This directory contains the "Formal Proofs on Floating Point"
library. It enables users to formally define types and operations of
floating arithmetic and to deduce properties using a mechanized proof
checker.

This library was first designed by Marc Daumas, Laurent Thery* and
Laurence Rideau*. Contributors now include Sylvie Boldo*, Jean-Michel
Muller, Laurent Fousse*, Ren-Cang Li and Guillaume
Melquiond*. Contributors that were directly interacting with Coq are
outlined by a star.

Latest developments include:

- Proved algorithms for argument reduction

- Software emulation of the FMA using rounding to odd

- A few properties from Prof. W. Kahan

- Results on the ulp of the inverse of a floating-point number from
  Dr. P. Markstein

- A link to the IEEE-754 standard

- Radix-independent algorithm to compute the exact error of a
  multiplication

- Error of a floating-point division or square root

- Conditions for a*x+y to be faithfully computed

Sources are in the "source" directory and are to be compiled with
version 8.1 of Coq.

http://coq.inria.fr/

Detailed documentations are in the "doc" directory and start with the
"index.html" file.
