Python Driver for GGNFS and MSIEVE

[ThomRuley]

Originally Posted by Mini-Geek

E.g. save the poly in example.poly and run "factmsieve.py example".

I'm in the poly file. Where should I put the polynomial? I have something that looks like this:

n: 1274763026968934326151324118033407286595278689008840450094212736567521356866670969828447383237740497765219290449667250482736883
skew: 37753.92
# norm 7.62e+17
c5: 98280
c4: 199470847944
c3: -19048750405790957
c2: -552833905381153759066
c1: -3965419990275258864286632
c0: 242806764000963487621896000
# alpha -6.41
Y1: 2354367642517
Y0: -1669523796316893484875767
# Murphy_E 1.18e-10
# M 87512351163596190002772862008417178331014973225376024777088517549445209342610626721037744367487909704796538658205186796915468
type: gnfs
rlim: 6100000
alim: 6100000
lpbr: 28
lpba: 28
mfbr: 55
mfba: 55
rlambda: 2.5
alambda: 2.5
qintsize: 100000

The polynomial that was recommended was (P743-1)*. How would I enter that?

[mdettweiler]

Originally Posted by ThomRuley

I'm in the poly file. Where should I put the polynomial? I have something that looks like this:



The polynomial that was recommended was (P743-1)*. How would I enter that?

What you put there in the quote box is your polynomial--you're all set. Just plop that in a text file "example.poly" (if you're using notepad, make sure to put quotes around the file name so it doesn't save it as example.poly.txt) in your GGNFS folder, then navigate to that directory with a command prompt and run:

factmsieve.py example

as Mini-Geek said. Note that if you're on Linux, you'll have to make that "python factmsieve.py example" instead.

[axn]

Originally Posted by ThomRuley

I'm in the poly file. Where should I put the polynomial? I have something that looks like this:



The polynomial that was recommended was (P743-1)*. How would I enter that?

As it so happens, I have a better poly for your number

n: 1274763026968934326151324118033407286595278689008840450094212736567521356866670969828447383237740497765219290449667250482736883
skew: 1242974.02
c5:  1560
c4:  430237510
c3: -16372306488770745
c2: -368655929109599815398
c1:  5504316802175350943141886440
c0:  161908777538241104503791860967792
Y1:  33744527270039
Y0: -3823497061361310432906205
type: gnfs
rlim: 6100000
alim: 6100000
lpbr: 28
lpba: 28
mfbr: 55
mfba: 55
rlambda: 2.5
alambda: 2.5
qintsize: 100000

[ThomRuley]

Originally Posted by axn

As it so happens, I have a better poly for your number

n: 1274763026968934326151324118033407286595278689008840450094212736567521356866670969828447383237740497765219290449667250482736883
skew: 1242974.02
c5:  1560
c4:  430237510
c3: -16372306488770745
c2: -368655929109599815398
c1:  5504316802175350943141886440
c0:  161908777538241104503791860967792
Y1:  33744527270039
Y0: -3823497061361310432906205
type: gnfs
rlim: 6100000
alim: 6100000
lpbr: 28
lpba: 28
mfbr: 55
mfba: 55
rlambda: 2.5
alambda: 2.5
qintsize: 100000

Thank you, but I am curious how you got this.

[axn]

Originally Posted by ThomRuley

Thank you, but I am curious how you got this.

I was testing msieve gpu version (http://sourceforge.net/projects/msieve/files/) with some of the odd perfect gnfs candidates (http://oddperfect.org/composites.html)

It took me about an hour on a GT 240 to find this poly.

[WraithX]

Brian, I've been running many snfs factoring jobs with version 0.67 of your script. I've been playing around with what percentage of relations to look for before trying to do post-processing. I've set some digit levels to try post-processing after finding 80% of your estimated minimum, and I've tried some after 70% (about to try 60% on my next digit level). Some went straight to post-processing, some had to keep sieving. I'd like to give you some of this information, but I don't know how much info you'd like, or what specific info you would like. Would you like digit levels (like snfs-120, snfs-130), what percentages worked, maybe the last percentage to not work, the actual numbers I tried to factor, something else I can't think of right now? Some digit levels I only factored a dozen numbers, other digit levels I factored several dozen numbers. Let me know if you'd like this info, and what info you'd like, and I can write a little summary text file.

[Brian Gladman]

Originally Posted by WraithX

Brian, I've been running many snfs factoring jobs with version 0.67 of your script. I've been playing around with what percentage of relations to look for before trying to do post-processing. I've set some digit levels to try post-processing after finding 80% of your estimated minimum, and I've tried some after 70% (about to try 60% on my next digit level). Some went straight to post-processing, some had to keep sieving. I'd like to give you some of this information, but I don't know how much info you'd like, or what specific info you would like. Would you like digit levels (like snfs-120, snfs-130), what percentages worked, maybe the last percentage to not work, the actual numbers I tried to factor, something else I can't think of right now? Some digit levels I only factored a dozen numbers, other digit levels I factored several dozen numbers. Let me know if you'd like this info, and what info you'd like, and I can write a little summary text file.

Hi WraithX,

In fact I don't want much at all - all I need is a formula in the form:

SNFS_rels_needed = function(p1, p2, ...)

where the parameters p1, p2, ... are specified inputs to the factoring task that are in a decreasing order of significance in respect of their impact on the minrels value.

Seriously, I need to know what input parameters should be used to determine the minrels value and how these values participate in setting the value.

I doubt that I can do much with raw data as I don't have the specialist knowledge needed to convert this into plausible formulae for the minrels value. But others here may be in a position to do this so publsihing your data may still be very useful.

best regards,

Brian

[xilman]

Not sure whether this is a FAQ. My rather cursory search for it didn't turn up anything useful.

On a Windows 7 machine I created foo.poly with valid data (it was modified from a previous factorization) and kicked off the Python script. When the time came to attempt filtering and linear algebra it couldn't find either foo.ini or foo.fb files. Any idea why they had not been created?

Factorizations generally go to completion but this isn't the first occasion I've seen this failure mode.

Creating those files (again by modification of those from other factorizations) generally lets the factorization complete. On this occasion, msieve -nc1 generates a "Return value 0. Terminating..." message.

Any suggestions on what to try next, other than to move everything to a Linux box and use the Perl script?


Paul

[Brian Gladman]

Originally Posted by xilman

Not sure whether this is a FAQ. My rather cursory search for it didn't turn up anything useful.

On a Windows 7 machine I created foo.poly with valid data (it was modified from a previous factorization) and kicked off the Python script. When the time came to attempt filtering and linear algebra it couldn't find either foo.ini or foo.fb files. Any idea why they had not been created?

Factorizations generally go to completion but this isn't the first occasion I've seen this failure mode.

Creating those files (again by modification of those from other factorizations) generally lets the factorization complete. On this occasion, msieve -nc1 generates a "Return value 0. Terminating..." message.

Any suggestions on what to try next, other than to move everything to a Linux box and use the Perl script?

Hi Paul,

The log file will hopefully give some clues - have you looked at it? If it is not too big, can we see it?

Brian

[EdH]

I'm reaching, here, so bear with me. . .

Somewhere within the first few times I ran the script, it skipped creating at least one file (test.ini). Since then all has seemed fine.

Is there a possibility that the creation of any of the files is somehow dependent on their already existing? IOW, are any of the files conditionally created, or appended? Or, is it possible that the creation can be skipped over by condition?

[xilman]

Originally Posted by Brian Gladman

Hi Paul,

The log file will hopefully give some clues - have you looked at it? If it is not too big, can we see it?

Brian

It gave a clue about why the "Return value 0. Terminating..." message appeared. I'd made a typo in the .FB file and it didn't match the .POLY file. Now fixed and filtering is proceeding.

No idea why the files weren't created in the first place. The log file is attached; you can probably interpret its contents much better than I can. Note that I used "foo" in my earlier post partly to save typing and partly to indicate that it's a generic problem, though not one which always occurs.

Paul