Paving the Way for the Future of Supercomputing: New Advances in HPC
CiTIUS looks to the future of heterogeneous and quantum supercomputing with three new contributions aimed at portable programming of CPU+GPU architectures, as well as promoting the development of next-generation quantum software.
Three new contributions from CiTIUS to the field of heterogeneous and quantum computing have just been unveiled at the IEEE Cluster 2025 conference, one of the world's leading forums in the field of High-Performance Computing (HPC). The meeting, whose 27th edition recently took place in Edinburgh, brings together experts from universities, research centers, and industry each year to discuss the latest advances in cluster computing system architectures (sets of connected computers that work together as if they were a single machine, thus multiplying their computing power).
Quantum computing is one of the great technological challenges of our time. Although still in the early stages of development, it promises to profoundly transform society, with applications ranging from accelerating the discovery of new drugs and materials to optimizing logistical and energy networks or enhancing communication security. Its potential lies in the ability to solve in seconds problems that today would take years for the most powerful supercomputers. However, current quantum computers are limited and diicult to scale, which is why much of the research focuses on creating simulation tools and hybrid environments that allow for experimentation with quantum algorithms and lay the foundations for the future.
On the other hand, portable heterogeneous programming oers a promising future in the field of HPC. Given the wide variety of accelerators available today, this approach enables the development of heterogeneous solutions without the need to maintain separate codebases implemented using dierent languages or frameworks. Thanks to programming models such as SYCL, it is possible to develop a single source code that can be executed on dierent platforms regardless of the manufacturer, thus providing portability and improving software programmability. These advances pave the way for more versatile and eicient systems, capable of adapting to the growing diversity of hardware present across all scales of modern computing systems.
In this context, a team from CiTIUS (a research centre co-funded by the European Union through the Galicia ERDF Programme 2021–2027), has just presented three new scientific contributions with the aim of facilitating the development of next-generation heterogeneous and quantum solutions. The first of these, presented by researcher Silvia Rodríguez Alcaraz in collaboration with Franscisco F. Rivera (CiTIUS), David L. Vilariño (USC), and Rubén Laso (University of Vienna). The work focused on evaluating the performance of portable heterogeneous solutions in systems that combine general-purpose processors (CPUs) with graphics processing units (GPUs). To this end, they developed ihp-sycl using Intel oneAPI, Intel's implementation of SYCL. This program employs iterative mathematical problems with dierent computational characteristics as case studies, applying dynamic load balancing between the host and the device at runtime. The study was conducted on GPUs with dierent specifications and from dierent manufacturers, specifically Intel integrated devices (graphics included within the processor itself) and NVIDIA discrete devices (high-performance independent graphics cards).For his part, researcher Francisco Javier Cardama presented in Edinburgh the result of two works developed in collaboration with researchers Tomás Fernández Pena (CiTIUS) and Miguel Leal (USC). The first is NetQMPI, a Python library to program distributed quantum systems and inspired by MPI or Message Passing Interface, a supercomputing standard that allows dierent processors to work in parallel coordination through message sending and receiving. NetQMPI translates this concept to the quantum field and introduces simple commands (qsend and qrecv), which simplify the programming of distributed algorithms in simulated quantum networks, such as NetSquid and SimulaQron. Its goal is to accelerate the development of scalable and portable quantum software, allowing the programming of the future distributed quantum computers.
The second proposal is SYCL QPU, a framework that allows classical processors and simulated quantum units to collaborate in a single unified program. It's built on SYCL, a standard that facilitates parallel programming on dierent types of hardware, and uses Intel's oneAPI DPC++ implementation to make it compatible with multiple architectures; moreover, it integrates with Qiskit Aer: a simulator developed by IBM that replicates the functioning of quantum computers, allowing for the eicient coordination of quantum and classical tasks in the same environment. With this combination, the proposal oers a solid foundation to advance towards a future where quantum and classical computing work together in real supercomputing applications.
New scientific steps that confirm CiTIUS's eort to anticipate the major technological challenges of the future, in a field that advances rapidly and where the border between basic research and practical application becomes increasingly blurred. Works like those published in IEEE Cluster 2025 help pave the way for eicient use of portable heterogeneous programming or for quantum computing to stop being a promise and become a real tool at the service of science and society.
Three new contributions from CiTIUS to the field of heterogeneous and quantum computing have just been unveiled at the IEEE Cluster 2025 conference, one of the world's leading forums in the field of High-Performance Computing (HPC). The meeting, whose 27th edition recently took place in Edinburgh, brings together experts from universities, research centers, and industry each year to discuss the latest advances in cluster computing system architectures (sets of connected computers that work together as if they were a single machine, thus multiplying their computing power).
Quantum computing is one of the great technological challenges of our time. Although still in the early stages of development, it promises to profoundly transform society, with applications ranging from accelerating the discovery of new drugs and materials to optimizing logistical and energy networks or enhancing communication security. Its potential lies in the ability to solve in seconds problems that today would take years for the most powerful supercomputers. However, current quantum computers are limited and diicult to scale, which is why much of the research focuses on creating simulation tools and hybrid environments that allow for experimentation with quantum algorithms and lay the foundations for the future.
On the other hand, portable heterogeneous programming oers a promising future in the field of HPC. Given the wide variety of accelerators available today, this approach enables the development of heterogeneous solutions without the need to maintain separate codebases implemented using dierent languages or frameworks. Thanks to programming models such as SYCL, it is possible to develop a single source code that can be executed on dierent platforms regardless of the manufacturer, thus providing portability and improving software programmability. These advances pave the way for more versatile and eicient systems, capable of adapting to the growing diversity of hardware present across all scales of modern computing systems.
In this context, a team from CiTIUS (a research centre co-funded by the European Union through the Galicia ERDF Programme 2021–2027), has just presented three new scientific contributions with the aim of facilitating the development of next-generation heterogeneous and quantum solutions. The first of these, presented by researcher Silvia Rodríguez Alcaraz in collaboration with Franscisco F. Rivera (CiTIUS), David L. Vilariño (USC), and Rubén Laso (University of Vienna). The work focused on evaluating the performance of portable heterogeneous solutions in systems that combine general-purpose processors (CPUs) with graphics processing units (GPUs). To this end, they developed ihp-sycl using Intel oneAPI, Intel's implementation of SYCL. This program employs iterative mathematical problems with dierent computational characteristics as case studies, applying dynamic load balancing between the host and the device at runtime. The study was conducted on GPUs with dierent specifications and from dierent manufacturers, specifically Intel integrated devices (graphics included within the processor itself) and NVIDIA discrete devices (high-performance independent graphics cards).For his part, researcher Francisco Javier Cardama presented in Edinburgh the result of two works developed in collaboration with researchers Tomás Fernández Pena (CiTIUS) and Miguel Leal (USC). The first is NetQMPI, a Python library to program distributed quantum systems and inspired by MPI or Message Passing Interface, a supercomputing standard that allows dierent processors to work in parallel coordination through message sending and receiving. NetQMPI translates this concept to the quantum field and introduces simple commands (qsend and qrecv), which simplify the programming of distributed algorithms in simulated quantum networks, such as NetSquid and SimulaQron. Its goal is to accelerate the development of scalable and portable quantum software, allowing the programming of the future distributed quantum computers.
The second proposal is SYCL QPU, a framework that allows classical processors and simulated quantum units to collaborate in a single unified program. It's built on SYCL, a standard that facilitates parallel programming on dierent types of hardware, and uses Intel's oneAPI DPC++ implementation to make it compatible with multiple architectures; moreover, it integrates with Qiskit Aer: a simulator developed by IBM that replicates the functioning of quantum computers, allowing for the eicient coordination of quantum and classical tasks in the same environment. With this combination, the proposal oers a solid foundation to advance towards a future where quantum and classical computing work together in real supercomputing applications.
New scientific steps that confirm CiTIUS's eort to anticipate the major technological challenges of the future, in a field that advances rapidly and where the border between basic research and practical application becomes increasingly blurred. Works like those published in IEEE Cluster 2025 help pave the way for eicient use of portable heterogeneous programming or for quantum computing to stop being a promise and become a real tool at the service of science and society.