Inno4scale

Innovative Algorithms for Applications on European Exascale Supercomputers
Logo Inno4scale

Inno4scale identified and provided funding to support the development of advanced algorithms and applications for upcoming European exascale systems.

Project Aims

Inno4scale aimed to enable disruptive innovation in HPC application software by supporting high-risk, high-reward algorithmic studies focused on extreme scalability. Through an open call, the project selected 22 "innovation studies" that explored unconventional algorithmic concepts aimed at overcoming core challenges in extreme-scale computing: increasing parallelism, reducing communication bottlenecks, enabling heterogeneity, and optimizing performance and efficiency. The project served as a testing ground for novel HPC algorithm designs that might otherwise be too risky for traditional R&D frameworks. HLRS led the communication, PR, and outreach activities to invite participation in Inno4scale across the scientific community, and to spread awareness of the project’s outcomes and impact by disseminating results as proofs of concept.

Project Achievements

  • 22 innovation studies conducted, covering a broad range of domains including CFD, quantum simulation, particle physics, and uncertainty quantification.
  • Proof-of-concept implementations led to algorithmic improvements yielding up to 100x performance speedups in selected use cases (e.g., XSCALE, CBM4Scale).
  • Several studies introduced asynchronous and mixed-precision algorithms, improving efficiency and enabling better use of GPU-CPU systems (e.g., SCALE-TRACK, Ex3S).
  • Developed novel numerical methods for uncertainty quantification and Bayesian inference, achieving over 90% cost savings in industrial and geophysical use cases.
  • Results from Inno4scale are already being considered for integration into production-level HPC applications and inspire ongoing and future research activities.

Future Objectives

  •  Many promising results from Inno4scale studies are being further developed and integrated into domain-specific HPC codes.
  • Follow-up proposals are in preparation under EuroHPC JU and national funding lines to continue work on algorithmic innovation at exascale.
  • Several studies initiated collaborations that are continuing beyond the project’s end, fostering a growing research community around scalable HPC algorithms.
  • Ongoing dissemination activities include scientific publications, workshops, and potential open-source releases of the developed code prototypes.
Runtime

01. July 2023 -
30. June 2025

Website
Departments

Public Relations

Funding

EuroHPC Joint Undertaking
HORIZON-EUROHPC-JU-2022-ALG-02

Project partners

  • Barcelona Supercomputing Center (BSC)
  • Scapos AG
  • High-Performance Computing Center Stuttgart (HLRS)
  • Partnership for Advanced Computing in Europe (PRACE)

Funding