Europe has clearly jumped into the global race to achieve practical quantum, though perhaps a step later (by a year or two) than the U.S. and China. Impressively, the European quantum effort has made up ground quickly and carved out a distinct early focus on HPC-QC integration, exploring different quantum modalities, and leveraging government-academic-industry collaboration. These were among the take-home messages from the ISC 2023 Birds of as Feather (BOF)[i] session, The First Generation of European Quantum Computers – A Discussion with the Hosting Sites.
Members from each of the six EuroHPC JU designated quantum computing hosting sites were present to provide brief updates, and Rene Spencer Chatwell, program officer for the EuroHPC Joint Undertaking, provided an overview. Don’t forget that the early call for proposals from potential hosting sites was only issued in late 2021. EuroHPC JU oversees the project.
Thus far, the six designated hosting sites include:
Chatwell provided a good overview, “We are looking at €100 million in total investment for the upcoming years. Our main strategic goals included: to pursue this integration of quantum computers into HPC systems; the most important thing is that each quantum computer has to be integrated in a genuine HPC system at a datacenter, [and] that it’s not in some generic lab environment at the university, or the Research Institute. Each hosting HPC system was…to provide an operational performance of at least 4 petaflops.”
“We want to see the development of the technology-agnostic software stack – that is the user experience should be embedded in all the six systems. We want to see the adaptation and optimization of load schedulers that prioritize these hybrid computations. And, of course, we want to see the identification of genuine use cases with high impact. The final goal we set out was the development of hybrid algorithms and applications that will be upcoming that would directly steer towards these developments. And what might be important for you as a user community is how to get access to these machines. So, I’m just talking about the European Union’s share, of course,” said Chatwell.
Broadly, the six EuroHPC quantum centers participate in regional consortia, have both distinct and overlapping goals, and seek to create a kind of pan-Europe federated group of hybrid HPC-quantum resources. Not surprisingly, actual installation of quantum computers (procurement and development) is at different stages at the various centers. Providing reasonable access to the new systems is also a key objective. Chatwell said a process involving peer review, at least early on, would be involved in allocating time on the various systems.
Developing benchmarks is another broad goal.
Chatwell said, “We’re looking at benchmarks [and] set out a set of very generic rules, which we believe everyone can rally behind. The benchmarks we have envisioned should be standardized, transparent, and non-discriminatory with respect to technology. They should have the widest possible intellectual acceptance, and they should of course evolve with technological developments. They should then measure the efficiency of the hybrid integration and provide feedback to the user communities today. In this [sense] they should allow for comparison of performance across the different technologies. And we would strive for the identification of specific algorithms that would benefit from quantum computing, and that eventually demonstrate quantum computing advantage. We have also envisioned some measures of standardization because we have seen that standardization [provides] a great opportunity for international collaboration.”
These are clearly ambitious plans and many of the details are still evolving. The GENCI-CEA-led EuroQCS-France consortium effort is a good example, and Sabine Mehr, chief quantum projects officer, GENCI, provided some insight into plans and progress. Planned quantum systems will be integrated with GENCI’s Joliot-Curie 22 petaflops supercomputer.
“The overall goal of what we were trying to achieve with this consortium was to collaborate with the other consortium in setting up pan-European federated infrastructure as was expected by EuroHPC. We plan to acquire a photonic quantum computer based on a linear optical scheme, because this technology we saw as quite mature and there’s a clear path for fault-tolerant quantum computing. We will be integrating this new device in the same flavor as was done or as it’s being done for the Pascal neutral atom quantum simulator,” said Mehr.
“We will try to build as much as possible on the experience that we gained through the HPCQS initiative to extend it to this new device, which is no longer a quantum simulator, [but will be a] quantum computer. We will also rely on national initiatives such as the HQI (hybrid HPC-QC) initiative in France. For our approach, we wanted also to [deliver] the user experience on two tracks: one of them being called agnostic through the Qaptiva offering from Eviden for people who want to just program in a very agnostic way; but for users [who] want to dive a bit deeper into the physics we, will also be providing a hardware-specific environment for photonic quantum computing.”
Different centers are exploring different qubit modalities (photonic, neutral atom, superconducting, etc.) as it is not clear which type, if any, will become predominant or if some may emerge as best for specific applications. All of the speakers cited challenges around bringing quantum devices into a traditional HPC environment.
“Installing and operating new devices that we don’t know much about in computing centers, which are not very friendly to these systems initially, is a challenge. We also want to support the adoption of these new systems in the HPC-QC converged environment, from a broad base of users throughout Europe and also in France,” said Mehr, reflecting a chorus of similar statements from host center speakers.
The IT4Innovations effort (Czech Republic) is working with a broad consortium of 14 partners, including nine European countries, noted Branislav Jansik, supercomputing services director, IT4Innovations National Supercomputing Center (Czech Republic). “We’ll have access to a number of hardware systems, at least four quantum computers within the context of consortium,” he said. “Our approach to the quantum computing at this point is primarily attributed to the accelerator approach, where basically we envisage the offloading of the GPU of tasks that may be accelerated by a quantum computing,” said Jansik.
The project is developing a job scheduler, based on a cost-scheduling approach. “When we put it [into operation] we’ll support the classical computing tasks that go along with the quantum computing. We plan to deploy this system in such a way that once the software is deployed, for the integration with a quantum computer, we’ll be able to extend the same software stack across the set of supercomputers within the consortium so that our users will be able to access the quantum device from multiple locations,” said Jansik, who envisions a kind of common software console acting as a gateway to these resources.
Current IT4I plans call for deploying semiconductor-based superconducting qubit systems. IT4I is working on a star-shaped system topology using ~12 qubits. This topology, said Jansik, allow a reduction in the number of sub-operations required.
Generally speaking, the BOF discussion did not dig deep into technical details. The conversation reflected common challenges around: early deployment of quantum devices in HPC centers; extensive software development with a goal of a single stack able to support hybrid HPC-quantum computing; building a workforce; and developing contacts with user communities to develop firm use cases. In this broad sense, the EuroHPC quantum efforts are still young but have accomplished much.
The broad goal of creating a federated system of accessible hybrid HPC-quantum systems throughout Europe, coupled with Euro-centric technology, is the central tenet. With the planning mostly completed, the build-out is underway at the six designated centers. It will be interesting to check on progress at ISC next year.
Unfortunately, the BOF was not recorded by ISC. Perhaps its organizers will post a summary.
[i] Description of BOF – “The EuroHPC JU has recently announced the hosting sites for the first six European quantum computers. The systems will be installed at IT4I-Czechia, GENCI-CEA-France, LRZ-Germany, CINECA-Italy, PSNC-Poland and BSC-CNS-Spain. With the systems also comes the mandate to integrate them with the existing HPC resources at the sites, establishing and enabling true HPCQC capabilities for the European user community. These systems will further Europe’s leading efforts in HPCQC integration and the projects tackle the challenges of seating and operating these systems: from where to locate these quantum system with respect to differing infrastructures, cooling systems and facilities to determining the level and type of physical and logical integration between the HPC and QC systems; from establishing suitable programming abstractions not solely relying on quantum physics to dynamic local and global scheduling environments; and from user education and training to the creation of a pan-European HPCQC community. In this BOF, we invite representatives from the sites to give an overview of their approaches to host and operate their systems as well as the motivation that drove the site-specific choices. This will be followed by a discussion between the centers and the HPC community, especially the targeted European user community in the audience, on how to make best use of these systems and how to integrate them into daily scientific workflows. The goal of the session is to increase awareness for the planned systems and to get early user input on how these new services will have to be offered.”