20210405, 08:49  #12  
Apr 2021
17 Posts 
Quote:
Last fiddled with by Unitome on 20210405 at 08:55 

20210405, 09:02  #13 
Romulan Interpreter
Jun 2011
Thailand
10010011100000_{2} Posts 
Well, that's kind of common sense, you will run NFS only after you picked all "low hanging fruits", i.e. extracted all small factors you could. Otherwise it makes no sense to waste the time doing NFS, and the algorithms assume that at least some basic TF was done. The RSA numbers are by definition "hard", they don't have small factors. But if 3 is not a factor, it means that the number is either 1 or 2 (mod 3), so changing a digit from 1 to 2 (remember the school rule of divisibility to 3? sum the digits?) will add 1 to the modulus, you get respective 2 or 0 (mod 3), and so on. So you have a 50% chance that your "change" introduced a small prime factor 3 (and add more for 7, 11, etc, or even 5, if you did that for the last digit), causing NFS to go on the weeds.
Last fiddled with by LaurV on 20210405 at 09:05 
20210406, 00:43  #14  
Apr 2021
11_{16} Posts 
Quote:
Last fiddled with by Unitome on 20210406 at 00:54 

20210406, 01:12  #15 
"Curtis"
Feb 2005
Riverside, CA
2×3^{2}×5×53 Posts 
GGNFS is exactly and only the siever program that finds relations.
It is not, and never was, a package to run an entire NFS job. If you want something foolproof, use YAFU. In fact, beginners should use it anyway, and only try what you're doing once they have enough experience to think they can choose better settings themselves to get faster factorization jobs or, those who prefer to run individual steps manually (such as, ahem, ECM). 
20210406, 08:28  #16  
Apr 2021
17 Posts 
Quote:
Last fiddled with by Unitome on 20210406 at 08:52 

20210406, 08:40  #17 
Just call me Henry
"David"
Sep 2007
Cambridge (GMT/BST)
2^{2}×3^{2}×163 Posts 
Once upon a time GGNFS was a collection of programs that would do a complete factorization itself using the factLat.pl script to pull them together. Many of these tools haven't been used in years for anything meaningful. I am not sure which programs were written from scratch for ggnfs. The siever wasn't and I suspect the polynomial selection wasn't(pol5) although it had an alternative that probably was.

20210406, 11:57  #18 
"Ed Hall"
Dec 2009
Adirondack Mtns
3713_{10} Posts 
You can do just the ECM phase with YAFU by using ECM(). You can run ECM directly, but it will run singlethreaded. You can also use ecm.py to multithread ECM. Check near the last post for the latest revision.
YAFU will step through several B1 values, while the others will perform your given number of curves on your given B1 [B2]. 
20210406, 15:42  #19  
Sep 2009
3767_{8} Posts 
Quote:
Chris 

20210407, 08:03  #20  
Apr 2021
17 Posts 
Quote:
Also just for any newbies who read this someday, I tested YAFU and GGNFS and GGNFS is much faster than YAFU for numbers that GGNFS can factor. Roughly 40% faster on 100 digit numbers. Last fiddled with by Unitome on 20210407 at 08:08 

20210407, 14:24  #21  
"Curtis"
Feb 2005
Riverside, CA
2·3^{2}·5·53 Posts 
Quote:
Did YAFU use quadratic sieve, instead of NFS? Perhaps you don't have YAFU "tuned" yet for your hardware. Sorry that I don't recall how to do so. 

20210407, 15:40  #22  
Sep 2009
2,039 Posts 
Quote:
Another thing worth doing is to read all the INSTALL and Readme files bundled with ggnfs and msieve. Then read them again (it will take several readings for it to all sink in). And every year or so reread them for bits you didn't quite understand last time. The same is true of the readme etc that come with ECM. And the README and docfile bundled with YAFU. Chris 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
factoring 2ⁿ2 equivalent to factoring 2ⁿ1(I think)  baih  Miscellaneous Math  9  20200921 07:11 
OpenCL GPU P1 Factoring and ECM Factoring  xx005fs  GPU Computing  3  20181027 14:49 