Pari lib with Math-BigInt (win32)

[gnu valued customer]

hello,

Win2k sp3
Perl 5.8.1 (ActiveState extensions)

using Math-Pari 2.010500

and the latest Math-BigInt i've
tried to install and use
Math-BigInt-Pari.

Everything installs fine, but I get
this warning when I call like so

use Math::BigInt lib => 'Pari';

warning:
Cannot load outdated Math::BigInt:ari v1.10, please
upgrade at ... (points to about 'use')

The script is reverting to the Pure-Perl version of
Math-BigInt (slow).

tia for your help,
Mark Pryor

 

[Sisyphus]

warning:
Cannot load outdated Math::BigInt:ari v1.10, please
upgrade at ... (points to about 'use')

Version 1.10 was put on CPAN earlier this year, and is the latest
available ... calling it "outdated" seems a little severe to me

I get the same. The error is produced by Math::BigInt-1.70 - search
'BigInt.pm' for "outdated". I tried Math::BigInt-1.64 and the error did
not arise, though 'nmake test' threw up some other errors. Perhaps using
version 1.69 or 1.68 instead of 1.70 would provide a temporary workaround.

The M::BI: author (Tels), who is also currently maintaining M::BI,
might be able to provide some insight if you don't get a satisfactory
answer here. (He should at least be informed of the problem.)

Cheers,
Rob

 

[gnu valued customer]

Hello Rob,

Version 1.10 was put on CPAN earlier this year, and is the latest
available ... calling it "outdated" seems a little severe to me

I get the same. The error is produced by Math::BigInt-1.70 - search
'BigInt.pm' for "outdated". I tried Math::BigInt-1.64 and the error did
not arise, though 'nmake test' threw up some other errors. Perhaps using
version 1.69 or 1.68 instead of 1.70 would provide a temporary workaround.

After having this suggestion seconded by Tels, I found v. 1.68 here
http://backpan.cpan.org/authors/id/T/TE/TELS/math/

Thanks Rob for pointing out such an easy workaround. Someone was saying
that I'd have to edit Math-BigInt-Pari myself, or wait 2 weeks!

By the way, using Pari underneath Math-BigInt is giving me a
several-order-of-magnitude speed increase. In the Pure-Perl case, computing
a 2048 bit Diffie-Hellman secret took over 1 minute; with (lib => 'Pari'), it
takes less than 1 second.

I'm prototyping the use of Perl(cgi) as a DH key server to the
Win32 CryptoAPI. With this speed increase I'm now planning on
finishing the prototyping, instead of giving up (in vain).

regards,
Mark

The M::BI: author (Tels), who is also currently maintaining M::BI,
might be able to provide some insight if you don't get a satisfactory
answer here. (He should at least be informed of the problem.)

I saw Tels post today in the ASPN Mailing list. Thanks to Tels for
confirmation the workaround of using an earlier version (<1.70)

 

[Sisyphus]

By the way, using Pari underneath Math-BigInt is giving me a
several-order-of-magnitude speed increase. In the Pure-Perl case, computing
a 2048 bit Diffie-Hellman secret took over 1 minute; with (lib => 'Pari'), it
takes less than 1 second.

Yes, pure-perl doesn't deal so well with modular exponentiation - which
I would think is what eats up the time in your Diffie-Hellman computations.

I generally find that using GMP instead of Pari provides a further
(significant) speed increase. In the script below, I find GMP to be 20
times faster than Pari, but results vary depending upon the actual size
of the operands and the types of computations being conducted.

----------------------------------
use Math::BigInt;
#use Math::BigInt lib => 'GMP'; # Plug M::BI into M::GMP
#use Math::BigInt lib => 'Pari'; # Plug M::BI into M:ari

use warnings;
use Benchmark;

$n = Math::BigInt->new('3' x 400);
$e = Math::BigInt->new('7' x 400);
$m = $e - 10;

timethese(1, {
'bench' => '$n->bmodpow($e, $m);',
});

print $n, "\n";
----------------------------------

If you want to try GMP on Win32, you can get the necesssary GMP dll from:
http://www.kalinabears.com.au/w32perl/GMP-4.1.3-msvc.tar.gz

Just extract 'libgmp-3.dll' from the tarball and place it somewhere in
your path. You won't need any of the other files contained in the tarball.

There's also Win32 binaries of Math::BigInt::GMP available.

Download for perl 5.6:
http://www.kalinabears.com.au/w32perl/Math-BigInt-GMP-56.tar.gz

Download for perl 5.8:
http://www.kalinabears.com.au/w32perl/Math-BigInt-GMP-58.tar.gz

Just install as per instructions in the Readme.

I've only just recently updated those files, and haven't yet checked to
see if anything has been broken in the process
If anyone does find there's something that's broke, please let me know.

Cheers,
Rob