A computing paradigm can be defined as the way a computing machine processes information. It is characterized by the coding scheme used to represent information, the architecture of the computer, and the technology it is based upon.
In the last few years, the architecture of standard computers (a.k.a. the von Neumann architecture) has evolved in many ways.
The clock frequency of single-core processors has reached its limit, and large multi-core microprocessors are now widely used, for example.
The introduction of a range of co-processors such as GPUs, TPUs, and FPGAs has enhanced the performance of these machines on more complex data structures like large matrices and tensors.
As standard computers have continued to improve, new
ways to process information—like neuromorphic and quantum computing—have also emerged, giving rise to radically different paradigms such as the one being developed under the European QuCube quantum computing project.
CEA research on new computing paradigms spans from advances in circuit design enabled by new technology to novel software and programming techniques.
In this multidisciplinary context, CEA benefits from experts in both embedded and high-performance computing, and is engaged in several advanced research programs on neuromorphic and quantum computing.
R&D on new computing paradigms will be key to the development of future high-added-value applications using more pervasive and energy efficient solutions for artificial intelligence, advanced cryptography and optimization techniques.