AERO - Developing an open-source software ecosystem for the emerging EU processor designs

HORIZON.2.4 - Digital, Industry and Space MAIN PROGRAMME
HORIZON.2.4.2 - Key Digital Technologies

Start date 1 January 2023
End date 31 December 2025

project website: https://aero-project.eu/

DII Team Leader: Sergio Saponara

 

The push for European innovation and autonomy in computational infrastructure and chip design has led to many fruitful projects working to achieve this goal. Crucial among these is the EU Processor Initiative (EPI) that is working on the design and development of the first EU processor.
AERO aims to complement the efforts of the EPI project by developing the open-source software ecosystem required to not only improve the efficiency of the EPI hardware but also accelerate and ease the processor’s integration into the cloud. AERO will enhance the performance, energy and security efficiency and facilitate and motivate user migration to the future European cloud platform, infrastructure and ecosystem.

 

 NEWS 

Pisa, Sept 18 2026  - The AERO project announces AXHEL (ARM aXceleration for Homomorphic Encryption Library), an open-source C++ library developed by the University of Pisa to accelerate the low-level arithmetic used by Homomorphic Encryption (HE) workloads on AArch64 processors. AXHEL exploits the Arm Scalable Vector Extension (SVE) to accelerate key modular-arithmetic and polynomial-transform operations used by Microsoft SEAL. These computational kernels are common to many Homomorphic Encryption libraries and can potentially be reused beyond SEAL, while its integration remains transparent to existing SEAL applications.

Technological Advancements in AERO
Accelerating Homomorphic Encryption Computation on Arm Processors AXHEL reduces the computational overhead of Homomorphic Encryption by applying architecture-aware SIMD vectorization based on the Arm Scalable Vector Extension (SVE) to modular and polynomial arithmetic.

Key features include:

  • SVE-vectorized modular addition, subtraction, multiplication, reduction, and fused multiply-add.
  • SVE-accelerated Number Theoretic Transform (NTT) and inverse NTT kernels.
  • Vector-length-agnostic AArch64 implementation with portable scalar fallbacks.
  • Integration with Microsoft SEAL without application-level code changes.

“AXHEL brings architecture-aware acceleration for Homomorphic Encryption to Arm Neoverse processors, achieving speedups of around 1.5× on key arithmetic kernels. It integrates transparently with Microsoft SEAL while targeting computational building blocks common to other HE libraries.” says Massimiliano Donati, University of Pisa 

partners

ICCS - National Technical University of Athens (NTUA)
The University of Manchester
SiPearl
IInstitute of Computer Science - Foundation for Research & Technology - Hellas
Codeplay Software
Sednai
KTM Innovation
UBITECH
University of Geneva
Università di Pisa
Red Hat
VIRTUAL OPEN SYSTEMS
European Health and Digital Executive Agency (HaDEA)
The University of Manchester Faculty of Science and Engineering