large float's

[yawnmoth]

According to php.net, the largest possible float is ~1.8e308.
wikipedia.org offers more insight, suggesting that is, exactly,
0x7FEFFFFFFFFFF FFF. This sorta begs the question... how can this
number be represented in PHP without resorting to the decimal notation?

<?
echo 0x7FEFFFFFFFFFF FFF;
?>

....doesn't work. Nor does...

<?
echo hexdec('7FEFFFF FFFFFFFFF');
?>

So... any ideas as to what will work?

[Oli Filth]

Re: large float's

yawnmoth said the following on 23/02/2006 06:00:[color=blue]
> According to php.net, the largest possible float is ~1.8e308.
> wikipedia.org offers more insight, suggesting that is, exactly,
> 0x7FEFFFFFFFFFF FFF. This sorta begs the question... how can this
> number be represented in PHP without resorting to the decimal notation?
>[/color]

The maximum floating-point value is platform dependent. For a 64-bit
platform, it's ~1.8e308, but for a 32-bit platform it's ~3.4e38
(assuming you're using the IEEE 754 standard).

This code converts from a hex-representation to a floating-point
representation (for 32-bit):


define("LENGTH" , 32);
define("SIGN_BI T", 31);
define("EXPONEN T_MSB", 30);
define("EXPONEN T_LSB", 23);
define("MANTISS A_MSB", 22);
define("MANTISS A_LSB", 0);

function hex2float($strH ex)
{
$dec = hexdec($strHex) ;
$sign = ($dec & (1 << SIGN_BIT)) != 0;
$exp = (($dec & ((2 << EXPONENT_MSB) - (1 << EXPONENT_LSB))) >>
EXPONENT_LSB)
- (1 << (EXPONENT_MSB - EXPONENT_LSB))
- (MANTISSA_MSB - MANTISSA_LSB);
$man = (($dec & ((2 << MANTISSA_MSB) - (1 << MANTISSA_LSB))) >>
MANTISSA_LSB)
+ (2 << (MANTISSA_MSB - MANTISSA_LSB));
$float = floatval($man * pow(2, $exp) * ($sign ? -1 : 1));
}


For a 64-bit version, you'll need to alter the constants; you can get
the required values from
http://stevehollasch.com/cgindex/coding/ieeefloat.html.

Note that this code doesn't deal with special cases, i.e. NaN, infinity,
etc.

--
Oli

[Chung Leong]

Re: large float's

Oli Filth wrote:[color=blue]
> The maximum floating-point value is platform dependent. For a 64-bit
> platform, it's ~1.8e308, but for a 32-bit platform it's ~3.4e38
> (assuming you're using the IEEE 754 standard).[/color]

Actually the bit-ness of a CPU only refers to its integer handling
capability. Nearly all FPUs--even the ancient 8087--can handle both
single and double precision floating points.

Float points in PHP are defined as doubles, which in C are always
64-bit IIRC.