The B method takes up floating-point numbers

Lilian Burdy (1) , Jean-Louis Dufour (2) , Thierry Lecomte (3)

1 ClearSy Systems Engineering
2 SAGEM Défense Sécurité [Massy]
3 GICC UMR 7292 CNRS - Génétique, immunothérapie, chimie et cancer (GICC), UMR 7292 CNRS [2012-2017]

For a long time, formal methods have ignored floating-point computations. About ten years ago this has changed, and today specification languages and tools are in use in research and pre-industrial contexts. Better late than never: the B method, which has been the first formal method to prove real-size software, will soon be able to prove the correctness of floating-point computations. This paper gives the motivations, the philosophy and the first impacts on the AtelierB tool.

Mots clés: formal methods; B method; floating-point arithmetic; DO-178

Domaines: Informatique [cs]; Systèmes embarqués

Lilian Burdy, Jean-Louis Dufour, Thierry Lecomte. The B method takes up floating-point numbers. Embedded Real Time Software and Systems (ERTS2012), Feb 2012, Toulouse, France. ⟨hal-02263404⟩