Frama-C:
Plug-ins:
Libraries:

Frama-C API - Fc_float

Implementation of floating-point values of different precision, using the standard ocaml floating-point numbers in double precision. Long_Double and Real are inexact.

include Float_sig.S with type t = float
type t = float
val packed_descr : Structural_descr.pack
val hash : t -> int
val pretty : Stdlib.Format.formatter -> t -> unit
val is_exact : Float_sig.prec -> bool

Is the representation of floating-point numbers of a given precision exact? Should at least be false for Real.

val of_float : Float_sig.round -> Float_sig.prec -> float -> t

Converts a caml float into a floating-point number of the given precision. Must work on infinities and on negative and positive zeros.

val to_float : t -> float

Converts a floating-point number of single or double precision into a caml float. Never called on better precision than double.

val cmp_ieee : t -> t -> int

IEEE comparison: do not distinguish between -0.0 and 0.0. This comparison must work between values of different precisions.

val compare : t -> t -> int

Comparison that distinguishes -0.0 and 0.0. This comparison must work between values of different precisions.

val is_nan : t -> bool
val is_finite : t -> bool
val is_infinite : t -> bool
val is_negative : t -> bool

Is a number negative? Never called on NaN, but must be correct on infinities and zeros.

val round_to_precision : Float_sig.round -> Float_sig.prec -> t -> t

Rounds a number to a given precision.

val next_float : Float_sig.prec -> t -> t

First value above the argument in the given precision. Returns the minimum finite value on -infty. Returns +0 on -0, and the minimum strictly positive value on +0. Returns +infty on the maximum finite value. Returns +infty on +infty. Undefined on NaN.

val prev_float : Float_sig.prec -> t -> t

First value below the argument in the given precision. Inversed behavior as next_float.

val widen_up : ?hint:Datatype.Float.Set.t -> Float_sig.prec -> t -> t

widen_up f returns a value strictly larger than f, such that successive applications of widen_up converge rapidly to infinity. The first optional argument gives a set of relevant thresholds that could be used.

val widen_down : ?hint:Datatype.Float.Set.t -> Float_sig.prec -> t -> t

Floating-point operations.

val neg : t -> t
val abs : t -> t
val floor : t -> t
val ceil : t -> t
val trunc : t -> t
val fround : t -> t
val add : Float_sig.round -> Float_sig.prec -> t -> t -> t
val sub : Float_sig.round -> Float_sig.prec -> t -> t -> t
val mul : Float_sig.round -> Float_sig.prec -> t -> t -> t
val div : Float_sig.round -> Float_sig.prec -> t -> t -> t
val fmod : Float_sig.round -> Float_sig.prec -> t -> t -> t
val exp : Float_sig.round -> Float_sig.prec -> t -> t
val log : Float_sig.round -> Float_sig.prec -> t -> t
val log10 : Float_sig.round -> Float_sig.prec -> t -> t
val sqrt : Float_sig.round -> Float_sig.prec -> t -> t
val pow : Float_sig.round -> Float_sig.prec -> t -> t -> t
val cos : Float_sig.round -> Float_sig.prec -> t -> t
val sin : Float_sig.round -> Float_sig.prec -> t -> t
val acos : Float_sig.round -> Float_sig.prec -> t -> t
val asin : Float_sig.round -> Float_sig.prec -> t -> t
val atan : Float_sig.round -> Float_sig.prec -> t -> t
val atan2 : Float_sig.round -> Float_sig.prec -> t -> t -> t