[QUOTE=CedricVonck]What I like to is that the program can sieve a > 150 digit number in more doable time (like say 10 hours) in a distributed approach.[/QUOTE]
In which case, I hope you have a [i]very[/i] big cluster available to do the sieving.

The world record MPQS factorization, as far as I know, is of a c135. It took us six [i]months[/i] of sieving on dozens of machines. We did it only to explore the benefits, if any, of using three large primes in relations. The conclusion is that although 3LP speeded the computation, GNFS would have been several times faster still.

Please give up your quest for MPQS to be used on large integers. It's an excellent choice for factorizations up to 100 digits or so, but a very sub-optimal one for numbers over 120 digits, or even less.


Paul

[QUOTE=CedricVonck]What I like to is that the program can sieve a > 150 digit number in more doable time (like say 10 hours) in a distributed approach.[/QUOTE]
I'd like that too. But GNFS with a bunch of machines takes 1-2 months to do the same job, with today's hardware.

jasonp

Ok sorry guys for my stUpid questions...

[QUOTE=CedricVonck]Ok sorry guys for my stUpid questions...[/QUOTE]
It's not easy at all to get an idea of the scale of work you have to do for a really big factorization, unless you're sufficiently familiar with the methods involved to estimate it in theory. If it's any consolation, the TWIRL paper gave sample NFS parameters for factoring RSA-1024, and I had no earthly idea it was so much work. I probably still wouldn't have an idea if I hadn't played around with NFS sieving.

jasonp

msieve 1.0 beta
 

I have a beta available of the next msieve version. This is more a bug fix release than anything else, and I'm sufficiently happy with the state of the code to declare version 1.0

[url]www.boo.net/~jasonp/msieve100b1.tar.gz[/url]
[url]www.boo.net/~jasonp/msieve100b1.exe[/url]

Be sure to read the changelog and make sure your bug report has been addressed. Highlights of the new release include

- The factorization limit has been raised to 120 digits. I've done this becase several people have asked for more digits, *not* because it's a good idea. I've also added extra tuning for larger factorizations, though none of it is tested

- The binary can now be given a time limit to sieve. This should make it easier to schedule things to run only overnight, for example.

Big factorizations (100+ digits) may be faster, but I don't know by how much (if at all). In any case, no formats have changed, so the new binary can be used immediately.

This will probably be the last release. Thanks to everybody who put up with all the trouble early on, and thanks for all the kind words and encouragement. This hobby stuff turned out much better because of it.

jasonp

My computer will complete the C114 with msieve 0.88 nearly in five days :)

[QUOTE=XYYXF]My computer will complete [B]the C114[/B] with msieve 0.88 nearly in five days :)[/QUOTE]

What c114 are you talking about?

Msieve 1.0 released
 

Now available at [url]www.boo.net/~jasonp/qs.html[/url]

Changes from the beta:

- upped the limit to 125 digits. The extra 5 digits are free given the size hardwired into the multiple precision library, so there's no reason to disallow them (no reason to use them either :)

- added a readme with instructions, credits and paper references that nobody here needs anymore.

On to bigger and better things, once I figure out what they are.

jasonp

[QUOTE=jasonp]
On to bigger and better things, once I figure out what they are.

jasonp[/QUOTE]

Thanks Jason, I just can't wait! :rolleyes:

Luigi

It's a thing of beauty... it's really come a long way in the past 8 months. Thanks for the great work! :w00t:

[code]Number digits msieve-0.7 msieve-1.0
-----------------------------------------
(6^67-1)/5 52 0:00:04 0:00:03
(10^59-1)/9 59 0:00:16 0:00:10
(3^137+1)/4 65 0:00:53 0:00:38
(10^71-1)/9 71 0:02:53 0:02:08
(12^71+1)/13 76 0:17:11 0:08:51
(10^83-1)/9 83 1:14:44 0:39:38[/code]

[QUOTE=trilliwig]It's a thing of beauty...[/QUOTE]You want improvement?

[code]Number digits msieve-0.1 msieve-1.0
-----------------------------------------
(6^67-1)/5 52 0:01:36 0:00:05
(10^59-1)/9 59 0:09:21 0:00:19
[/code]
I didn't have the heart to try more :smile:

jasonp