Differences between revisions 11 and 12
Revision 11 as of 2008-06-05 21:36:38
Size: 7271
Editor: MartinSebor
Comment: Copied relevant portions of C99.
Revision 12 as of 2009-09-20 23:02:58
Size: 7273
Editor: localhost
Comment: converted to 1.6 markup
Deletions are marked like this. Additions are marked like this.
Line 5: Line 5:
In preparation for resolving [http://issues.apache.org/jira/browse/STDCXX-51 STDCXX-51] this page outlines the platform-specific details describing infinities and NaNs produced as the result of certain floating point calculations, usually triggered by invalid operands. In preparation for resolving [[http://issues.apache.org/jira/browse/STDCXX-51|STDCXX-51]] this page outlines the platform-specific details describing infinities and NaNs produced as the result of certain floating point calculations, usually triggered by invalid operands.

WORK IN PROGRESS

Floating Point Numbers, Infinities and NaNs

In preparation for resolving STDCXX-51 this page outlines the platform-specific details describing infinities and NaNs produced as the result of certain floating point calculations, usually triggered by invalid operands.

libc symbols

C99 specifies the following interface to infinities and NaNs:

7.12 Mathematics <math.h>

-4- The macro INFINITY expands to a constant expression of type float representing positive or unsigned infinity, if available; else to a positive constant of type float that overflows at translation time.

-5- The macro NAN is defined if and only if the implementation supports quiet NaNs for the float type. It expands to a constant expression of type float representing a quiet NaN.

7.12.11.2 The nan functions

-1- Synopsis

#include <math.h> 
double nan(const char *tagp); 
float nanf(const char *tagp); 
long double nanl(const char *tagp);

Description

-2- The call nan("n-char-sequence") is equivalent to strtod("NAN(n-charsequence)", (char**) NULL); the call nan("") is equivalent to strtod("NAN()", (char**) NULL). If tagp does not point to an n-char sequence or an empty string, the call is equivalent to strtod("NAN", (char**) NULL). Calls to nanf and nanl are equivalent to the corresponding calls to strtof and strtold.

Returns

-3- The nan functions return a quiet NaN, if available, with content indicated through tagp. If the implementation does not support quiet NaNs, the functions return zero.

The table below lists the public symbols defined on each platform for Infinity, quiet NaN, and signaling NaN.

libc symbols

AIX

HP-UX

IRIX

Linux

Solaris

Tru64 UNIX

Windows

header

<float.h>

<math.h>

N/A

<math.h>

<sunmath.h>

<float.h>

Infinity

INFINITY, FLT_INF, DBL_INF, LDBL_INF

INFINITY

N/A

INFINITY

INFINITY, infinity()

FLT_INFINITY, DBL_INFINITY, LDBL_INFINITY

N/A

Quiet NaN

FLT_QNAN, DBL_QNAN, LDBL_QNAN 2

nan(const char*) 1

N/A

nan(const char*) 1

quiet_nan(long) 2

FLT_QNAN, DBL_QNAN, LDBL_QNAN

N/A

Signaling NaN

FLT_SNAN, DBL_SNAN, LDBL_SNAN

N/A

N/A

signaling_nan(long)

FLT_SNAN, DBL_SNAN, LDBL_SNAN

N/A

1 Supports all signatures required by C99 and POSIX: float nanf(const char*), double nan(const char*), and long double nanl(const char*).

2 Recent versions also support the C99 interface.

printf() formatting

printf() formatting

AIX

HP-UX

IRIX

Linux

Solaris

Tru64 UNIX

Windows/x86

Windows/x64

Infinity

+/- INF

+/- inf

+/- inf

+/- inf

+/- Inf

+/- INF

+/- 1.#INF

+/- 1.#INF

Quiet NaN

+/- NaNQ

+/- nan

+/- nan0xxx

+/- nan

+/- NaN

NaNQ

-/- 1.#IND

-/- 1.#IND

Signaling NaN

+/- NaNS

+/- nan

+/- nan0xxx

+/- nan

+/- NaN

NaNS

+/- 1.#QNAN

+/- 1.#SNAN

numeric_limits<double> values/formatting

numeric_limits<double> values/formatting

AIX

HP-UX

IRIX

Linux

Solaris

Tru64 UNIX

Windows/x86

Windows/x64

XLC++ 9.0

aCC 3,6

MIPSpro

gcc

Sun C++

HP C++

MSVC

MSVC

infinity()

INF

inf

inf

inf

Inf

0

1.#INF

1.#INF

quiet_NaN()

-NaNQ

nan

nan0xxx

nan

NaN

0

1.#QNAN

1.#QNAN

signaling_NaN()

-INF

nan

nan0xxx

nan

NaN

0

1.#QNAN

1.#SNAN

Bit Patterns

The table below shows the bit patterns for Infinity, Quiet NaN, and Signaling NaN on each platform.

IEEE 754 double precision bit patterns

Number

Sign

Exponent

Fraction

AIX/Power

Infinity

0

0x7ff

0

Quiet NaN

0

0x7ff

0x80000

Signaling NaN

0

0x7ff

0x5555555500055555

HP-UX

Infinity

0

0x7ff

0

Quiet NaN

0

0x7ff

0x40000

Signaling NaN

0

0x7ff

0x80000

IRIX/MIPS

Infinity

Quiet NaN

Signaling NaN

Linux/x86

Infinity

0

0x7ff

0

Quiet NaN1

0

0x7ff

0x80000

Signaling NaN1

0

0x7ff

0x40000

Solaris/SPARC

Infinity

Quiet NaN

Signaling NaN

Solaris/x86

Infinity

Quiet NaN1

Signaling NaN1

Tru64 UNIX/Alpha

Infinity

0

0

0x7ff0000000000000

Quiet NaN

0

0

0xfff8000000000000

Signaling NaN

1

0x2aa

0x7ff5555500055555

Windows/x86

Infinity

Quiet NaN

Signaling NaN

1 Intel and AMD processors set the first fraction bit to 1 for Quiet NaNs and clear it for Signaling NaNs.

FloatingPoint (last edited 2009-09-20 23:02:58 by localhost)