Platform to analyse Floating-Point arithmetic – INTERFLOP
Intensive usage of floating-point (FP) arithmetic in today’s real software affect numerical quality and reproducibility of program. The problem of the detection, localization and correction of numerical bug is entering a new area with the tremendous increases in computational horsepower needed to address new classes of problems, combined with the emergence of new multicore processors and new application-specific floating-point formats. InterFLOP project aims at providing a modular and scalable platform to both analyze and control the costs of FP behavior of today’s real programs facing those new paradigms. The platform will propose interoperable tools, starting from existing one developed by the partners combined with new composite one, to analyze and optimize FP calculus. By making those tools interoperable, it will be possible to take advantage of the properties and information specific to each of them in order to build composite and new analyses, inaccessible otherwise.
Project coordination
DAVID DEFOUR (LABORATOIRE DE MATHEMATIQUES, PHYSIQUE ET SYSTEMES)
The author of this summary is the project coordinator, who is responsible for the content of this summary. The ANR declines any responsibility as for its contents.
Partner
ANEO ANEO
LAMPS LABORATOIRE DE MATHEMATIQUES, PHYSIQUE ET SYSTEMES
Intel INTEL Corporation SAS / Data Center Group – Entreprise & Gouvernement
LI-PaRAD Laboratoire d'informatique Parallélisme, Réseaux et Algorithmique Distribuée
CEA LIST Commissariat à l'énergie atomique et aux énergies alternatives (CEA) / Laboratoire d'Intégration des Systèmes et des Technologies
TRISCALE INNOV
LIP6 Laboratoire d'informatique de Paris 6
EDF SA ELECTRICITE DE FRANCE
Help of the ANR 642,878 euros
Beginning and duration of the scientific project:
- 48 Months