Debian Codegraph @

upstream/debian-unstable/man3/fma.3

raw
1 '\" t
2 .\" Copyright 2002, Walter Harms <[email protected]>
3 .\" Copyright, the authors of the Linux man-pages project
4 .\"
5 .\" SPDX-License-Identifier: GPL-1.0-or-later
6 .\"
7 .TH fma 3 2026-02-08 "Linux man-pages 6.18"
8 .SH NAME
9 fma, fmaf, fmal \- floating-point multiply and add
10 .SH LIBRARY
11 Math library
12 .RI ( libm ,\~ \-lm )
13 .SH SYNOPSIS
14 .nf
15 .B #include <math.h>
16 .P
17 .BI "double fma(double " x ", double " y ", double " z );
18 .BI "float fmaf(float " x ", float " y ", float " z );
19 .BI "long double fmal(long double " x ", long double " y ", long double " z );
20 .fi
21 .P
22 .RS -4
23 Feature Test Macro Requirements for glibc (see
24 .BR feature_test_macros (7)):
25 .RE
26 .P
27 .BR fma (),
28 .BR fmaf (),
29 .BR fmal ():
30 .nf
31 _ISOC99_SOURCE || _POSIX_C_SOURCE >= 200112L
32 .fi
33 .SH DESCRIPTION
34 These functions compute
35 .IR x " * " y " + " z .
36 The result is rounded as one ternary operation according to the
37 current rounding mode (see
38 .BR fenv (3)).
39 .SH RETURN VALUE
40 These functions return the value of
41 .IR x " * " y " + " z ,
42 rounded as one ternary operation.
43 .P
44 If
45 .I x
46 or
47 .I y
48 is a NaN, a NaN is returned.
49 .P
50 If
51 .I x
52 times
53 .I y
54 is an exact infinity, and
55 .I z
56 is an infinity with the opposite sign,
57 a domain error occurs,
58 and a NaN is returned.
59 .P
60 .\" POSIX.1-2008 allows some possible differences for the following two
61 .\" domain error cases, but on Linux they are treated the same (AFAICS).
62 .\" Nevertheless, we'll mirror POSIX.1 and describe the two cases
63 .\" separately.
64 If one of
65 .I x
66 or
67 .I y
68 is an infinity, the other is 0, and
69 .I z
70 is not a NaN,
71 a domain error occurs, and
72 a NaN is returned.
73 .\" POSIX.1 says that a NaN or an implementation-defined value shall
74 .\" be returned for this case.
75 .P
76 If one of
77 .I x
78 or
79 .I y
80 is an infinity, and the other is 0, and
81 .I z
82 is a NaN,
83 .\" POSIX.1 makes the domain error optional for this case.
84 a domain error occurs, and
85 a NaN is returned.
86 .P
87 If
88 .I x
89 times
90 .I y
91 is not an infinity times zero (or vice versa), and
92 .I z
93 is a NaN,
94 a NaN is returned.
95 .P
96 If the result overflows,
97 a range error occurs, and
98 an infinity with the correct sign is returned.
99 .P
100 If the result underflows,
101 a range error occurs, and
102 a signed 0 is returned.
103 .SH ERRORS
104 See
105 .BR math_error (7)
106 for information on how to determine whether an error has occurred
107 when calling these functions.
108 .P
109 The following errors can occur:
110 .TP
111 Domain error: \f[I]x * y + z\f[] \
112 or \f[I]x * y\f[] is invalid.
113 .\" .I errno
114 .\" is set to
115 .\" .BR EDOM .
116 An invalid floating-point exception
117 .RB ( FE_INVALID )
118 is raised.
119 .TP
120 Range error: result overflow
121 .\" .I errno
122 .\" is set to
123 .\" .BR ERANGE .
124 An overflow floating-point exception
125 .RB ( FE_OVERFLOW )
126 is raised.
127 .TP
128 Range error: result underflow
129 .\" .I errno
130 .\" is set to
131 .\" .BR ERANGE .
132 An underflow floating-point exception
133 .RB ( FE_UNDERFLOW )
134 is raised.
135 .P
136 These functions do not set
137 .IR errno .
138 .\" FIXME . Is it intentional that these functions do not set errno?
139 .\" Bug raised: https://www.sourceware.org/bugzilla/show_bug.cgi?id=6801
140 .SH ATTRIBUTES
141 For an explanation of the terms used in this section, see
142 .BR attributes (7).
143 .TS
144 allbox;
145 lbx lb lb
146 l l l.
147 Interface Attribute Value
148 T{
149 .na
150 .nh
151 .BR fma (),
152 .BR fmaf (),
153 .BR fmal ()
154 T} Thread safety MT-Safe
155 .TE
156 .SH STANDARDS
157 C11, POSIX.1-2008.
158 .SH HISTORY
159 glibc 2.1.
160 C99, POSIX.1-2001.
161 .SH SEE ALSO
162 .BR remainder (3),
163 .BR remquo (3)