Image
As global environmental, urban, and computing challenges expand, traditional computational frameworks frequently hit structural limits. Simulating entire cityscapes, tracing microscale atmospheric pollution, or predicting the flash progression of a wildfire requires massive processing capacity. However, raw power is no longer enough; next-generation engineering demands scalability, transparency, and energy awareness.
As a premier EuroHPC Joint Undertaking Center of Excellence, HiDALGO2 bridges this gap. By co-designing next-generation numerical methods, workflow automation pipelines, and machine learning architectures, our global consortium is unlocking extreme-scale computing for open science.
To help researchers, developers, and policymakers navigate our project milestones, we have consolidated our latest 16 core publications, benchmarks, and open datasets into five interconnected scientific pillars.
Theme I: Urban Digital Twins & Sustainable Infrastructure
Developing viable mitigation plans for modern smart cities requires modelling physical infrastructure at an individual building resolution. HiDALGO2 accelerates this deployment by introducing automated cloud-native engineering into classic supercomputing environments, eliminating manual deployment friction.
1. Ktirio Urban Building: A Computational Framework for City Energy Simulations Enhanced by CI/CD Innovations on EuroHPC Systems, by Luca Berti, Vincent Chabannes, Gwennolé Chappron, Javier Cladellas, Abdoulaye Diallo, Maryam Maslek Elayam, Philippe Pinçon & Christophe Prud’homme.
2. Open-source complex-geometry 3D fluid dynamics for applications with unpredictable heterogeneous dynamic high-performance computing loads, by J. Bakosi, Széchenyi Egyetem, University of Győr, Hungary
Theme II: Microscale Environmental Modelling & Climate Adaptation
From vehicular emissions to active forest fires, climate adaptation relies heavily on capturing high-resolution, real-time spatial trends. Our research couples microscale atmospheric dispersion tracking with hyper-fast predictive models to support environmental policy and crisis response.
3. Estimating the air quality standard exceedance areas and the spatial representativeness of urban air quality stations applying microscale modelling, by F. Martín, V. Rodrigues, J.L. Santiago, J. Sousa, J. Stocker, S. Janssen, R. Jackson, F. Russo, M.G. Villani, G. Tinarelli, D. Barbero, R. San José, J.L. Pérez-Camanyo, G. Sousa-Santos, L. Tarrason, J. Bartzis, I. Sakellaris, Z. Horváth, L. Környei, X. Jurado, P. Thunis
4. Using dispersion models at microscale to assess long-term air pollution in urban hot spots: A FAIRMODE joint intercomparison exercise for a case study in Antwerp, by F. Martín, S. Janssen, V. Rodrigues, J.Sousa, J.L. Santiago, E.Rivas, J.Stocker, R.Jackson, F.Russo, M.G.Villani, G.Tinarelli, D.Barbero, R.San José, J.L.Pérez-Camanyo, G.Sousa Santos, J.Bartzis, I.Sakellaris, Z.Horváth, L.Környei, B.Liszkai, C.Cuvelier
5. AI for Global Challenges: Case Studies in Urban Solar Exposure & Wildfire Management, by Giorgos Filandrianos, Angeliki Dimitriou, Vasiliki Kostoula, Nikolaos Chalvantzis, David Caballero, Luis Torres, Michal Kulczewski, Javier Cladellas, Zoltán Horváth, Harald Köstler, Konstantinos Nikas, Dimitrios Tsoumakos, Giorgos Stamou.
6. Simulation of Wildfires Using EuroHPC Resources: Challenges and Opportunities, by David Caballero, Leydi Laura Salazar, Ángela Rivera & Luis Torres.
Theme III: Graph Analytics, Network Science & Explainable AI (XAI)
The intersection of HPC and AI demands deep transparency. As deep learning models expand into physical domains, HiDALGO2 is actively building frameworks that replace traditional “black-box” systems with explainable, semantic structural networks.
7. Structure Your Data: Towards Semantic Graph Counterfactuals, by Angeliki Dimitriou, Maria Lymperaiou, Georgios Filandrianos, Konstantinos Thomas, Giorgos Stamou (Proceedings of the 41st International Conference on Machine Learning (ICML), PMLR 235:10897-10926, 2024)
8. Affected Buildings Network Topology: Utilising Graph Convolutional Networks (GCNs) and transductive link prediction to map and discover hidden topological shading dependencies across dense cityscape networks.
Theme IV: Energy-Aware Supercomputing & Scalability Profiling
True supercomputing innovation is a balancing act. Pushing simulation meshes across multi-thousand-core setups dramatically drives up energy footprints, meaning codes must become energy-aware to remain operationally sustainable.
9. Prediction model of performance–energy trade-off for CFD codes on AMD-based cluster, by Marcin Lawenda, Łukasz Szustak, László Környei
10. Evaluating AMD EPYC CPU architectures on CFD applications, by Marcin Lawenda, Łukasz Szustak, László Környei, Flavio Cesar Cunha Galeazzo, Paweł Bratek
Theme V: Advanced Parallel Engineering & Numerical Portability
At the very foundation of the HiDALGO2 project lies core parallel software engineering. By tackling cross-platform portability, memory latency bottlenecks, and load imbalances, we ensure our workflows run fluidly across heterogeneous clusters.
11. Partition deactivation with load balancing for parallel flow simulations, by J. Bakosi, Széchenyi Egyetem, University of Győr, Hungary
12. Breaking Down LLM Inference: A preliminary performance analysis of sparsified transformers, by Ioanna Tasou; Petros Anastasiadis; Panagiotis Mpakos; Dimitrios Galanopoulos; Nectarios Koziris; Georgios Goumas.
13. Complex-Geometry 3D Computational Fluid Dynamics with Automatic Load Balancing, by József Bakosi, Mátyás Constans, Zoltán Horváth, Ákos Kovács, László Környei, Marc Charest, Aditya Pandare, Paula Rutherford and Jacob Waltz
14. Fostering Uncertainty Quantification in Global Challenges with mUQSA Toolkit, by Michał Kulczewski, Bartosz Bosak, Piotr Kopta, Wojciech Szeliga & Tomasz Piontek
15. Profiling and Optimisation of Multicard GPU Machine Learning Jobs, by Marcin Lawenda, Kyrylo Khloponin, Krzesimir Samborski, Łukasz Szustak on the Journal Wiley – Concurrency and Computation: Practice and Experience.
16. Efficient allocation of image recognition and LLM tasks on multi-GPU system, by Marcin Lawenda, Krzesimir Samborski, Kyrylo Khloponin, Łukasz Szusta
17. Uncut-GEMMs: Communication-aware matrix multiplication on multi-GPU nodes, by Petros Anastasiadis, Nikela Papadopoulou, Nectarios Koziris and Georgios Goumas.
Bringing Open Science into the Future
Every script, dataset, and manuscript produced under the HiDALGO2 umbrella is designed to be fully reproducible, transparent, and aligned with European Open Science standards. We invite the broader HPC, data science, and environmental modelling communities to deploy our open-source tools, review our energy benchmarks, and collaborate with us as we continue to co-design the future of high-performance simulations.