XS and -Duse64bitint

syphilis has asked for the wisdom of the Perl Monks concerning the following question:

Hi,
This is a re-hash of a question I asked on the XS list a few hours ago. (No replies yet.) It's an issue only with perls built with -Duse64bitint.

Consider the following code and output:

use warnings;
use Inline C => Config =>
    BUILD_NOISY => 1;

use Inline C => <<'EOC';

int foo(SV * integer) {

     char * str;

     str = SvPV_nolen(integer);

     if(SvIOK(integer)) {
       printf("IV: %s\n", str);
       return 1;
       }

     if(SvNOK(integer)) {
       printf("NV: %s\n", str);
       return 2;
       }

     printf("Neither IV nor NV");
     return 3;
}

EOC

{
use integer;
$x = 2 ** 57 + 12345;
}

$y = 2 ** 57 + 12345;

foo($x);
foo($y);

__END__
Outputs:

IV: 144115188075868217
NV: 1.44115188075868e+17
[download]

Note that 'foo($y)' creates the string "1.44115188075868e+17". I need to make it create the same string as 'foo($x)' does - ie "144115188075868217". How do I do that from within the foo() function ? (I'm not interested in solutions that involve fiddling with the perl side of things.)

Cheers,
Rob

Comment on XS and -Duse64bitint Download Code

Replies are listed 'Best First'.
Re: XS and -Duse64bitint by almut (Canon) on Feb 13, 2007 at 11:55 UTC
You could add an explicit conversion to IV, e.g. `... char * str; sv_setiv(integer, SvIV(integer)); // <--- here str = SvPV_nolen(integer); ...` [download] However, when doing that, I get (note the difference in the last two digits): `IV: 144115188075868217 IV: 144115188075868224` [download] I suppose this is due to the insufficient precision in the NV representation... Actually, I think it's too late to deal with this within your XS routine, as the value will already have been converted to NV when the SV gets passed to `foo()` - unless you `use integer` on the Perl side (but you already figured that out :)	[reply] [d/l] [select]
Re^2: XS and -Duse64bitint by syphilis (Archbishop) on Feb 13, 2007 at 23:00 UTC
I suppose this is due to the insufficient precision in the NV representation Yes, you're right - the precision has been lost. I was fooled by the fact that the following tests re $x (the IV) and $y (the NV) return "true": `$x == $y $x - 1 == $y - 1 $x + 1 == $y + 1` [download] I took that to indicate that perl knows that $y contains the value 144115188075868217, and that my problem was merely one of how to get at that value. The following tests also return "true": `$x == $y + 1 $x == $y - 1` [download] which doesn't fit too well with the hypothesis that "no precision was lost" :-) Thanks almut, for setting me straight. Cheers, Rob	[reply] [d/l] [select]
Re: XS and -Duse64bitint by Anonymous Monk on Feb 13, 2007 at 09:40 UTC
use integer	[reply]