1 [section:next_float Floating-Point Representation Distance (ULP),
2 and Finding Adjacent Floating-Point Values]
4 [@http://en.wikipedia.org/wiki/Unit_in_the_last_place Unit of Least Precision or Unit in the Last Place]
5 is the gap between two different, but as close as possible, floating-point numbers.
7 Most decimal values, for example 0.1, cannot be exactly represented as floating-point values,
8 but will be stored as the
9 [@http://en.wikipedia.org/wiki/Floating_point#Representable_numbers.2C_conversion_and_rounding
10 closest representable floating-point].
12 Functions are provided for finding adjacent greater and lesser floating-point values,
13 and estimating the number of gaps between any two floating-point values.
15 The floating-point type FPT must have has a fixed number of bits in the representation.
16 The number of bits may set at runtime, but must be the same for all numbers.
17 For example, __NTL_quad_float type (fixed 128-bit representation)
18 or __NTL_RR type (arbitrary but fixed decimal digits, default 150)
19 but *not* a type that extends the representation to provide an exact representation
20 for any number, for example [@http://keithbriggs.info/xrc.html XRC eXact Real in C].
22 [section:nextafter Finding the Next Representable Value in a Specific Direction (nextafter)]
27 #include <boost/math/special_functions/next.hpp>
30 namespace boost{ namespace math{
33 FPT nextafter(FPT val, FPT direction);
37 [h4 Description - nextafter]
39 This is an implementation of the `nextafter` function included in the C99 standard.
40 (It is also effectively an implementation of the C99 'nexttoward' legacy function
41 which differs only having a long double direction,
42 and can generally serve in its place if required).
44 [note The C99 functions must use suffixes f and l to distinguish float and long double versions.
45 C++ uses the template mechanism instead.]
47 Returns the next representable value after /x/ in the direction of /y/. If
48 `x == y` then returns /x/. If /x/ is non-finite then returns the result of
49 a __domain_error. If there is no such value in the direction of /y/ then
50 returns an __overflow_error.
52 [warning The template parameter FTP must be a floating-point type.
53 An integer type, for example, will produce an unhelpful error message.]
55 [tip Nearly always, you just want the next or prior representable value,
56 so instead use `float_next` or `float_prior` below.]
58 [h4 Examples - nextafter]
60 The two representations using a 32-bit float either side of unity are:
62 The nearest (exact) representation of 1.F is 1.00000000
63 nextafter(1.F, 999) is 1.00000012
64 nextafter(1/f, -999) is 0.99999994
66 The nearest (not exact) representation of 0.1F is 0.100000001
67 nextafter(0.1F, 10) is 0.100000009
68 nextafter(0.1F, 10) is 0.099999994
73 [section:float_next Finding the Next Greater Representable Value (float_next)]
78 #include <boost/math/special_functions/next.hpp>
81 namespace boost{ namespace math{
84 FPT float_next(FPT val);
88 [h4 Description - float_next]
90 Returns the next representable value which is greater than /x/.
91 If /x/ is non-finite then returns the result of
92 a __domain_error. If there is no such value greater than /x/ then
93 returns an __overflow_error.
95 Has the same effect as
97 nextafter(val, (std::numeric_limits<FPT>::max)());
99 [endsect] [/section:float_next Finding the Next Greater Representable Value (float_prior)]
101 [section:float_prior Finding the Next Smaller Representable Value (float_prior)]
106 #include <boost/math/special_functions/next.hpp>
109 namespace boost{ namespace math{
112 FPT float_prior(FPT val);
117 [h4 Description - float_prior]
119 Returns the next representable value which is less than /x/.
120 If /x/ is non-finite then returns the result of
121 a __domain_error. If there is no such value less than /x/ then
122 returns an __overflow_error.
124 Has the same effect as
126 nextafter(val, -(std::numeric_limits<FPT>::max)()); // Note most negative value -max.
128 [endsect] [/section:float_prior Finding the Next Smaller Representable Value (float_prior)]
130 [section:float_distance Calculating the Representation Distance
131 Between Two Floating Point Values (ULP) float_distance]
133 Function float_distance finds the number of gaps/bits/ULP between any two floating-point values.
134 If the significands of floating-point numbers are viewed as integers,
135 then their difference is the number of ULP/gaps/bits different.
140 #include <boost/math/special_functions/next.hpp>
143 namespace boost{ namespace math{
146 FPT float_distance(FPT a, FPT b);
150 [h4 Description - float_distance]
152 Returns the distance between /a/ and /b/: the result is always
153 a signed integer value (stored in floating-point type FPT)
154 representing the number of distinct representations between /a/ and /b/.
158 * `float_distance(a, a)` always returns 0.
159 * `float_distance(float_next(a), a)` always returns -1.
160 * `float_distance(float_prior(a), a)` always returns 1.
162 The function `float_distance` is equivalent to calculating the number
163 of ULP (Units in the Last Place) between /a/ and /b/ except that it
164 returns a signed value indicating whether `a > b` or not.
166 If the distance is too great then it may not be able
167 to be represented as an exact integer by type FPT,
168 but in practice this is unlikely to be a issue.
170 [endsect] [/section:float_distance Calculating the Representation Distance
171 Between Two Floating Point Values (ULP) float_distance]
173 [section:float_advance Advancing a Floating Point Value by a Specific
174 Representation Distance (ULP) float_advance]
176 Function float_advance advances a floating point number by a specified number
182 #include <boost/math/special_functions/next.hpp>
185 namespace boost{ namespace math{
188 FPT float_advance(FPT val, int distance);
192 [h4 Description - float_advance]
194 Returns a floating point number /r/ such that `float_distance(val, r) == distance`.
196 [endsect] [/section:float_advance]
198 [endsect] [/ section:next_float Floating-Point Representation Distance (ULP),
199 and Finding Adjacent Floating-Point Values]
202 Copyright 2008 John Maddock and Paul A. Bristow.
203 Distributed under the Boost Software License, Version 1.0.
204 (See accompanying file LICENSE_1_0.txt or copy at
205 http://www.boost.org/LICENSE_1_0.txt).