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
|
'\" t
.\" Copyright 2002 Walter Harms (walter.harms@informatik.uni-oldenburg.de)
.\" and Copyright 2008, Linux Foundation, written by Michael Kerrisk
.\" <mtk.manpages@gmail.com>
.\"
.\" SPDX-License-Identifier: GPL-1.0-or-later
.\"
.\" Modified 2004-11-15, Added further text on FLT_ROUNDS
.\" as suggested by AEB and Fabian Kreutz
.\"
.TH fma 3 2024-05-02 "Linux man-pages (unreleased)"
.SH NAME
fma, fmaf, fmal \- floating-point multiply and add
.SH LIBRARY
Math library
.RI ( libm ", " \-lm )
.SH SYNOPSIS
.nf
.B #include <math.h>
.P
.BI "double fma(double " x ", double " y ", double " z );
.BI "float fmaf(float " x ", float " y ", float " z );
.BI "long double fmal(long double " x ", long double " y ", long double " z );
.fi
.P
.RS -4
Feature Test Macro Requirements for glibc (see
.BR feature_test_macros (7)):
.RE
.P
.BR fma (),
.BR fmaf (),
.BR fmal ():
.nf
_ISOC99_SOURCE || _POSIX_C_SOURCE >= 200112L
.fi
.SH DESCRIPTION
These functions compute
.IR x " * " y " + " z .
The result is rounded as one ternary operation according to the
current rounding mode (see
.BR fenv (3)).
.SH RETURN VALUE
These functions return the value of
.IR x " * " y " + " z ,
rounded as one ternary operation.
.P
If
.I x
or
.I y
is a NaN, a NaN is returned.
.P
If
.I x
times
.I y
is an exact infinity, and
.I z
is an infinity with the opposite sign,
a domain error occurs,
and a NaN is returned.
.P
.\" POSIX.1-2008 allows some possible differences for the following two
.\" domain error cases, but on Linux they are treated the same (AFAICS).
.\" Nevertheless, we'll mirror POSIX.1 and describe the two cases
.\" separately.
If one of
.I x
or
.I y
is an infinity, the other is 0, and
.I z
is not a NaN,
a domain error occurs, and
a NaN is returned.
.\" POSIX.1 says that a NaN or an implementation-defined value shall
.\" be returned for this case.
.P
If one of
.I x
or
.I y
is an infinity, and the other is 0, and
.I z
is a NaN,
.\" POSIX.1 makes the domain error optional for this case.
a domain error occurs, and
a NaN is returned.
.P
If
.I x
times
.I y
is not an infinity times zero (or vice versa), and
.I z
is a NaN,
a NaN is returned.
.P
If the result overflows,
a range error occurs, and
an infinity with the correct sign is returned.
.P
If the result underflows,
a range error occurs, and
a signed 0 is returned.
.SH ERRORS
See
.BR math_error (7)
for information on how to determine whether an error has occurred
when calling these functions.
.P
The following errors can occur:
.TP
Domain error: \fIx\fP * \fIy\fP + \fIz\fP, \
or \fIx\fP * \fIy\fP is invalid and \fIz\fP is not a NaN
.\" .I errno
.\" is set to
.\" .BR EDOM .
An invalid floating-point exception
.RB ( FE_INVALID )
is raised.
.TP
Range error: result overflow
.\" .I errno
.\" is set to
.\" .BR ERANGE .
An overflow floating-point exception
.RB ( FE_OVERFLOW )
is raised.
.TP
Range error: result underflow
.\" .I errno
.\" is set to
.\" .BR ERANGE .
An underflow floating-point exception
.RB ( FE_UNDERFLOW )
is raised.
.P
These functions do not set
.IR errno .
.\" FIXME . Is it intentional that these functions do not set errno?
.\" Bug raised: https://www.sourceware.org/bugzilla/show_bug.cgi?id=6801
.SH ATTRIBUTES
For an explanation of the terms used in this section, see
.BR attributes (7).
.TS
allbox;
lbx lb lb
l l l.
Interface Attribute Value
T{
.na
.nh
.BR fma (),
.BR fmaf (),
.BR fmal ()
T} Thread safety MT-Safe
.TE
.SH STANDARDS
C11, POSIX.1-2008.
.SH HISTORY
glibc 2.1.
C99, POSIX.1-2001.
.SH SEE ALSO
.BR remainder (3),
.BR remquo (3)
|
