enables
interoperability with different compilers and alterna-
tive schedulers to SLURM, optimizing computational
efficiency based on the specific requirements of each
application.
The infrastructure implemented at TGCC and JSC
demonstrates that quantum-classical computing is not
only viable but also constitutes a scalable solution
adaptable to technological advancements. Through
tools like Qaptiva Access, the integration of quan-
tum computing into classical supercomputing envi-
ronments is facilitated, enabling the efficient and flex-
ible development of hybrid algorithms. This con-
vergence between classical and quantum computing
not only accelerates the exploration of new applica-
tions but also contributes to the advancement of ap-
plied quantum computing, laying the foundation for
broader adoption of these systems in the future.
REFERENCES
Aasen, D., Aghaee, M., Alam, Z., Andrzejczuk, M., An-
tipov, A., Astafev, M., and et al. (2025). Roadmap to
fault tolerant quantum computation using topological
qubit arrays. arXiv.
AbuGhanem, M. (2024). Ibm quantum computers: Evolu-
tion, performance, and future directions. https://arxiv.
org/abs/2410.00916.
AI, G. Q. (2024). Cirq: Quantum circuit simulation. https:
//quantumai.google/cirq/simulate/simulation. Ac-
cessed: Apr. 01, 2025.
Alice & Bob Quantum Computing (2024). Alice &
bob: Quantum computing technology. https://
alice-bob.com/quantum-computing/technology/. Ac-
cessed: Apr. 01, 2025.
(AQT), A. Q. T. (2024). Marmot - the first commercial 19-
inch rack-mounted quantum computer. https://www.
aqt.eu/products/marmot/. Accessed: Apr. 01, 2025.
Barcelona Supercomputing Center (2025). Marenostrum
supercomputer. https://www.bsc.es/es/marenostrum/
marenostrum. Accessed: Apr. 01, 2025.
Beck, T., Baroni, A., Bennink, R., Buchs, G., P
´
erez, E.
A. C., Eisenbach, M., and et al. (2024). Integrat-
ing quantum computing resources into scientific hpc
ecosystems. Future Generation Computer Systems,
161:11–25.
Bidzhiev, K., Wennersteen, A., Beji, M., Dagrada, M.,
D’Arcangelo, M., Grijalva, S., and et al. (2023).
Cloud on-demand emulation of quantum dynamics
with tensor networks. arXiv.
Bravo-Montes, J., Bastante, M., Botella, G., del Barrio, A.,
and Garc
´
ıa-Herrero, F. (2024). A methodology to se-
lect and adjust quantum noise models through emula-
tors: benchmarking against real backends. EPJ Quan-
tum Technology, 11(1):71.
CEA (2025a). Centre de calcul recherche et technologie.
https://www-ccrt.cea.fr/. Accessed: Apr. 03, 2025.
CEA (2025b). Joliot-curie supercomputer. https://
www-hpc.cea.fr/en/Joliot-Curie.html. Accessed: Apr.
03, 2025.
CEA TGCC (2025). Tr
`
es grand centre de cal-
cul (tgcc). https://hpc.cea.fr/tgcc-public/en/html/
tgcc-public.html. Accessed: Apr. 02, 2025.
De Raedt, H., Jin, F., Willsch, D., Willsch, M., Yoshioka,
N., Ito, N., and et al. (2019). Massively parallel quan-
tum computer simulator, eleven years later. Computer
Physics Communications, 237:47–61.
Diraq (2024). Diraq. https://diraq.com/. Accessed: Apr. 01,
2025.
IQSOFT 2025 - 1st International Conference on Quantum Software
160
Eni S.p.A. (2025). Supercomputing and artifi-
cial intelligence at eni. https://www.eni.com/
en-IT/actions/energy-transition-technologies/
supercomputing-artificial-intelligence/
supercomputer.html. Accessed: Apr. 01, 2025.
Eviden (2024). Qaptiva 800 range. https:
//eviden.com/wp-content/uploads/2024/09/
Eviden-Qaptiva-800-Quantum-Computing-brochure-en.
pdf. Accessed Apr. 03, 2025.
Eviden (2025a). Eviden website. https://eviden.com/. Ac-
cessed Apr. 03, 2025.
Eviden (2025b). myqlm documentation. https://myqlm.
github.io/index.html. Accessed: Apr. 02, 2025.
Forcer, T., Hey, T., Ross, D., and Smith, P. (2002). Su-
perposition, entanglement and quantum computation.
Quantum Information & Computation, 2.
Gao, Y. and Zhang, P. (2016). A survey of homogeneous
and heterogeneous system architectures in high per-
formance computing. In Proceedings of the IEEE
International Conference on Smart Cloud (Smart-
Cloud), pages 170–175.
Georgopoulos, K., Emary, C., and Zuliani, P. (2021). Mod-
eling and simulating the noisy behavior of near-term
quantum computers. Physical Review A, 104(6).
Gill, S. S., Cetinkaya, O., Marrone, S., Claudino, D.,
Haunschild, D., Schlote, L., and et al. (2024). Quan-
tum computing: Vision and challenges. arXiv.
Hewlett Packard Enterprise (2025). Lawrence
livermore national laboratory. https:
//www.hpe.com/es/es/compute/hpc/
lawrence-livermore-national-laboratory.html. Ac-
cessed: Apr. 01, 2025.
Humble, T. S., McCaskey, A., Lyakh, D. I., Gowrishankar,
M., Frisch, A., and Monz, T. (2021). Quantum com-
puters for high-performance computing. IEEE Micro,
41(5):15–23.
Insider, T. Q. (2025). Public annual report 2025. The Quan-
tum Insider. Accessed: May 26, 2025.
IonQ (2022). Ionq forte. https://ionq.com/
quantum-systems/forte. Accessed: Apr. 01, 2025.
J
¨
ulich, F. (2025). J
¨
ulich supercomputing centre (jsc). https:
//www.fz-juelich.de/en/ias/jsc. Accessed: Apr. 03,
2025.
LaRose, R. (2018). Distributed memory techniques for clas-
sical simulation of quantum circuits. https://arxiv.org/
abs/1801.01037. Accessed: Apr. 01, 2025.
Linke, N. M., Maslov, D., Roetteler, M., Debnath, S., Fig-
gatt, C., Landsman, K. A., and et al. (2017). Experi-
mental comparison of two quantum computing archi-
tectures. Proceedings of the National Academy of Sci-
ences, 114(13):3305–3310.
OpenSuperQPlus (2023). Opensuperqplus. https://
opensuperqplus.eu/. Accessed: Apr. 04, 2025.
ParityQC (2025). Parityos. https://parityqc.com/parityos.
Accessed: Apr. 02, 2025.
Park, B.-H., Kim, Y., Kim, B.-D., Hong, T., Kim, S., and
Lee, J. (2012). High performance computing: Infras-
tructure, application, and operation. Journal of Com-
puting Science and Engineering, 6:280–286.
Pasqal (2025). Pasqal quantum computing pro-
cessor brochure. https://www.pasqal.com/
wp-content/uploads/2025/03/2503 Pasqal
Quantum-Computing-Processor Borchure.pdf.
Accessed: Apr. 02, 2025.
PsiQuantum (2025). Psiquantum announces omega, a
manufacturable chipset for photonic quantum com-
puting. https://www.psiquantum.com/featured-news/
omega. Accessed: Apr. 01, 2025.
Pulser Development Team (2022). Pulser documen-
tation. https://pulser.readthedocs.io/en/stable/index.
html. Accessed: Apr. 02, 2025.
QRuise (2025). Qruise os. https://www.qruise.com/
products/qruise-os. Accessed: Apr. 02, 2025.
QSolid (2022). Qsolid. https://www.q-solid.de/. Accessed:
Apr. 04, 2025.
Quandela (2025). Perceval documentation. https://perceval.
quandela.net/docs/v0.13/index.html. Accessed: Apr.
02, 2025.
Quantinuum (2024). Quantinuum system model h2 product
data sheet. https://docs.quantinuum.com/systems/
data sheets/Quantinuum%20H2%20Product%
20Data%20Sheet.pdf. Accessed: Apr. 02, 2025.
Quantum, I. (2024). Qiskit aersimulator. https:
//docs.quantum.ibm.com/api/qiskit/0.30/qiskit.
providers.aer.AerSimulator. Accessed: Apr. 01, 2025.
QuEra (2022). Aquila, our 256-qubit analog quantum com-
puter. https://www.quera.com/aquila. Accessed: Apr.
01, 2025.
Quobly (2024). Vlsi: Our path to scaling
up silicon spin qubits. https://quobly.io/
vlsi-our-path-to-scaling-up-silicon-spin-qubits/.
Accessed: Apr. 01, 2025.
Research, G. (2023). Google unveils willow quan-
tum chip. https://blog.google/technology/research/
google-willow-quantum-chip/. Accessed: Apr. 01,
2025.
Rigetti (2024). Quantum computing. solved. https://www.
rigetti.com/. Accessed: Apr. 01, 2025.
Romero, J. and Milburn, G. (2024). Photonic quantum com-
puting. arXiv.
SchedMD (2025). Slurm workload manager documenta-
tion. https://slurm.schedmd.com/documentation.html.
Accessed: Apr. 04, 2025.
Systems, D.-W. (2024a). D-wave quantum: Quantum com-
puting solutions. https://www.dwavequantum.com/.
Accessed: Apr. 01, 2025.
Systems, D.-W. (2024b). Per-qpu solver proper-
ties and schedules: Advantage2 prototype2.6.
https://docs.dwavequantum.com/en/latest/
quantum research/solver properties specific.html#
advantage2-prototype2-6. Accessed: Apr. 02, 2025.
TOP500 Project (2024). Top500 list - november 2024.
https://top500.org/lists/top500/2024/11/. Accessed:
Apr. 01, 2025.
Xanadu (2024). X-series quantum computers. https://www.
xanadu.ai/products/x-series. Accessed: Apr. 01, 2025.
Architectural Design and Orchestration of Heterogeneous Quantum-Classical Computing Systems
161