How to tell if a float is SNAN or QNAN

▲
4
▼

I am trying to figure out how to print out if a floating point number is QNAN or SNAN. I have already separated out the bits into the signBit exponentBit and the FractBits.

unsigned int sign = (i & 0x80000000) >> 31;
unsigned int exponent = (i & 0x7f800000) >> 23;
unsigned int fraction = (i & 0x007FFFFF);
printf("signBit %d, expBits %d, fractBits 0x%08X\n",sign, exponent, fraction);
c++
c
gcc
assembly
ieee-754
asked on Stack Overflow Apr 16, 2015 by thepi

2 Answers

▲
8
▼

GNU provides a facility which was recently standardized:

Macro: int issignaling (float-type x)

Preliminary: | MT-Safe | AS-Safe | AC-Safe | See POSIX Safety Concepts.

This macro returns a nonzero value if x is a signaling NaN (sNaN). It is based on draft TS 18661 and currently enabled as a GNU extension.

The final draft TS mentions that you might need to opt-in with a macro to get it:

#define __STDC_WANT_IEC_60559_BFP_EXT__
#include <math.h> 
int issignaling(real-floating x);
answered on Stack Overflow Apr 16, 2015 by Potatoswatter • edited Apr 16, 2015 by Potatoswatter
▲
2
▼

The format of the floating-point real numbers is processor-dependent. On x86 / x86-64 ISA, the top bit of the significand = 1 for quiet, 0 for signalling. (The rest of the NaN payload is still arbitrary).

  • float type:

    enter image description here

  • double type:

    enter image description here

  • long double type (not supported by Microsoft compilers):

    enter image description here

answered on Stack Overflow Sep 20, 2018 by ahmd0 • edited Sep 20, 2018 by Peter Cordes

User contributions licensed under CC BY-SA 3.0