d/freebsd-skq
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Sanitize the markup, as prompted.

Browse Source
This commit is contained in:
Ruslan Ermilov 2005-01-11 20:16:03 +00:00
parent 7a13f01f6f
commit 4e05ab77a8
1 changed files with 174 additions and 183 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

351
lib/msun/man/math.3
View File

@ -35,15 +35,13 @@
.Dd June 11, 2004
.Dt MATH 3
.Os
.ds up \fIulp\fR
.de If
.if n \\
\\$1Infinity\\$2
.if t \\
\\$1\\(if\\$2
..
.if n \{\
.char \[if] "Infinity
.char \[sr] "sqrt
.\}
.Sh NAME
math \- floating-point mathematical library
.Nm math
.Nd "floating-point mathematical library"
.Sh LIBRARY
.Lb libm
.Sh SYNOPSIS
@ -58,14 +56,14 @@ functions has a
counterpart with an
.Ql f
appended to the name and a
.Vt long double
.Vt "long double"
counterpart with an
.Ql l
appended.
As an example, the
.Vt float
and
.Vt long double
.Vt "long double"
counterparts of
.Ft double
.Fn acos "double x"
@ -73,19 +71,22 @@ are
.Ft float
.Fn acosf "float x"
and
.Ft long double
.Ft "long double"
.Fn acosl "long double x" ,
respectively.
.Pp
The programs are accurate to within the numbers
of \*(ups tabulated below; an \*(up is one \fIU\fRnit in the \fIL\fRast
\fIP\fRlace.
.sp 2
.nf
.ta \w'nexttoward'u+10n +\w'remainder with partial quot'u
\fIName\fP \fIDescription\fP \fIError Bound (ULPs)\fP
.ta \w'nexttoward'u+4n +\w'remainder with partial quotient'u+6nC
.sp 5p
of
.Em ulp Ns s
tabulated below; an
.Em ulp
is one
.Em U Ns nit
in the
.Em L Ns ast
.Em P Ns lace .
.Bl -column "nexttoward" "remainder with partial quotient"
.Em "Name Description Error Bound (ULPs)"
.\" XXX Many of these error bounds are wrong for the current implementation!
acos inverse trigonometric function ???
acosh inverse hyperbolic function ???
@ -133,7 +134,7 @@ modf extract fractional part 0
nearbyint round to integer 0
nextafter next representable value 0
.\" nexttoward next representable value 0
pow exponential x**y 60\-500
pow exponential x**y 60-500
remainder remainder 0
.\" remquo remainder with partial quotient ???
rint round to nearest integer 0
@ -150,8 +151,7 @@ trunc round towards zero 0
y0 bessel function ???
y1 bessel function ???
yn bessel function ???
.ta
.fi
.El
.Sh NOTES
Virtually all modern floating-point units attempt to support
IEEE Standard 754 for Binary Floating-Point Arithmetic.
@ -161,168 +161,144 @@ except for the few documented in
it primarily defines representations of numbers and abstract
properties of arithmetic operations relating to precision, rounding,
and exceptional cases, as described below.
.Pp
\fBIEEE STANDARD 754 Floating\-Point Arithmetic:\fR
.Pp
.Ss IEEE STANDARD 754 Floating-Point Arithmetic
.\" XXX mention single- and extended-/quad- precisions
Properties of IEEE 754 Double\-Precision:
.Bd -filled -offset indent
Wordsize: 64 bits, 8 bytes. Radix: Binary.
.br
Precision: 53
.if n \
sig.
.if t \
significant
bits, roughly like 16
.if n \
sig.
.if t \
significant
decimals.
.Bd -filled -offset indent -compact
If x and x' are consecutive positive Double\-Precision
numbers (they differ by 1 \*(up), then
.br
Properties of IEEE 754 Double-Precision:
.Bd -ragged -offset indent -compact
Wordsize: 64 bits, 8 bytes.
.Pp
Radix: Binary.
.Pp
Precision: 53 significant bits,
roughly like 16 significant decimals.
.Bd -ragged -offset indent -compact
If x and x' are consecutive positive Double-Precision
numbers (they differ by 1
.Em ulp ) ,
then
.Bd -ragged -compact
1.1e\-16 < 0.5**53 < (x'\-x)/x \(<= 0.5**52 < 2.3e\-16.
.Ed
.nf
.ta \w'Range:'u+1n +\w'Underflow threshold'u+1n +\w'= 2.0**1024'u+1n
.Ed
.Pp
.Bl -column "XXX" -compact
Range: Overflow threshold = 2.0**1024 = 1.8e308
Underflow threshold = 0.5**1022 = 2.2e\-308
.ta
.fi
.Bd -filled -offset indent -compact
Overflow goes by default to a signed
.If "" .
.br
Underflow is \fIGradual,\fR rounding to the nearest
.El
.Bd -ragged -offset indent -compact
Overflow goes by default to a signed \(if.
Underflow is
.Em Gradual ,
rounding to the nearest
integer multiple of 0.5**1074 = 4.9e\-324.
.Ed
.Pp
Zero is represented ambiguously as +0 or \-0.
.Bd -filled -offset indent -compact
.Bd -ragged -offset indent -compact
Its sign transforms correctly through multiplication or
division, and is preserved by addition of zeros
with like signs; but x\-x yields +0 for every
finite x. The only operations that reveal zero's
sign are division by zero and copysign(x,\(+-0).
In particular, comparison (x > y, x \(>= y, etc.)
finite x.
The only operations that reveal zero's
sign are division by zero and
.Fn copysign x \(+-0 .
In particular, comparison (x > y, x \(>= y, etc.)\&
cannot be affected by the sign of zero; but if
finite x = y then
.If
\&= 1/(x\-y)
.if n \
!=
.if t \
\(!=
\-1/(y\-x) =
.If \- .
finite x = y then \(if = 1/(x\-y) \(!= \-1/(y\-x) = \-\(if.
.Ed
.If
is signed.
.Bd -filled -offset indent -compact
it persists when added to itself
or to any finite number. Its sign transforms
.Pp
\(if is signed.
.Bd -ragged -offset indent -compact
It persists when added to itself
or to any finite number.
Its sign transforms
correctly through multiplication and division, and
.If (finite)/\(+- \0=\0\(+-0
(nonzero)/0 =
.If \(+- .
(finite)/\(+-\(if\0=\0\(+-0
(nonzero)/0 = \(+-\(if.
But
.if n \
Infinity\-Infinity, Infinity\(**0 and Infinity/Infinity
.if t \
\(if\-\(if, \(if\(**0 and \(if/\(if
are, like 0/0 and sqrt(\-3),
invalid operations that produce \*(Na. ...
.Ed
.Pp
Reserved operands:
.Bd -filled -offset indent -compact
.Bd -ragged -offset indent -compact
there are 2**53\-2 of them, all
called \*(Na (\fIN\fRot \fIa N\fRumber).
Some, called Signaling \*(Nas, trap any floating\-point operation
called \*(Na
.Em ( N Ns ot Em a N Ns umber ) .
Some, called Signaling \*(Nas, trap any floating-point operation
performed upon them; they are used to mark missing
or uninitialized values, or nonexistent elements
of arrays. The rest are Quiet \*(Nas; they are
of arrays.
The rest are Quiet \*(Nas; they are
the default results of Invalid Operations, and
propagate through subsequent arithmetic operations.
If x
.if n \
!=
.if t \
\(!=
x then x is \*(Na; every other predicate
If x \(!= x then x is \*(Na; every other predicate
(x > y, x = y, x < y, ...) is FALSE if \*(Na is involved.
.br
.Pp
NOTE: Trichotomy is violated by \*(Na.
.Bd -filled -offset indent -compact
Besides being FALSE, predicates that entail ordered
comparison, rather than mere (in)equality,
signal Invalid Operation when \*(Na is involved.
.Ed
.Ed
.Pp
Rounding:
.Bd -filled -offset indent -compact
.Bd -ragged -offset indent -compact
Every algebraic operation (+, \-, \(**, /,
.if n \
sqrt)
.if t \
\(sr)
is rounded by default to within half an \*(up, and
when the rounding error is exactly half an \*(up then
is rounded by default to within half an
.Em ulp ,
and when the rounding error is exactly half an
.Em ulp
then
the rounded value's least significant bit is zero.
This kind of rounding is usually the best kind,
sometimes provably so; for instance, for every
x = 1.0, 2.0, 3.0, 4.0, ..., 2.0**52, we find
(x/3.0)\(**3.0 == x and (x/10.0)\(**10.0 == x and ...
despite that both the quotients and the products
have been rounded. Only rounding like IEEE 754
can do that. But no single kind of rounding can be
have been rounded.
Only rounding like IEEE 754 can do that.
But no single kind of rounding can be
proved best for every circumstance, so IEEE 754
provides rounding towards zero or towards
.If +
or towards
.If \-
at the programmer's option. And the
+\(if or towards \-\(if
at the programmer's option.
And the
same kinds of rounding are specified for
Binary\-Decimal Conversions, at least for magnitudes
Binary-Decimal Conversions, at least for magnitudes
between roughly 1.0e\-10 and 1.0e37.
.Ed
.Pp
Exceptions:
.Bd -filled -offset indent -compact
IEEE 754 recognizes five kinds of floating\-point exceptions,
.Bd -ragged -offset indent -compact
IEEE 754 recognizes five kinds of floating-point exceptions,
listed below in declining order of probable importance.
.Bd -filled -offset indent -compact
.nf
.ta \w'Invalid Operation'u+6n +\w'Gradual Underflow'u+2n
Exception Default Result
.tc \(ru
.tc
.Bl -column -offset indent "Invalid Operation" "Gradual Underflow"
.Em "Exception Default Result"
Invalid Operation \*(Na, or FALSE
.if n \{\
Overflow \(+-Infinity
Divide by Zero \(+-Infinity \}
.if t \{\
Overflow \(+-\(if
Divide by Zero \(+-\(if \}
Divide by Zero \(+-\(if
Underflow Gradual Underflow
Inexact Rounded value
.ta
.fi
.Ed
.El
.Pp
NOTE: An Exception is not an Error unless handled
badly. What makes a class of exceptions exceptional
badly.
What makes a class of exceptions exceptional
is that no single default response can be satisfactory
in every instance. On the other hand, if a default
in every instance.
On the other hand, if a default
response will serve most instances satisfactorily,
the unsatisfactory instances cannot justify aborting
computation every time the exception occurs.
.Ed
.Pp
For each kind of floating\-point exception, IEEE 754
For each kind of floating-point exception, IEEE 754
provides a Flag that is raised each time its exception
is signaled, and stays raised until the program resets
it. Programs may also test, save and restore a flag.
it.
Programs may also test, save and restore a flag.
Thus, IEEE 754 provides three ways by which programs
may cope with exceptions for which the default result
might be unsatisfactory:
@ -336,20 +312,17 @@ since the program last reset its flag.
.It
Test a result to see whether it is a value that only
an exception could have produced.
.RS
.Pp
CAUTION: The only reliable ways to discover
whether Underflow has occurred are to test whether
products or quotients lie closer to zero than the
underflow threshold, or to test the Underflow
flag. (Sums and differences cannot underflow in
IEEE 754; if x
.if n \
!=
.if t \
\(!=
y then x\-y is correct to
flag.
(Sums and differences cannot underflow in
IEEE 754; if x \(!= y then x\-y is correct to
full precision and certainly nonzero regardless of
how tiny it may be.) Products and quotients that
how tiny it may be.)
Products and quotients that
underflow gradually can lose accuracy gradually
without vanishing, so comparing them with zero
(as one might on a VAX) will not reveal the loss.
@ -358,19 +331,22 @@ destined to be added to something bigger than the
underflow threshold, as is almost always the case,
digits lost to gradual underflow will not be missed
because they would have been rounded off anyway.
So gradual underflows are usually \fIprovably\fR ignorable.
So gradual underflows are usually
.Em provably
ignorable.
The same cannot be said of underflows flushed to 0.
.RE
.El
.Pp
At the option of an implementor conforming to IEEE 754,
other ways to cope with exceptions may be provided:
.Bl -hang -width 3n
.It 4.
ABORT. This mechanism classifies an exception in
.Bl -enum
.It
ABORT.
This mechanism classifies an exception in
advance as an incident to be handled by means
traditionally associated with error\-handling
statements like "ON ERROR GO TO ...". Different
traditionally associated with error-handling
statements like "ON ERROR GO TO ...".
Different
languages offer different forms of this statement,
but most share the following characteristics:
.Bl -dash
@ -378,28 +354,32 @@ but most share the following characteristics:
No means is provided to substitute a value for
the offending operation's result and resume
computation from what may be the middle of an
expression. An exceptional result is abandoned.
expression.
An exceptional result is abandoned.
.It
In a subprogram that lacks an error\-handling
In a subprogram that lacks an error-handling
statement, an exception causes the subprogram to
abort within whatever program called it, and so
on back up the chain of calling subprograms until
an error\-handling statement is encountered or the
an error-handling statement is encountered or the
whole task is aborted and memory is dumped.
.El
.It 5.
STOP. This mechanism, requiring an interactive
.It
STOP.
This mechanism, requiring an interactive
debugging environment, is more for the programmer
than the program. It classifies an exception in
than the program.
It classifies an exception in
advance as a symptom of a programmer's error; the
exception suspends execution as near as it can to
the offending operation so that the programmer can
look around to see how it happened. Quite often
look around to see how it happened.
Quite often
the first several exceptions turn out to be quite
unexceptionable, so the programmer ought ideally
to be able to resume execution after each one as if
execution had not been stopped.
.It 6.
.It
\&... Other ways lie beyond the scope of this document.
.El
.Ed
@ -407,14 +387,14 @@ execution had not been stopped.
Ideally, each
elementary function should act as if it were indivisible, or
atomic, in the sense that ...
.Bl -tag -width "iii)"
.It i)
.Bl -enum
.It
No exception should be signaled that is not deserved by
the data supplied to that function.
.It ii)
.It
Any exception signaled should be identified with that
function rather than with one of its subroutines.
.It iii)
.It
The internal behavior of an atomic function should not
be disrupted when a calling program changes from
one to another of the five or so ways of handling
@ -423,51 +403,60 @@ of the function may be correlated intentionally
with exception handling.
.El
.Pp
The functions in \fIlibm\fR are only approximately atomic.
The functions in
.Nm libm
are only approximately atomic.
They signal no inappropriate exception except possibly ...
.Bd -filled -offset indent -compact
.Bl -tag -width indent -offset indent -compact
.It Xo
Over/Underflow
.Bd -filled -offset indent -compact
.Xc
when a result, if properly computed, might have lain barely within range, and
.Ed
Inexact in \fIcabs\fR, \fIcbrt\fR, \fIhypot\fR, \fIlog10\fR and \fIpow\fR
.Bd -filled -offset indent -compact
.It Xo
Inexact in
.Fn cabs ,
.Fn cbrt ,
.Fn hypot ,
.Fn log10
and
.Fn pow
.Xc
when it happens to be exact, thanks to fortuitous cancellation of errors.
.Ed
.Ed
.El
Otherwise, ...
.Bd -filled -offset indent -compact
.Bl -tag -width indent -offset indent -compact
.It Xo
Invalid Operation is signaled only when
.Bd -filled -offset indent -compact
.Xc
any result but \*(Na would probably be misleading.
.Ed
.It Xo
Overflow is signaled only when
.Bd -filled -offset indent -compact
.Xc
the exact result would be finite but beyond the overflow threshold.
.Ed
Divide\-by\-Zero is signaled only when
.Bd -filled -offset indent -compact
.It Xo
Divide-by-Zero is signaled only when
.Xc
a function takes exactly infinite values at finite operands.
.Ed
.It Xo
Underflow is signaled only when
.Bd -filled -offset indent -compact
.Xc
the exact result would be nonzero but tinier than the underflow threshold.
.Ed
.It Xo
Inexact is signaled only when
.Bd -filled -offset indent -compact
.Xc
greater range or precision would be needed to represent the exact result.
.Ed
.Ed
.El
.Sh BUGS
Several functions required by
.St -isoC-99
are missing, and many functions are not available in their
.Vt long double
.Vt "long double"
variants.
.Pp
On some architectures, trigonometric argument reduction is not
performed accurately, resulting in errors greater than 1 ulp for large
arguments to
performed accurately, resulting in errors greater than 1
.Em ulp
for large arguments to
.Fn cos ,
.Fn sin ,
and
@ -478,8 +467,8 @@ and
.Pp
An explanation of IEEE 754 and its proposed extension p854
was published in the IEEE magazine MICRO in August 1984 under
the title "A Proposed Radix\- and Word\-length\-independent
Standard for Floating\-point Arithmetic" by W. J. Cody et al.
the title "A Proposed Radix- and Word-length-independent
Standard for Floating-point Arithmetic" by W. J. Cody et al.
The manuals for Pascal, C and BASIC on the Apple Macintosh
document the features of IEEE 754 pretty well.
Articles in the IEEE magazine COMPUTER vol. 14 no. 3 (Mar.\&
@ -488,8 +477,10 @@ Oct. 1979, may be helpful although they pertain to
superseded drafts of the standard.
.Sh HISTORY
A math library with many of the present functions appeared in
Version 7 AT&T UNIX.
The library was substantially rewritten for 4.3BSD to provide
.At v7 .
The library was substantially rewritten for
.Bx 4.3
to provide
better accuracy and speed on machines supporting either VAX
or IEEE 754 floating-point.
Most of this library was replaced with FDLIBM, developed at Sun