The teams / labs (and their representatives) involved in the HOLIGRAIL consortium are:
- Univ. Rennes, IRISA, Inria Taran – Olivier Sentieys (coordinator), Silviu Filip
- computer arithmetic including number representations, computer architecture especially, hardware acceleration, design space exploration, architecture optimization, and deep neural network compression including quantization error evaluation, and low-precision training.
- CEA List/LIAE, Université Paris Saclay – Olivier Bichler
- quantization-aware training and coding theory in deep neural networks and their link with technologies (non-volatile memory, in-memory computing) and software/hardware implementation (N2D2 framework).
- Inria Corse, Université Grenoble Alpes – Fabrice Rastello, Christophe Guillon
- compiler optimization, data-movement, pattern-specific analysis and optimization.
- Grenble-INP, TIMA/SLS, Université Grenoble Alpes – Adrien Prost-Boucle, Olivier Muller
- number representation, hardware acceleration, hardware support for quantization, memory accesses, and throughput-friendly compression.
- INSA Lyon, CITI, Inria Emeraude – Florent De Dinechin, Anastasia Volkova
- computer arithmetic, operator optimization and fusion, and high-level synthesis.
- CEA List/LVML, Université Paris Saclay – Mohamed Tamaazousti
- tensor methods, distillation techniques for neural network compression, deep learning representations.
