xref: /netbsd-src/lib/libm/man/exp.3 (revision 1ca5c1b28139779176bd5c13ad7c5f25c0bcd5f8)
1.\" Copyright (c) 1985, 1991 Regents of the University of California.
2.\" All rights reserved.
3.\"
4.\" Redistribution and use in source and binary forms, with or without
5.\" modification, are permitted provided that the following conditions
6.\" are met:
7.\" 1. Redistributions of source code must retain the above copyright
8.\"    notice, this list of conditions and the following disclaimer.
9.\" 2. Redistributions in binary form must reproduce the above copyright
10.\"    notice, this list of conditions and the following disclaimer in the
11.\"    documentation and/or other materials provided with the distribution.
12.\" 3. All advertising materials mentioning features or use of this software
13.\"    must display the following acknowledgement:
14.\"	This product includes software developed by the University of
15.\"	California, Berkeley and its contributors.
16.\" 4. Neither the name of the University nor the names of its contributors
17.\"    may be used to endorse or promote products derived from this software
18.\"    without specific prior written permission.
19.\"
20.\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21.\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22.\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23.\" ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24.\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25.\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26.\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27.\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28.\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29.\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30.\" SUCH DAMAGE.
31.\"
32.\"     from: @(#)exp.3	6.12 (Berkeley) 7/31/91
33.\"	$NetBSD: exp.3,v 1.16 1999/08/15 22:23:54 kleink Exp $
34.\"
35.Dd July 31, 1991
36.Dt EXP 3
37.Os
38.Sh NAME
39.Nm exp ,
40.Nm expf ,
41.Nm expm1 ,
42.Nm expm1f ,
43.Nm log ,
44.Nm logf ,
45.Nm log10 ,
46.Nm log10f ,
47.Nm log1p ,
48.Nm log1pf ,
49.Nm pow ,
50.Nm powf
51.Nd exponential, logarithm, power functions
52.Sh LIBRARY
53.Lb libm
54.Sh SYNOPSIS
55.Fd #include <math.h>
56.Ft double
57.Fn exp "double x"
58.Ft float
59.Fn expf "float x"
60.Ft double
61.Fn expm1 "double x"
62.Ft float
63.Fn expm1f "float x"
64.Ft double
65.Fn log "double x"
66.Ft float
67.Fn logf "float x"
68.Ft double
69.Fn log10 "double x"
70.Ft float
71.Fn log10f "float x"
72.Ft double
73.Fn log1p "double x"
74.Ft float
75.Fn log1pf "float x"
76.Ft double
77.Fn pow "double x" "double y"
78.Ft float
79.Fn powf "float x" "float y"
80.Sh DESCRIPTION
81The
82.Fn exp
83function computes the exponential value of the given argument
84.Fa x .
85.Pp
86The
87.Fn expm1
88function computes the value exp(x)\-1 accurately even for tiny argument
89.Fa x .
90.Pp
91The
92.Fn log
93function computes the value of the natural logarithm of argument
94.Fa x.
95.Pp
96The
97.Fn log10
98function computes the value of the logarithm of argument
99.Fa x
100to base 10.
101.Pp
102The
103.Fn log1p
104function computes
105the value of log(1+x) accurately even for tiny argument
106.Fa x .
107.Pp
108The
109.Fn pow
110computes the value
111of
112.Ar x
113to the exponent
114.Ar y .
115.Sh ERROR (due to Roundoff etc.)
116exp(x), log(x), expm1(x) and log1p(x) are accurate to within
117an
118.Em ulp ,
119and log10(x) to within about 2
120.Em ulps ;
121an
122.Em ulp
123is one
124.Em Unit
125in the
126.Em Last
127.Em Place .
128The error in
129.Fn pow x y
130is below about 2
131.Em ulps
132when its
133magnitude is moderate, but increases as
134.Fn pow x y
135approaches
136the over/underflow thresholds until almost as many bits could be
137lost as are occupied by the floating\-point format's exponent
138field; that is 8 bits for
139.Tn "VAX D"
140and 11 bits for IEEE 754 Double.
141No such drastic loss has been exposed by testing; the worst
142errors observed have been below 20
143.Em ulps
144for
145.Tn "VAX D" ,
146300
147.Em ulps
148for
149.Tn IEEE
150754 Double.
151Moderate values of
152.Fn pow
153are accurate enough that
154.Fn pow integer integer
155is exact until it is bigger than 2**56 on a
156.Tn VAX ,
1572**53 for
158.Tn IEEE
159754.
160.Sh RETURN VALUES
161These functions will return the appropriate computation unless an error
162occurs or an argument is out of range.
163The functions
164.Fn exp ,
165.Fn expm1
166and
167.Fn pow
168detect if the computed value will overflow,
169set the global variable
170.Va errno
171to
172.Er ERANGE
173and cause a reserved operand fault on a
174.Tn VAX
175or
176.Tn Tahoe .
177The function
178.Fn pow x y
179checks to see if
180.Fa x
181< 0 and
182.Fa y
183is not an integer, in the event this is true,
184the global variable
185.Va errno
186is set to
187.Er EDOM
188and on the
189.Tn VAX
190and
191.Tn Tahoe
192generate a reserved operand fault.
193On a
194.Tn VAX
195and
196.Tn Tahoe ,
197.Va errno
198is set to
199.Er EDOM
200and the reserved operand is returned
201by log unless
202.Fa x
203> 0, by
204.Fn log1p
205unless
206.Fa x
207> \-1.
208.Sh NOTES
209The functions exp(x)\-1 and log(1+x) are called
210expm1 and logp1 in
211.Tn BASIC
212on the Hewlett\-Packard
213.Tn HP Ns \-71B
214and
215.Tn APPLE
216Macintosh,
217.Tn EXP1
218and
219.Tn LN1
220in Pascal, exp1 and log1 in C
221on
222.Tn APPLE
223Macintoshes, where they have been provided to make
224sure financial calculations of ((1+x)**n\-1)/x, namely
225expm1(n\(**log1p(x))/x, will be accurate when x is tiny.
226They also provide accurate inverse hyperbolic functions.
227.Pp
228The function
229.Fn pow x 0
230returns x**0 = 1 for all x including x = 0,
231.if n \
232Infinity
233.if t \
234\(if
235(not found on a
236.Tn VAX ) ,
237and
238.Em NaN
239(the reserved
240operand on a
241.Tn VAX ) .  Previous implementations of pow may
242have defined x**0 to be undefined in some or all of these
243cases.  Here are reasons for returning x**0 = 1 always:
244.Bl -enum -width indent
245.It
246Any program that already tests whether x is zero (or
247infinite or \*(Na) before computing x**0 cannot care
248whether 0**0 = 1 or not. Any program that depends
249upon 0**0 to be invalid is dubious anyway since that
250expression's meaning and, if invalid, its consequences
251vary from one computer system to another.
252.It
253Some Algebra texts (e.g. Sigler's) define x**0 = 1 for
254all x, including x = 0.
255This is compatible with the convention that accepts a[0]
256as the value of polynomial
257.Bd -literal -offset indent
258p(x) = a[0]\(**x**0 + a[1]\(**x**1 + a[2]\(**x**2 +...+ a[n]\(**x**n
259.Ed
260.Pp
261at x = 0 rather than reject a[0]\(**0**0 as invalid.
262.It
263Analysts will accept 0**0 = 1 despite that x**y can
264approach anything or nothing as x and y approach 0
265independently.
266The reason for setting 0**0 = 1 anyway is this:
267.Bd -filled -offset indent
268If x(z) and y(z) are
269.Em any
270functions analytic (expandable
271in power series) in z around z = 0, and if there
272x(0) = y(0) = 0, then x(z)**y(z) \(-> 1 as z \(-> 0.
273.Ed
274.It
275If 0**0 = 1, then
276.if n \
277infinity**0 = 1/0**0 = 1 too; and
278.if t \
279\(if**0 = 1/0**0 = 1 too; and
280then \*(Na**0 = 1 too because x**0 = 1 for all finite
281and infinite x, i.e., independently of x.
282.El
283.Sh SEE ALSO
284.Xr math 3
285.Sh STANDARDS
286The
287.Fn exp ,
288.Fn log ,
289.Fn log10
290and
291.Fn pow
292functions conform to
293.St -ansiC .
294.Sh HISTORY
295A
296.Fn exp ,
297.Fn log
298and
299.Fn pow
300functions
301appeared in
302.At v6 .
303A
304.Fn log10
305function
306appeared in
307.At v7 .
308The
309.Fn log1p
310and
311.Fn expm1
312functions appeared in
313.Bx 4.3 .
314