1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
|
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<title>Floating-point Comparison</title>
<link rel="stylesheet" href="../math.css" type="text/css">
<meta name="generator" content="DocBook XSL Stylesheets V1.79.1">
<link rel="home" href="../index.html" title="Math Toolkit 4.2.1">
<link rel="up" href="../utils.html" title="Chapter 2. Floating Point Utilities">
<link rel="prev" href="next_float/ulp.html" title="Obtaining the Size of a Unit In the Last Place - ULP">
<link rel="next" href="cond.html" title="Condition Numbers">
<meta name="viewport" content="width=device-width, initial-scale=1">
</head>
<body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF">
<table cellpadding="2" width="100%"><tr>
<td valign="top"><img alt="Boost C++ Libraries" width="277" height="86" src="../../../../../boost.png"></td>
<td align="center"><a href="../../../../../index.html">Home</a></td>
<td align="center"><a href="../../../../../libs/libraries.htm">Libraries</a></td>
<td align="center"><a href="http://www.boost.org/users/people.html">People</a></td>
<td align="center"><a href="http://www.boost.org/users/faq.html">FAQ</a></td>
<td align="center"><a href="../../../../../more/index.htm">More</a></td>
</tr></table>
<hr>
<div class="spirit-nav">
<a accesskey="p" href="next_float/ulp.html"><img src="../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../utils.html"><img src="../../../../../doc/src/images/up.png" alt="Up"></a><a accesskey="h" href="../index.html"><img src="../../../../../doc/src/images/home.png" alt="Home"></a><a accesskey="n" href="cond.html"><img src="../../../../../doc/src/images/next.png" alt="Next"></a>
</div>
<div class="section">
<div class="titlepage"><div><div><h2 class="title" style="clear: both">
<a name="math_toolkit.float_comparison"></a><a class="link" href="float_comparison.html" title="Floating-point Comparison">Floating-point Comparison</a>
</h2></div></div></div>
<p>
Comparison of floating-point values has always been a source of endless difficulty
and confusion.
</p>
<p>
Unlike integral values that are exact, all floating-point operations will potentially
produce an inexact result that will be rounded to the nearest available binary
representation. Even apparently innocuous operations such as assigning 0.1
to a double produces an inexact result (as this decimal number has no exact
binary representation).
</p>
<p>
Floating-point computations also involve rounding so that some 'computational
noise' is added, and hence results are also not exact (although repeatable,
at least under identical platforms and compile options).
</p>
<p>
Sadly, this conflicts with the expectation of most users, as many articles
and innumerable cries for help show all too well.
</p>
<p>
Some background reading is:
</p>
<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
<li class="listitem">
Knuth D.E. The art of computer programming, vol II, section 4.2, especially
Floating-Point Comparison 4.2.2, pages 198-220.
</li>
<li class="listitem">
<a href="http://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html" target="_top">David
Goldberg, "What Every Computer Scientist Should Know About Floating-Point
Arithmetic"</a>
</li>
<li class="listitem">
<a href="http://adtmag.com/articles/2000/03/16/comparing-floatshow-to-determine-if-floating-quantities-are-close-enough-once-a-tolerance-has-been-r.aspx" target="_top">Alberto
Squassabia, Comparing floats listing</a>
</li>
<li class="listitem">
<a href="https://code.google.com/p/googletest/wiki/AdvancedGuide#Floating-Point_Comparison" target="_top">Google
Floating-Point_Comparison guide</a>
</li>
<li class="listitem">
<a href="https://www.boost.org/doc/libs/release/libs/test/doc/html/boost_test/testing_tools/extended_comparison/floating_point.html" target="_top">Boost.Test
Floating-Point_Comparison</a>
</li>
</ul></div>
<p>
Boost provides a number of ways to compare floating-point values to see if
they are tolerably close enough to each other, but first we must decide what
kind of comparison we require:
</p>
<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
<li class="listitem">
Absolute difference/error: the absolute difference between two values
<span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span> is simply <code class="computeroutput"><span class="identifier">fabs</span><span class="special">(</span><span class="identifier">a</span><span class="special">-</span><span class="identifier">b</span><span class="special">)</span></code>. This is the only meaningful comparison
to make if we know that the result may have cancellation error (see below).
</li>
<li class="listitem">
The edit distance between the two values: i.e. how many (binary) floating-point
values are between two values <span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span>?
This is provided by the function <a href="../../../../../libs/math/doc/html/math_toolkit/next_float/float_distance.html" target="_top">Boost.Math
float_distance</a>, but is probably only useful when you know that
the distance should be very small. This function is somewhat difficult
to compute, and doesn't scale to values that are very far apart. In other
words, use with care.
</li>
<li class="listitem">
The relative distance/error between two values. This is quick and easy
to compute, and is generally the method of choice when checking that your
results are "tolerably close" to one another. However, it is
not as exact as the edit distance when dealing with small differences,
and due to the way floating-point values are encoded can "wobble"
by a factor of 2 compared to the "true" edit distance. This is
the method documented below: if <code class="computeroutput"><span class="identifier">float_distance</span></code>
is a surgeon's scalpel, then <code class="computeroutput"><span class="identifier">relative_difference</span></code>
is more like a Swiss army knife: both have important but different use
cases.
</li>
</ul></div>
<h6>
<a name="math_toolkit.float_comparison.h0"></a>
<span class="phrase"><a name="math_toolkit.float_comparison.fp_relative"></a></span><a class="link" href="float_comparison.html#math_toolkit.float_comparison.fp_relative">Relative
Comparison of Floating-point Values</a>
</h6>
<p>
<code class="computeroutput"><span class="preprocessor">#include</span> <span class="special"><</span><span class="identifier">boost</span><span class="special">/</span><span class="identifier">math</span><span class="special">/</span><span class="identifier">special_functions</span><span class="special">/</span><span class="identifier">relative_difference</span><span class="special">.</span><span class="identifier">hpp</span><span class="special">></span></code>
</p>
<pre class="programlisting"><span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">,</span> <span class="keyword">class</span> <span class="identifier">U</span><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">a</span><span class="special">,</span> <span class="identifier">U</span> <span class="identifier">b</span><span class="special">);</span>
<span class="keyword">template</span> <span class="special"><</span><span class="keyword">class</span> <span class="identifier">T</span><span class="special">,</span> <span class="keyword">class</span> <span class="identifier">U</span><span class="special">></span>
<a class="link" href="result_type.html" title="Calculation of the Type of the Result"><span class="emphasis"><em>calculated-result-type</em></span></a> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">T</span> <span class="identifier">a</span><span class="special">,</span> <span class="identifier">U</span> <span class="identifier">b</span><span class="special">);</span>
</pre>
<p>
The function <code class="computeroutput"><span class="identifier">relative_difference</span></code>
returns the relative distance/error <span class="emphasis"><em>E</em></span> between two values
as defined by:
</p>
<div class="blockquote"><blockquote class="blockquote"><p>
<span class="serif_italic">E = fabs((a - b) / min(a,b))</span>
</p></blockquote></div>
<p>
The function <code class="computeroutput"><span class="identifier">epsilon_difference</span></code>
is a convenience function that returns <code class="computeroutput"><span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span>
<span class="identifier">b</span><span class="special">)</span> <span class="special">/</span> <span class="identifier">eps</span></code> where
<code class="computeroutput"><span class="identifier">eps</span></code> is the machine epsilon
for the result type.
</p>
<p>
The following special cases are handled as follows:
</p>
<div class="itemizedlist"><ul class="itemizedlist" style="list-style-type: disc; ">
<li class="listitem">
If either of <span class="emphasis"><em>a</em></span> or <span class="emphasis"><em>b</em></span> is a NaN,
then returns the largest representable value for T: for example for type
<code class="computeroutput"><span class="keyword">double</span></code>, this is <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">double</span><span class="special">>::</span><span class="identifier">max</span><span class="special">()</span></code>
which is the same as <code class="computeroutput"><span class="identifier">DBL_MAX</span></code>
or <code class="computeroutput"><span class="number">1.7976931348623157e+308</span></code>.
</li>
<li class="listitem">
If <span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span> differ in sign then
returns the largest representable value for T.
</li>
<li class="listitem">
If both <span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span> are both infinities
(of the same sign), then returns zero.
</li>
<li class="listitem">
If just one of <span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span> is an
infinity, then returns the largest representable value for T.
</li>
<li class="listitem">
If both <span class="emphasis"><em>a</em></span> and <span class="emphasis"><em>b</em></span> are zero then
returns zero.
</li>
<li class="listitem">
If just one of <span class="emphasis"><em>a</em></span> or <span class="emphasis"><em>b</em></span> is a zero
or a denormalized value, then it is treated as if it were the smallest
(non-denormalized) value representable in T for the purposes of the above
calculation.
</li>
</ul></div>
<p>
These rules were primarily designed to assist with our own test suite, they
are designed to be robust enough that the function can in most cases be used
blindly, including in cases where the expected result is actually too small
to represent in type T and underflows to zero.
</p>
<h6>
<a name="math_toolkit.float_comparison.h1"></a>
<span class="phrase"><a name="math_toolkit.float_comparison.examples"></a></span><a class="link" href="float_comparison.html#math_toolkit.float_comparison.examples">Examples</a>
</h6>
<p>
Some using statements will ensure that the functions we need are accessible.
</p>
<pre class="programlisting"><span class="keyword">using</span> <span class="keyword">namespace</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">;</span>
</pre>
<p>
or
</p>
<pre class="programlisting"><span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">relative_difference</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">epsilon_difference</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">float_next</span><span class="special">;</span>
<span class="keyword">using</span> <span class="identifier">boost</span><span class="special">::</span><span class="identifier">math</span><span class="special">::</span><span class="identifier">float_prior</span><span class="special">;</span>
</pre>
<p>
The following examples display values with all possibly significant digits.
Newer compilers should provide <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">FPT</span><span class="special">>::</span><span class="identifier">max_digits10</span></code> for this purpose, and here we
use <code class="computeroutput"><span class="keyword">float</span></code> precision where <code class="computeroutput"><span class="identifier">max_digits10</span></code> = 9 to avoid displaying a distracting
number of decimal digits.
</p>
<div class="note"><table border="0" summary="Note">
<tr>
<td rowspan="2" align="center" valign="top" width="25"><img alt="[Note]" src="../../../../../doc/src/images/note.png"></td>
<th align="left">Note</th>
</tr>
<tr><td align="left" valign="top"><p>
Older compilers can use this formula to calculate <code class="computeroutput"><span class="identifier">max_digits10</span></code>
from <code class="computeroutput"><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">FPT</span><span class="special">>::</span><span class="identifier">digits10</span></code>: <code class="computeroutput"><span class="keyword">int</span>
<span class="identifier">max_digits10</span> <span class="special">=</span>
<span class="number">2</span> <span class="special">+</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="identifier">FPT</span><span class="special">>::</span><span class="identifier">digits10</span>
<span class="special">*</span> <span class="number">3010</span><span class="special">/</span><span class="number">10000</span><span class="special">;</span></code>
</p></td></tr>
</table></div>
<p>
One can set the display including all trailing zeros (helpful for this example
to show all potentially significant digits), and also to display <code class="computeroutput"><span class="keyword">bool</span></code> values as words rather than integers:
</p>
<pre class="programlisting"><span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span><span class="special">.</span><span class="identifier">precision</span><span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">float</span><span class="special">>::</span><span class="identifier">max_digits10</span><span class="special">);</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">boolalpha</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">showpoint</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
When comparing values that are <span class="emphasis"><em>quite close</em></span> or <span class="emphasis"><em>approximately
equal</em></span>, we could use either <code class="computeroutput"><span class="identifier">float_distance</span></code>
or <code class="computeroutput"><span class="identifier">relative_difference</span></code>/<code class="computeroutput"><span class="identifier">epsilon_difference</span></code>, for example with type
<code class="computeroutput"><span class="keyword">float</span></code>, these two values are adjacent
to each other:
</p>
<pre class="programlisting"><span class="keyword">float</span> <span class="identifier">a</span> <span class="special">=</span> <span class="number">1</span><span class="special">;</span>
<span class="keyword">float</span> <span class="identifier">b</span> <span class="special">=</span> <span class="number">1</span> <span class="special">+</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">float</span><span class="special">>::</span><span class="identifier">epsilon</span><span class="special">();</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"a = "</span> <span class="special"><<</span> <span class="identifier">a</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"b = "</span> <span class="special"><<</span> <span class="identifier">b</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"float_distance = "</span> <span class="special"><<</span> <span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon_difference = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
Which produces the output:
</p>
<pre class="programlisting">a = 1.00000000
b = 1.00000012
float_distance = 1.00000000
relative_difference = 1.19209290e-007
epsilon_difference = 1.00000000
</pre>
<p>
In the example above, it just so happens that the edit distance as measured
by <code class="computeroutput"><span class="identifier">float_distance</span></code>, and the
difference measured in units of epsilon were equal. However, due to the way
floating point values are represented, that is not always the case:
</p>
<pre class="programlisting"><span class="identifier">a</span> <span class="special">=</span> <span class="number">2.0f</span> <span class="special">/</span> <span class="number">3.0f</span><span class="special">;</span> <span class="comment">// 2/3 inexactly represented as a float</span>
<span class="identifier">b</span> <span class="special">=</span> <span class="identifier">float_next</span><span class="special">(</span><span class="identifier">float_next</span><span class="special">(</span><span class="identifier">float_next</span><span class="special">(</span><span class="identifier">a</span><span class="special">)));</span> <span class="comment">// 3 floating point values above a</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"a = "</span> <span class="special"><<</span> <span class="identifier">a</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"b = "</span> <span class="special"><<</span> <span class="identifier">b</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"float_distance = "</span> <span class="special"><<</span> <span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon_difference = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
Which produces the output:
</p>
<pre class="programlisting">a = 0.666666687
b = 0.666666865
float_distance = 3.00000000
relative_difference = 2.68220901e-007
epsilon_difference = 2.25000000
</pre>
<p>
There is another important difference between <code class="computeroutput"><span class="identifier">float_distance</span></code>
and the <code class="computeroutput"><span class="identifier">relative_difference</span><span class="special">/</span><span class="identifier">epsilon_difference</span></code>
functions in that <code class="computeroutput"><span class="identifier">float_distance</span></code>
returns a signed result that reflects which argument is larger in magnitude,
where as <code class="computeroutput"><span class="identifier">relative_difference</span><span class="special">/</span><span class="identifier">epsilon_difference</span></code>
simply return an unsigned value that represents how far apart the values are.
For example if we swap the order of the arguments:
</p>
<pre class="programlisting"><span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"float_distance = "</span> <span class="special"><<</span> <span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">b</span><span class="special">,</span> <span class="identifier">a</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">b</span><span class="special">,</span> <span class="identifier">a</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon_difference = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">b</span><span class="special">,</span> <span class="identifier">a</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
The output is now:
</p>
<pre class="programlisting">float_distance = -3.00000000
relative_difference = 2.68220901e-007
epsilon_difference = 2.25000000
</pre>
<p>
Zeros are always treated as equal, as are infinities as long as they have the
same sign:
</p>
<pre class="programlisting"><span class="identifier">a</span> <span class="special">=</span> <span class="number">0</span><span class="special">;</span>
<span class="identifier">b</span> <span class="special">=</span> <span class="special">-</span><span class="number">0</span><span class="special">;</span> <span class="comment">// signed zero</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">a</span> <span class="special">=</span> <span class="identifier">b</span> <span class="special">=</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">float</span><span class="special">>::</span><span class="identifier">infinity</span><span class="special">();</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="special">-</span><span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
Which produces the output:
</p>
<pre class="programlisting">relative_difference = 0.000000000
relative_difference = 0.000000000
relative_difference = 3.40282347e+038
</pre>
<p>
Note that finite values are always infinitely far away from infinities even
if those finite values are very large:
</p>
<pre class="programlisting"><span class="identifier">a</span> <span class="special">=</span> <span class="special">(</span><span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">float</span><span class="special">>::</span><span class="identifier">max</span><span class="special">)();</span>
<span class="identifier">b</span> <span class="special">=</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">float</span><span class="special">>::</span><span class="identifier">infinity</span><span class="special">();</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"a = "</span> <span class="special"><<</span> <span class="identifier">a</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"b = "</span> <span class="special"><<</span> <span class="identifier">b</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon_difference = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
Which produces the output:
</p>
<pre class="programlisting">a = 3.40282347e+038
b = 1.#INF0000
relative_difference = 3.40282347e+038
epsilon_difference = 3.40282347e+038
</pre>
<p>
Finally, all denormalized values and zeros are treated as being effectively
equal:
</p>
<pre class="programlisting"><span class="identifier">a</span> <span class="special">=</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">numeric_limits</span><span class="special"><</span><span class="keyword">float</span><span class="special">>::</span><span class="identifier">denorm_min</span><span class="special">();</span>
<span class="identifier">b</span> <span class="special">=</span> <span class="identifier">a</span> <span class="special">*</span> <span class="number">2</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"a = "</span> <span class="special"><<</span> <span class="identifier">a</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"b = "</span> <span class="special"><<</span> <span class="identifier">b</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"float_distance = "</span> <span class="special"><<</span> <span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon_difference = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">a</span> <span class="special">=</span> <span class="number">0</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"a = "</span> <span class="special"><<</span> <span class="identifier">a</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"b = "</span> <span class="special"><<</span> <span class="identifier">b</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"float_distance = "</span> <span class="special"><<</span> <span class="identifier">float_distance</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"relative_difference = "</span> <span class="special"><<</span> <span class="identifier">relative_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
<span class="identifier">std</span><span class="special">::</span><span class="identifier">cout</span> <span class="special"><<</span> <span class="string">"epsilon_difference = "</span> <span class="special"><<</span> <span class="identifier">epsilon_difference</span><span class="special">(</span><span class="identifier">a</span><span class="special">,</span> <span class="identifier">b</span><span class="special">)</span> <span class="special"><<</span> <span class="identifier">std</span><span class="special">::</span><span class="identifier">endl</span><span class="special">;</span>
</pre>
<p>
Which produces the output:
</p>
<pre class="programlisting">a = 1.40129846e-045
b = 2.80259693e-045
float_distance = 1.00000000
relative_difference = 0.000000000
epsilon_difference = 0.000000000
a = 0.000000000
b = 2.80259693e-045
float_distance = 2.00000000
relative_difference = 0.000000000
epsilon_difference = 0.000000000</pre>
<p>
Notice how, in the above example, two denormalized values that are a factor
of 2 apart are none the less only one representation apart!
</p>
<p>
All the above examples are contained in <a href="../../../example/float_comparison_example.cpp" target="_top">float_comparison_example.cpp</a>.
</p>
<h6>
<a name="math_toolkit.float_comparison.h2"></a>
<span class="phrase"><a name="math_toolkit.float_comparison.small"></a></span><a class="link" href="float_comparison.html#math_toolkit.float_comparison.small">Handling
Absolute Errors</a>
</h6>
<p>
Imagine we're testing the following function:
</p>
<pre class="programlisting"><span class="keyword">double</span> <span class="identifier">myspecial</span><span class="special">(</span><span class="keyword">double</span> <span class="identifier">x</span><span class="special">)</span>
<span class="special">{</span>
<span class="keyword">return</span> <span class="identifier">sin</span><span class="special">(</span><span class="identifier">x</span><span class="special">)</span> <span class="special">-</span> <span class="identifier">sin</span><span class="special">(</span><span class="number">4</span> <span class="special">*</span> <span class="identifier">x</span><span class="special">);</span>
<span class="special">}</span>
</pre>
<p>
This function has multiple roots, some of which are quite predictable in that
both <code class="computeroutput"><span class="identifier">sin</span><span class="special">(</span><span class="identifier">x</span><span class="special">)</span></code> and <code class="computeroutput"><span class="identifier">sin</span><span class="special">(</span><span class="number">4</span><span class="identifier">x</span><span class="special">)</span></code> are zero
together. Others occur because the values returned from those two functions
precisely cancel out. At such points the relative difference between the true
value of the function and the actual value returned may be <span class="emphasis"><em>arbitrarily
large</em></span> due to <a href="http://en.wikipedia.org/wiki/Loss_of_significance" target="_top">cancellation
error</a>.
</p>
<p>
In such a case, testing the function above by requiring that the values returned
by <code class="computeroutput"><span class="identifier">relative_error</span></code> or <code class="computeroutput"><span class="identifier">epsilon_error</span></code> are below some threshold is
pointless: the best we can do is to verify that the <span class="emphasis"><em>absolute difference</em></span>
between the true and calculated values is below some threshold.
</p>
<p>
Of course, determining what that threshold should be is often tricky, but a
good starting point would be machine epsilon multiplied by the largest of the
values being summed. In the example above, the largest value returned by <code class="computeroutput"><span class="identifier">sin</span><span class="special">(</span><span class="identifier">whatever</span><span class="special">)</span></code> is 1, so simply using machine epsilon as the
target for maximum absolute difference might be a good start (though in practice
we may need a slightly higher value - some trial and error will be necessary).
</p>
</div>
<div class="copyright-footer">Copyright © 2006-2021 Nikhar Agrawal, Anton Bikineev, Matthew Borland,
Paul A. Bristow, Marco Guazzone, Christopher Kormanyos, Hubert Holin, Bruno
Lalande, John Maddock, Evan Miller, Jeremy Murphy, Matthew Pulver, Johan Råde,
Gautam Sewani, Benjamin Sobotta, Nicholas Thompson, Thijs van den Berg, Daryle
Walker and Xiaogang Zhang<p>
Distributed under the Boost Software License, Version 1.0. (See accompanying
file LICENSE_1_0.txt or copy at <a href="http://www.boost.org/LICENSE_1_0.txt" target="_top">http://www.boost.org/LICENSE_1_0.txt</a>)
</p>
</div>
<hr>
<div class="spirit-nav">
<a accesskey="p" href="next_float/ulp.html"><img src="../../../../../doc/src/images/prev.png" alt="Prev"></a><a accesskey="u" href="../utils.html"><img src="../../../../../doc/src/images/up.png" alt="Up"></a><a accesskey="h" href="../index.html"><img src="../../../../../doc/src/images/home.png" alt="Home"></a><a accesskey="n" href="cond.html"><img src="../../../../../doc/src/images/next.png" alt="Next"></a>
</div>
</body>
</html>
|
