Running Msieve

[Phil MjX]

Hi jasonp !

I'll quote the msieve faq :

Q. Why did you stop at 125 digits? I want to factor much bigger numbers.

Does this limit still holds for the nfs part of the code?
I am currently sieving a 128 digits composite with ggnfs and would be interested in the msieve classical sieve...

Thanks.
Philippe.

[jasonp]

Originally Posted by smh

I'm also interested in these instructions. Can you PM them to me too?

Maybe I should treat the people in this forum like adults. In fact, maybe I should treat all msieve users like adults. The next version will have a 155 digit limit. I'll only hear back once from users crazy enough to push that limit :)

Note that when the postprocessing can actually work, similar instructions will allow a GGNFS factorization to be completed using msieve. It would be interesting to compare the matrices generated by the two packages.

jasonp

------------------

Assuming you have a binary compiled with a larger limit on the input size (I can provide one if this is a problem), to run the line siever you just need to provide a factor base file. This is a text file with the following lines:

N <the number you are factoring>
R0 <low-order coefficient of rational polynomial>
R1 <high-order coefficient of rational polynomial>
A0 <low-order coefficient of algebraic polynomial>
A1 ...
A2 ...
A3 ...
A4 ...
A5 <high-order coefficient of algebraic polynomial>

Running

msieve -i <file_with_number> -nf <factor_base_file> -ns X,Y

(with X and Y the lower and upper limits of sieve lines) will then run the line siever. It will generate the rest of the factor base and sieving parameters, then dump any relations found to msieve.dat

Note that the first line of msieve.dat does not contain a relation, so hopefully this won't give procrels any trouble.

[smh]

Originally Posted by jasonp

Running

msieve -i <file_with_number> -nf <factor_base_file> -ns X,Y

(with X and Y the lower and upper limits of sieve lines) will then run the line siever.

Running msieve is the easy part, but what are usefull values for X,Y for factorizations at different sizes?

[frmky]

Thanks Jason for all the work you're putting into this! I have two questions. Does msieve support 4th or 6th order algebraic polynomials as well, or only 5th order? And is the 155 digit limit a limit of the MP code or could it be used with larger SNFS factorizations?

Thanks!
Greg

[jasonp]

smh,

I wouldn't use a range larger than [1,4000]. The line size will always be the same, but it should really be varied, so that smaller values get a larger line size

George,

I don't want to get to a lattice siever (or any other optimization) until the library can complete an end-to-end GNFS factorization. It's the first priority when I get to that point though

frmky,

Incrementing MAX_POLY_DEGREE should allow the code to handle 6th degree polynomials. It can only *generate* 5th degree, though; the whole poly generator is going to get overhauled someday, and I didn't see the point of bothering with 4th-degree polynomials until that has happened.

Thanks for the support everybody; if experience is any guide, the first few factorizations done with the new library are going to be very dicey, so please be patient if you want to help out.

jasonp

[RedGolpe]

I assume you can use the 1.10 msieve.dat file: I tried that and got no errors. Is that correct?

[jasonp]

Originally Posted by RedGolpe

I assume you can use the 1.10 msieve.dat file: I tried that and got no errors. Is that correct?

Correct; the only time that's not allowed is when changes happen to the QS multiplier selection. I think that multiplier selection has worked correctly since 1.10, but I also thought that the last four times that changes were needed :)

jasonp

[bigmic]

Originally Posted by smh

Thats 11 to 12 days for a c100.

I started a c102 around the same time, but canceled the run this morning.
After sieving a bit over 600000 lines i had about 4.09M relations. (all data still available).

I did ~10 minutes of polynomial search using GGNFS and restarted msieve with that polynomial.

Results look much more promising.
After 72 minutes: b = 1804, found 737262 relations

I'll post results if the factorization completes.

ok - perhaps i am too stupid - but i dont get the ggnfs polynomials converted to msieve - in my easy thinking i just put the A0-5 and R0,R1 and also N to a file (using the coefficients from ggnfs polynomial) i was rewarded with an segfault - any help how to do it correctly ?

may thanks !

[jasonp]

Originally Posted by bigmic

ok - perhaps i am too stupid - but i dont get the ggnfs polynomials converted to msieve - in my easy thinking i just put the A0-5 and R0,R1 and also N to a file (using the coefficients from ggnfs polynomial) i was rewarded with an segfault - any help how to do it correctly ?

Did you concatenate all of your relation files into a single huge file, with 'N <your_number>' as the first line in that file? If so, the command line should be

msieve -v -i <file with input number> -nf <file with polynomial> -s <huge savefile> -nc

If it still crashes, and takes a while to do so, then it's probably a bug somewhere (according to your post in the GGNFS mailing list, you're trying to boil down 40 million relations, or 5x more than anyone else has tried so far). If this is on a windows system, and the file is larger than 2GB or 4GB, there could be trouble (though the library only makes sequential passes through the file, on purpose).

jasonp

[bigmic]

hi jason, thanks for answering.

my first attempt was only the goal to use the polynomial i found with ggnfs and start sieving with msieve.

a 2nd goal would be trying to use the old relations (this was not what i was trying atm)

i will try in the evening as you pointed out .

Thanks for the help !

greets

Update:

i tried it as you said but same segfault
msieve -v -i <file with input number> -nf <file with polynomial> -s <huge savefile> -nc

size score = 0.000000e+00, Murphy alpha = 0.000000, combined = 0.000000e+00
generating factor base
Segmentation fault

in the <file with polynomial> i put

N 12345 (mynumber)
R0 12345
R1 12345
A0 12345
A1 12345
..
A5 12345

with R0 == Y0 and R1 == Y1 from ggnfs
and mapping Ax to cX from ggnfs

is this ok ?

[jasonp]

Originally Posted by bigmic

size score = 0.000000e+00, Murphy alpha = 0.000000, combined = 0.000000e+00
generating factor base
Segmentation fault

'-nc' will not do any sieving, '-ns' will. Run with '-h' to get a complete list of options. It appears that your polynomials are not getting read in, so this could be a library bug. Would you be willing to PM your file? (Shuffle the digits around if you want, I know better than to try such big jobs myself :)

jasonp