Factoring for a publication

[swishzzz]

Originally Posted by Max0526

c198 spun poly vs original polys. Because the two spun polys are close, and coming from different connected elliptic curves, it may be smart to test-sieve to decide which one to pick.

n: 442480456268307355970377486465969784979730470279409306982158698829691328120816007497057176155135483213248563103903861494465722951888948657031409779919681670455655411204200835023769409621132838744169

SPUN:
Line 4 : {'c4': '1', 'c3': '0', 'c2': '2', 'c1': '-12', 'c0': '10', 'y1': '3105317578628153423506739578983437982692791417722955507', 'y0': '1783785732675759131389551214804762300612349559114245473'}
Skew, Murphy E:
1.16292 2.41372557e-13

Line 25 : {'c4': '2', 'c3': '2', 'c2': '1', 'c1': '4', 'c0': '8', 'y1': '1195009498251726830929240449280801910673701770687009936', 'y0': '-6033442123246363595737194920672615397495021982445825621'}
Skew, Murphy E:
2.2025 2.37771575e-13
----------------------
BEST ORIGINAL:
Line 2 : {'c4': '2', 'c3': '-30', 'c2': '169', 'c1': '-420', 'c0': '392', 'y1': '1195009498251726830929240449280801910673701770687009936', 'y0': '-10813480116253270919454156717795823040189829065193865365'}
Skew, Murphy E:
4.9201 2.03968529e-13

Line 7 : {'c4': '1', 'c3': '-24', 'c2': '288', 'c1': '-4320', 'c0': '32400', 'y1': '220255307658732382019531394029779280346740309768118339', 'y0': '-15526587893140767117533697894917189913463957088614777535'}
Skew, Murphy E:
20.63567 1.77637452e-13

bsquared, are they detailed enough for you, or you need me to recreate them in a standard column format? Here the coefficients are c4,c3,c2,c1,c0,Y1,Y0; n is at the top.

If needed, I've modified the script slightly to print out the actual polys in a column format. Example output:

Line 1 : {'c4': '1', 'c3': '0', 'c2': '2', 'c1': '-12', 'c0': '10', 'Y1': '3105317578628153423506739578983437982692791417722955507', 'Y0': '1783785732675759131389551214804762300612349559114245473'}

n: 442480456268307355970377486465969784979730470279409306982158698829691328120816007497057176155135483213248563103903861494465722951888948657031409779919681670455655411204200835023769409621132838744169
# e = 2.41372557e-13
skew: 1.16292
type: snfs
c4: 1
c3: 0
c2: 2
c1: -12
c0: 10
Y1: 3105317578628153423506739578983437982692791417722955507
Y0: 1783785732675759131389551214804762300612349559114245473

[Max0526]

Originally Posted by swishzzz

If needed, I've modified the script slightly to print out the actual polys in a column format. Example output:

Line 1 : {'c4': '1', 'c3': '0', 'c2': '2', 'c1': '-12', 'c0': '10', 'Y1': '3105317578628153423506739578983437982692791417722955507', 'Y0': '1783785732675759131389551214804762300612349559114245473'}

n: 442480456268307355970377486465969784979730470279409306982158698829691328120816007497057176155135483213248563103903861494465722951888948657031409779919681670455655411204200835023769409621132838744169
# e = 2.41372557e-13
skew: 1.16292
type: snfs
c4: 1
c3: 0
c2: 2
c1: -12
c0: 10
Y1: 3105317578628153423506739578983437982692791417722955507
Y0: 1783785732675759131389551214804762300612349559114245473

swishzzz, you are so awesome!
Thank you so much!

[Max0526]

Originally Posted by bsquared

I'm running a t35 on all of the unbooked composties in unhighlighted rows 158, 160, and 168-172. This is good experience for avx-ecm.

So far (I will update as things progress):

(10,-5)
c168 = p47*c123

(10,-7)
c282 = p26*c256
c288 = p11*c277

(5,-10)
c283 = p27*p32*p34*p35*c157

(4,-10)
c277 (#1) = p19*p17*p28*p34*c181
c277 (#2) = p22*p28*c229

(3,-10)
c265 = p14*prp251

(2,-10)
c231 = p24*c207
c249 = p15*p23*c212

(1,-10)
c184 = p36*c149
c235 = p28*prp207

So far no complete factorizations but (10,-5) now has an easy c123 that I am not reserving.

All updated. Big thank you for the two primes and your perseverance!

[Max0526]

Originally Posted by bsquared

Advancing to t40 on unreserved composites on lines 158,160,163, and 167-172

Progress:

(10,-7)
c256 = p38*c219

(ongoing)

Updated, thank you!

[Max0526]

Originally Posted by swishzzz

Finished light ecm on all these (t25 or so):

(12, 5): c156, c169
(12, 4): c170, c173, c177
(12, 3): c130, c180
(12, 2): c121, c143
(12, 1): c125
(12, 0): c168, c174, c175
(12, -1): c121, c179, c190

Everything is updated in the sheet.
swishzzz, thank you for creating and ECMing these stage 12 numbers!

[bsquared]

Thanks for the spin! I will pick some parameters and test sieve. The number has survived 60k+ curves at B1=260M so it's probably ready for SNFS.

[bur]

Originally Posted by charybdis

bur - watch that "excess" figure which pops up each time filtering runs; it was -1183116 in the log entry that you posted. That number should get smaller and smaller (in absolute value) until it turns positive, at which point filtering should succeed unless required_excess is nonzero. Keeping track of the excess will give you some idea of how much longer you'll need to sieve.

Then hopefully the current sieving will be the last one, -536571 excess.

In that regard, I think at some point both cado and msieve go through several filtering iterations in which the excess steadily increases. But without any additional sieving. What is that about?

[bur]

It's finally at the linear algebra stage, took 31.1e6 relations instead of the initially estimated 17.6e6.

I let cado do the LA, ETA for krylov is 4 hours, so the total will be about 156 hours. Still better than GNFS which would likely haven taken around 200+ hours. But I'm sure glad when this is over. ;)

[Max0526]

Originally Posted by bur

It's finally at the linear algebra stage, took 31.1e6 relations instead of the initially estimated 17.6e6.

I let cado do the LA, ETA for krylov is 4 hours, so the total will be about 156 hours. Still better than GNFS which would likely haven taken around 200+ hours. But I'm sure glad when this is over. ;)

You just got a pretty tough number, bur. It is not shared by the connected elliptic curve, so only 4 (not 8) SNFS polys could be initially created. Out of these 4, the best spun poly has 3.3% higher E than what you started with. frmky says that even +10% may not make a difference in the total job time.
How much disk space, in MB/GB, is taken by the raw relations right now?

[bur]

I'd still like to see this run with the alternative parameters. Or with alternative poly.

The relations are 3.8 GB in gzip'ed form. If you're interested in them, I could upload them somewhere.

[Max0526]

Originally Posted by bur

I'd still like to see this run with the alternative parameters. Or with alternative poly.

The relations are 3.8 GB in gzip'ed form. If you're interested in them, I could upload them somewhere.

I should have asked in the last post, how many gzipped files are there?

When I factored some c168/snfs170 from plot 2, CADO failed on the last step (didn't find a suitable small prime [as it didn't exist at all probably] and did not proceed anyways [as msieve would]). The readme file says it's a known CADO issue with SNFS degree 4 polys with a particular Galois group and SNFS degree 8 polys. All suggested fixes on CADO didn't work. I had to transfer ~9000 gzipped files (just under 1 GB) to Windows, unzip them, create the first in the sequence one-line file with N: ... in it, glue them all together, and feed into msieve as raw relations. msieve was successful in that case and I thought it would always be an acceptable solution. Not true, it was just luck, as was pointed out by EdH: https://www.mersenneforum.org/showpo...0&postcount=94.

If it's not trouble for you, yes, I would like to look at those 3.8 GB (along with the used poly), ideally uploaded in maybe 4 separate chunks, ~1 GB each. I usually use Mega (https://mega.io/) for huge data transfers but anything will work (e.g., Google Drive, etc.) I'll try to take in the relations and build the matrix on msieve. A zipped matrix should also be smaller than 3.8 GB I hope. :-)