fun with snfs - Page 3

masser · 2007-11-09, 15:57

The candidates with n between 200 and 400 have only been ECM'd up to 35 digits. It might make sense to do more ECM before factoring them... just a warning.

michaf · 2007-11-10, 18:10

hmm... this one bothers me...
after four times the same return value (65280) with ggnfs

I haven't tried msieve yet, no idea if (and how) you can put the poly values in there yet. I will read into that first :)

In the meantime, does anyone know what is the problem here? :

Quote:

-> Found 12492 relation-sets versus minFF=159533.
-> More sieving needed.
-> Q0=625001, QSTEP=75000.
-> makeJobFile(): q0=625000, q1=700000.
-> makeJobFile(): Adjusted to q0=625001, q1=700000.
-> Lattice sieving q-values from q=625001 to 700000.
=> "../ggnfs/ggnfs/src/gnfs-lasieve4I12e.exe" -k -o spairs.out -v -n0 -a 24032_5_181p1.job
-> Return value 65280. Updating job file and terminating...
=>"c:/mingw/msys/1.0/bin/cat.exe" spairs.out >> spairs.add
-> makeJobFile(): q0=625000, q1=700000.
-> makeJobFile(): Adjusted to q0=625001, q1=700000.
Terminating...

smh · 2007-11-10, 20:43

Quote:

Originally Posted by michaf

hmm... this one bothers me...
after four times the same return value (65280) with ggnfs

I haven't tried msieve yet, no idea if (and how) you can put the poly values in there yet. I will read into that first :)

In the meantime, does anyone know what is the problem here? :

Don't know what's wrong here but i've seen this before. Just increase the special-q a couple of thousand and continue sieving.

frmky · 2007-11-11, 03:44

If you compiled your binaries from source, try removing -fomit-frame-pointer from CFLAGS in src/Makefile and recompiling. I'm not sure of the error number, but I was getting occasional crashes in the lattice sieve until I removed that compiler option.

Greg

konrad127123 · 2007-11-11, 22:19

53546*5^200-1: I had a load of errors again, so I tried commenting out the code where it detected the error, and ended up with a valid partial factorisation!:
[11/11 09:39:34] From dependence 24, sqrt obtained:
[11/11 09:39:34] r1=9371472846912361231466755306955909048138768280468851408539860585052203785
619765448104405489006204468163 (c103)
[11/11 09:39:34] r2=355565855561475148370692149966677312439523 (pp42)

Based on the time this took, at this level I should ecm part, but not all of the way up to 40 digits.

Reserving: 325846*5^201-1

mdettweiler · 2007-11-12, 00:21

Quote:

Originally Posted by konrad127123

53546*5^200-1: I had a load of errors again, so I tried commenting out the code where it detected the error, and ended up with a valid partial factorisation!:
[11/11 09:39:34] From dependence 24, sqrt obtained:
[11/11 09:39:34] r1=9371472846912361231466755306955909048138768280468851408
539860585052203785619765448104405489006204468163 (c103)
[11/11 09:39:34] r2=355565855561475148370692149966677312439523 (pp42)

Based on the time this took, at this level I should ecm part, but not all of the way up to 40 digits.

Reserving: 325846*5^201-1

I think for the purposes of this project, all we need is one factor, so the one factor you got (I'm assuming the pp42 listed there actually is a factor, and not something else) would be enough to whack this candidate out of the sieve file, I would think.

masser · 2008-02-01, 21:17

Any word on this final reservation with n < 200?

Code:

24032*5^181+1 reserved      (michaf)

michaf · 2008-02-02, 08:24

Quote:

Originally Posted by masser

Any word on this final reservation with n < 200?

Code:

24032*5^181+1 reserved      (michaf)

My apologies,

I've tried to get it fed to ggnfs withg snfs, but didn't succeed in the end. It just refused to get sieving. (iirc, something about a very bad poly)

If someone else is able to do it the right way, go ahead :)

konrad127123 · 2008-02-02, 13:27

I didn't manage to get 325846*5^201-1 to work, so I'm unreserving it.

fivemack · 2008-02-03, 01:10

This is interesting; you've found a new bug in the Franke siever, thanks! My original thought was that 24032*5^181+1 was behaving oddly when sieved with a quintic since 24032 is divisible by 32, and you might want to write it as 3755*(2*5^36)^5+1, but that doesn't seem to work either.

I'm trying jasonp's new implementation of line sieving on 24032*5^181+1, which is running slowly (~10 relations per second when I'd expect ~100 from the lattice sievers, but that's not accounting for the batch-factorisation step which probably adds another factor two to the yield) but does at least seem to be finding relations. If that works reasonably I'll try the same approach on 325846*5^201-1

michaf · 2008-02-03, 10:52

ah, that good (or bad, depending on the way you see things)
It never even occurred to me that it might NOT be me...

Do I need to report it anywhere, or will you take the upon you, to avoid future misses?

Quote:

Originally Posted by fivemack

This is interesting; you've found a new bug in the Franke siever, thanks! My original thought was that 24032*5^181+1 was behaving oddly when sieved with a quintic since 24032 is divisible by 32, and you might want to write it as 3755*(2*5^36)^5+1, but that doesn't seem to work either.

I'm trying jasonp's new implementation of line sieving on 24032*5^181+1, which is running slowly (~10 relations per second when I'd expect ~100 from the lattice sievers, but that's not accounting for the batch-factorisation step which probably adds another factor two to the yield) but does at least seem to be finding relations. If that works reasonably I'll try the same approach on 325846*5^201-1

2007-11-09, 15:57	#23
masser Jul 2003 wear a mask 2×829 Posts	200<=n<400 candidates The candidates with n between 200 and 400 have only been ECM'd up to 35 digits. It might make sense to do more ECM before factoring them... just a warning.

2008-02-01, 21:17	#29
masser Jul 2003 wear a mask 2·829 Posts	Any word on this final reservation with n < 200? Code: 24032*5^181+1 reserved (michaf)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
SNFS targets which need more ECM	XYYXF	XYYXF Project	57	2017-07-04 19:15
SNFS(27x) How much ECM before switching to NFS?	YuL	Factoring	24	2015-06-08 01:26
SNFS polynomials for k*b^n+-1	mdettweiler	Factoring	15	2010-01-14 21:13
SNFS Sample(10^4+1)	nuggetprime	Factoring	1	2007-06-11 16:31
Completed my first SNFS job	schickel	Factoring	0	2007-05-27 05:42

2007-11-11, 03:44	#26
frmky Jul 2003 So Cal 2·3⁴·13 Posts	If you compiled your binaries from source, try removing -fomit-frame-pointer from CFLAGS in src/Makefile and recompiling. I'm not sure of the error number, but I was getting occasional crashes in the lattice sieve until I removed that compiler option. Greg

2007-11-11, 22:19	#27
konrad127123 Jun 2005 21₁₆ Posts	535465^200-1: I had a load of errors again, so I tried commenting out the code where it detected the error, and ended up with a valid partial factorisation!: [11/11 09:39:34] From dependence 24, sqrt obtained: [11/11 09:39:34] r1=9371472846912361231466755306955909048138768280468851408539860585052203785 619765448104405489006204468163 (c103) [11/11 09:39:34] r2=355565855561475148370692149966677312439523 (pp42) Based on the time this took, at this level I should ecm part, but not all of the way up to 40 digits. Reserving: 3258465^201-1

2008-02-02, 13:27	#31
konrad127123 Jun 2005 3·11 Posts	I didn't manage to get 325846*5^201-1 to work, so I'm unreserving it.

2008-02-03, 01:10	#32
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 7²×131 Posts	This is interesting; you've found a new bug in the Franke siever, thanks! My original thought was that 240325^181+1 was behaving oddly when sieved with a quintic since 24032 is divisible by 32, and you might want to write it as 3755(25^36)^5+1, but that doesn't seem to work either. I'm trying jasonp's new implementation of line sieving on 240325^181+1, which is running slowly (~10 relations per second when I'd expect ~100 from the lattice sievers, but that's not accounting for the batch-factorisation step which probably adds another factor two to the yield) but does at least seem to be finding relations. If that works reasonably I'll try the same approach on 325846*5^201-1