Common — library colibrilib.fbin

Binary floating–point numbers, the format as data

The floats of the binary formats of IEEE 754, without ieee_float: a format is a pair of integers {eb; sb}, a float is NaN, an infinity or a sign and a magnitude. The value of the magnitude k is vom fm k, for every k >= 0 (the exponent is not bounded).

The rounding of a real r >= 0 is defined by its neighbours (IEEE 754 §4.3): the magnitudes k and k+1 with vom k <= r < vom (k+1). The overflow (§7.4) gives the infinity or the largest finite float. round_d computes it on the dyadic numbers m * 2^e, by shifting.

module Pow2r
module Format
module Float
module Round
module Impl
module Nearest
module BitsIv
module Special
module Integral
module AddBound
module Exact
module FpBase
module Fp
module Export

Statistics

Proved (2075) (colibri2 1698) (cvc4 70) (cvc5 307) (7min) (9%)