Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.lang.c > #25232

Re: OxCp syntax ?

Show all headers | View raw

On 08/28/2012 07:28 AM, Guillaume Dargaud wrote:
> Hello all,
> I just ran across this syntax:
> #define NAN 0xCp125f
> Can somebody explain ? A quick googling turned out nothing.

C99 added the capability of specifying floating point constants using
hexadecimal notation. They are case-insensitive, just like hexadecimal
constants and decimal floating point constants. The 0x is the
hexadecimal prefix. "C" is the hexadecimal digit sequence, corresponding
to a decimal value of 13. 'p' is the separator, corresponding to "e" in
decimal floating point constants. 125 is the exponent. "f" is a suffix
indicating that this constant has the type 'float'. This particular
constant represents a float value of 0xC*2^125. Compare to 5e60, which
represents 5*10^60. Fractional parts are also allowed, but unlike
decimal floating point constants, the binary exponent part cannot be
dropped:

#define PI 0x3.243F6A88p0

One of the most important characteristics of hexadecimal floating point
constants is explained in 6.4.4.2p3:
> ... For decimal floating constants, and also for
> hexadecimal floating constants when FLT_RADIX is not a power of 2, the result is either
> the nearest representable value, or the larger or smaller representable value immediately
> adjacent to the nearest representable value, chosen in an implementation-defined manner.
> For hexadecimal floating constants when FLT_RADIX is a power of 2, the result is
> correctly rounded. ...

Almost all real-world implementations of C have a power of 2 for
FLT_RADIX, so this guarantee means that using a hexadecimal floating
point value gives you more precise control over the value that it
represents than a decimal floating point value would.

Note: it's not clear to me that, on a system which supports true NaNs,
that 0xCp125f could be a NaN. If 0xC*2^125 is representable, then by
definition it isn't a NaN. If it is not representable, then a diagnostic
is required (the behavior being undefined at that point, it would be
permissible for a true NaN to be generated). However, if you ran into
this definition inside of <math.h>, then I might be wrong about that.
-- 
James Kuyper

Back to comp.lang.c | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread

Thread

OxCp syntax ? Guillaume Dargaud <use_the_contact_form@www.gdargaud.net> - 2012-08-28 13:28 +0200
  Re: OxCp syntax ? Les Cargill <lcargill99@comcast.com> - 2012-08-28 07:30 -0500
    Re: OxCp syntax ? Les Cargill <lcargill99@comcast.com> - 2012-08-28 07:41 -0500
  Re: OxCp syntax ? James Kuyper <jameskuyper@verizon.net> - 2012-08-28 09:00 -0400
    Re: OxCp syntax ? Eric Sosman <esosman@ieee-dot-org.invalid> - 2012-08-28 09:10 -0400
      Re: OxCp syntax ? James Kuyper <jameskuyper@verizon.net> - 2012-08-28 09:58 -0400
  Re: OxCp syntax ? Eric Sosman <esosman@ieee-dot-org.invalid> - 2012-08-28 09:02 -0400
    Re: OxCp syntax ? Vincenzo Mercuri <vincenzo.mercuri@yahoo.it> - 2012-08-29 19:16 +0200
  Re: OxCp syntax ? Nobody <nobody@nowhere.com> - 2012-08-28 14:17 +0100
  Re: OxCp syntax ? Guillaume Dargaud <use_the_contact_form@www.gdargaud.net> - 2012-08-29 17:50 +0200

csiph-web