Team sieve #18: c171 from 4788:2527

[10metreh]

Post-processing will be done by FactorEyes

31/3: reservations closed
7/4: Linear algebra has begun, factors due 13/4
13/4: Factorization complete, c171 = p56.p115

Poly:

n: 193180597261434437130723427223452983749001196443861486431050101233095959046579376056784397572419692971642911818089261513816572456073452501892312801518858596361529181158803
# norm 1.035357e-16 alpha -6.929733 e 2.847e-13
skew: 42972295.61
c0: -53546854924607693903542081910433567264000
c1:  522292118959486321900977916817558660
c2:  37609110301045736156878149024
c3: -940271047170212912673
c4: -17420632831592
c5:  122160
Y0: -1095990843863198982888397501015033
Y1:  2645241704855349311
rlim: 67108863
alim: 67108863
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
rlambda: 2.6
alambda: 2.6

Use siever 15e, sieve 20M-120M Q. Siever 15e guzzles much more memory than 14e, so only use it if you have more than 512MB RAM. (Most machines do these days.)

Command line:

gnfs-lasieve4I15e -a 4788_2527.poly -o <output_file_name> -f <start_of_range> -c <length_of_range>

where 4788_2527.poly is the poly file above.

Reservations:

  Range      Who
*  20M- 21M  Andi47 (done, 1237296 relations)
*  21M- 40M  FactorEyes (done, 25225898 relations)
*  40M- 41M  Greebley (done, 1369715 relations)
*  41M- 50M  bsquared (done, 12434568 relations)
*  50M- 51M  Andi47 (done, 1377211 relations)
*  51M- 60M  bsquared (done, 12515899 relations)
*  60M- 84M  FactorEyes (done, 32347496 relations)
*  84M- 85M  Greebley (done, 1274669 relations)
*  85M- 86M  Andi47 (done, 1262309 relations)
*  86M-100M  FactorEyes (done, 17187765 relations)
* 100M-110M  fivemack (done, 11092013 relations)
* 110M-115M  FactorEyes (done, 5692052 relations)
* 115M-120M  bsquared (done, 5566276 relations)

Total relations received: 128583167 (97752679 unique)

Use a site such as Sendspace to upload relations.
Please do not use Rapidshare if possible as it has time consuming restrictions.

[Andi47]

reserving 20-21M

[FactorEyes]

I'll do 21M-40M.

[Greebley]

I will take 40-41.

I estimate it will take 2 weeks for me to do 1 million relations with one processor working on it.

[FactorEyes]

Originally Posted by Greebley

I estimate it will take 2 weeks for me to do 1 million relations with one processor working on it.

You'll beat the crowd by a mile, then.

The C163 we just did took almost a month to sieve. I'll be astounded if this one finishes within 6 weeks.

[jasonp]

That's a really nice poly! Did you use pol51 or msieve?

[Batalov]

This poly was found by JRK with msieve/GPU. It is truly nice. It has a combined score that befits an average c169.

Not every number (or every search range) is so lucky...

[henryzz]

Originally Posted by Batalov

It has a combined score that befits an average c169.

What simliarities does this make this factorization have with an average c169?
Should the params be based on c169 params? Will it sieve like a c169? I realize that multiprecision arithmatic might be a little slower as it is modulo a c171 not a c169.

[FactorEyes]

Originally Posted by henryzz

Should the params be based on c169 params? Will it sieve like a c169? I realize that multiprecision arithmatic might be a little slower as it is modulo a c171 not a c169.

The lattice sieving is almost entirely low-precision, except for perhaps for some modular inverses, so little to no difference there.

Even for the small fraction of calculations which actually occur at the full precision of the target composite, a C169 is probably the same number of limbs as a C171.

[jrk]

Originally Posted by jasonp

That's a really nice poly! Did you use pol51 or msieve?

The poly was found with msieve GPU. But with different norm targets than the main msieve uses.

My local msieve changes this:

	{170, 5.00E+025, 1.58E+024, 1.20E-013},
	{175, 3.00E+026, 1.00E+025, 6.00E-014},

To:

	{170, 5.00E+025, 1.58E+024, 1.50E-013},
	{175, 2.00E+026, 9.00E+024, 6.40E-014},

For a c171 though the changes are very small. I was also toying with the leading rational coefficient factors which is of no consequence to the overall search efficiency.

For this number, I figured a good expectation from a GPU search is about 2.65e-13, which is about what frmky found. But 2.847e-13 should probably not be realistically expected in another search.

[henryzz]

Originally Posted by FactorEyes

The lattice sieving is almost entirely low-precision, except for perhaps for some modular inverses, so little to no difference there.

Even for the small fraction of calculations which actually occur at the full precision of the target composite, a C169 is probably the same number of limbs as a C171.

I thought that might be the case.