(ref. ANR-18-CE25-0004)
The main objective of this project is to build tools to automatically adapt data format in numerical codes, in order to better benefit from modern vector micro-architectures and improve code performances while not sacrifying accuracy.
Click here to read the full project proposal.
Members
- DALI, Univ. Perpignan Via Domitia / LIRMM : Philippe LANGLOIS, David PARELLO and Guillaume REVY (project leader)
- IN2P3 / LUPM : Luisa ARRABITO, Johan BREGEON
Publications
International conferences with review committees and proceedings
-
Analyzing the impact of floating-point precision adaptation in iterative programs.
G. Revy.
28th IEEE Symposium on Computer Arithmetic (ARITH 2021), Conférence virtuelle, 2021.
-
Augmenter l'impact de la vectorisation par transformation automatique de programmes.
H. L. Boudi, Ph. Langlois, and G. Revy..
Conférence d'informatique en Parallélisme, Architecture et Système (Compas'2020), Lyon, France, 2020.
Some talks
-
Présentation du projet PADOC.
G. Revy.
Session projet des RAIM 2021, Conférence virtuelle, 2021.
-
A LLVM tool to automatically combine floating-point with multiple-precision computations.
G. Revy.
2e face-to-face meeting du projet REPRISES, Lyon, France, 2019
Meetings
-
29-05-2019: Kick-off meeting, Perpignan, France.
Participants: L. Arrabito, J. Bregeon, Ph. Langlois, D. Parello, G. Revy.