The European Rebecca project aims to democratize the development of new, next-generation artificial intelligence (AI) systems. To do this, the project will develop a complete European hardware and software stack around a RISC-V CPU, which will provide significantly higher levels of performance, energy efficiency, security and protection.
The project will contribute to realizing business and social opportunities by validating and demonstrating its approach against real-world use cases and application-based benchmarks from the domains of smart home appliances, power generation, infrastructure inspection, aviation, automotive, and health.
According to information published in the Community Research and Development Information Service (Cordisfor its acronym in English) of the European Commission, cutting-edge technologies and significant scientific and technological advances will be used in several key relevant domains, including processing units, hardware accelerators, reconfigurable hardware, tightly coupled interconnected chiplets, co-design and co-development tools of hardware/software, middleware, libraries and artificial intelligence frameworks.
Creating a new chip
In terms of hardware, Rebecca will develop a new chip that will consist of two tightly coupled chipsets incorporating a multi-core RISC-V, a neuromorphic AI accelerator, a programmable matrix AI accelerator, an AI accelerator that uses a hierarchical processing architecture. , a DNN accelerator, hardware reconfigurable, near-memory processing, and memory encryption.
For its part, the software will integrate optimized libraries of software, middleware and artificial intelligence into the system, taking full advantage of the underlying novel hardware.
The Rebecca platform will be complemented by a novel hardware/software design space exploration tool that will enable the development of highly efficient Rebecca-based systems. The project will also provide the means for security and protection modeling and verification for hardware and software created from the earliest design stages.
Led by the Greek University Institute for Telecommunications Systems Research, the Rebecca project has a consortium made up of 18 entities from Greece, Spain, Germany, the Netherlands, Italy, Sweden, Turkey and Lithuania. Spanish participation is represented by the National Supercomputing Center, the University of Castilla-La Mancha and the companies Solver Machine Learning and Fent Innovative Software Solutions.
Likewise, the project, which began in February 2023 and will end in July 2026, has a budget of 8,498,328 euros, of which 2,744,319 euros are financed by the Horizon research program of the European Commission.
