Factoring for a publication - Page 23

frmky · 2021-06-12, 19:14

Quote:

Originally Posted by Max0526

This is what we call a nice split! c156=2*p78!

Brilliant even!

EdH · 2021-06-12, 20:18

Quote:

Originally Posted by EdH

I will have to play in that arena later today.

For now, all the smaller composites are factored and the c312 is nearing t50.

My Bash script failed (miserably) to spin the tested SNFS polynomial on one of my machines. I'm trying the same machine on the latest GNFS requested poly (in the msieve thread) to see if that returns anything at least equal to the initial one.

As to the c312, it is at t50. Moving to the c326. . .

firejuggler · 2021-06-12, 20:36

i'm jumping around but
(10,9)c204= 184374369527696042002774666493*P175

Max0526 · 2021-06-12, 22:25

Quote:

Originally Posted by Max0526

This is what we call a nice split! c156=2*p78! Done and done!
Thank you, fivemack!
Could you please send me that c156 GNFS poly? How did you select it? msieve GPU or CADO?
Could you also take a look at https://mersenneforum.org/showpost.p...postcount=228?

EDIT:
fivemack, I hope you haven't started line 88 (5, -8) yet. If you haven't, could you please hold on for a bit, just to give me a chance to look at all available SNFS polys and their spins for that c209? It would be a perfect training ground for the definitely possible (just checked) polys 5-8 and the newly baked spinner.

Done! The spinner helped a lot.
The three best spun polys are below, all above 9e-13 (22% better).

All are in the format c4,c3,c2,c1,c0,Y1,Y0. All need test-sieving of course.
A full list is attached in a text file.

Code:

16, 24, 17, -9, 1, 9307684042037302713836537085227216320008466867809161, -3786905681649011959990757640861865401828382570683684
0.40317 9.62467489e-13

4, 12, 17, -18, 4, 9307684042037302713836537085227216320008466867809161, -7573811363298023919981515281723730803656765141367368
0.80637 9.55967455e-13

1, 6, 17, -36, 16, 9307684042037302713836537085227216320008466867809161, -15147622726596047839963030563447461607313530282734736
1.61274 9.09034083e-13

The other curve is too weak (snfs214). Two best from the original 1-4 set:

Code:

(5, -8) c209 / snfs211 --> poly 1
n: 16565142879290504125073762823767688357919065144905054524206541804478927498295871968740614207813511789416877059764381461295498833321565662655267058786104869169924987607710560565878863482569594348028360005558029
# a = 41975122822309585932179048083335413789237187395320622/5839938684558745126126493478220245287305063414925575
Y0: -41975122822309585932179048083335413789237187395320622
Y1: 5839938684558745126126493478220245287305063414925575
# poly x^4 - 6*x^3 + 17*x^2 - 84*x + 196
c4: 1
c3: -6
c2: 17
c1: -84
c0: 196
skew: 4.63621
# E = 7.86184221e-13 <-- best in 1-4
---------------------------------------------
(5, -8) c209 / snfs211 --> poly 4
n: 16565142879290504125073762823767688357919065144905054524206541804478927498295871968740614207813511789416877059764381461295498833321565662655267058786104869169924987607710560565878863482569594348028360005558029
# a = 43070674852707955981472641819129110567338930886162219/9307684042037302713836537085227216320008466867809161
Y0: -43070674852707955981472641819129110567338930886162219
Y1: 9307684042037302713836537085227216320008466867809161
# poly x^4 - 6*x^3 + 17*x^2 - 84*x + 196
c4: 1
c3: -6
c2: 17
c1: -84
c0: 196
skew: 4.03775
# E = 7.79756343e-13 <-- second best in 1-4

swishzzz · 2021-06-12, 23:29

Code:

Sat Jun 12 19:06:21 2021  p66 factor: 665143820458149897367011706930190969209165571709435291862102921201
Sat Jun 12 19:06:21 2021  p80 factor: 44893098614779250493405940343123948769333444856278197198665021119423641761101073

Murphy e-score = 1.051e-11, 41.6M raw relations processed into a 3.3M matrix.

swishzzz · 2021-06-12, 23:54

Quote:

Originally Posted by Max0526

I built the SNFS spinner and it works!

The two polys above easily spin up 9%, and above 1e-16.

Code:

1,0,2,-12,10,64721779846090870584687710177775610620232746158905507772251959674640,-635923536702326787817298118516458633781564513999305084724198320018881
2.80011 1.01256249e-16

2,2,1,4,8,126098023634219809899835084852688502789500914867461314423771553552960,-601881864613820064535786766952632109810808546510423702878545714894641
2.23217 1.01166588e-16

1,0,8,-96,160,32360889923045435292343855088887805310116373079452753886125979837320,-635923536702326787817298118516458633781564513999305084724198320018881
5.58749 1.00195264e-16

If anybody wants to try spinning a poly, please let me know.

@swishzzz
Is it possible to try with Python please?
1) feed in the text file with 7 comma separated values for c4,c3,c2,c1,c0,Y1,Y0 on each line, the test file is attached;
2) connect to http://myfactorcollection.cownoise.c...lculators.html, the top form Optimal Skew;
3) send records through Optimal Skew form and receive the skew and E score for each;
4) pick the line with the best E score;
5) how long would it take to process 100 records?

Can try with this script (change .txt to .py and make sure you install the "requests" package). Less than 1 min for 100 records.

Max0526 · 2021-06-13, 02:36

Quote:

Originally Posted by swishzzz

Can try with this script (change .txt to .py and make sure you install the "requests" package). Less than 1 min for 100 records.

Love it!
It's so easily scalable to c5 and c6!

Code:

# modify the following two values
fields = "c4,c3,c2,c1,c0,y1,y0"
entries = """
...

It will be a great asset for everybody who does SNFS!
I hope you don't mind us advertising it a bit on your behalf.
Thank you so much, swishzzz!

bur · 2021-06-13, 05:34

The sieving goes on... 21,608,085/22,162,816

I was hoping the estimation was good, because I got a similar value of 17,000,000 with the function that yafu uses.

Well, as long as it won't take longer than a comparable GNFS run, I'm fine. :D The rels/q are already down to 1100 rels / 10000 q though.

fivemack · 2021-06-13, 09:42

Quote:

Originally Posted by Max0526

This is what we call a nice split! c156=2*p78! Done and done!
Thank you, fivemack!
Could you please send me that c156 GNFS poly? How did you select it? msieve GPU or CADO?
Could you also take a look at https://mersenneforum.org/showpost.p...postcount=228?

This is the polynomial I used, which I found in a few hours with a version of msieve CPU that I found under a rock (I think an x86_64 executable running on my ARM Mac)

Code:

n: 202825684447247563197193403459240192747123552357366811463743019450351963962508510059107114168102414937775338054288748526215596251938048377945059566855020817
skew: 236437581.96
c0: 39888441025249897937155045491418965541289397
c1: 79310553754964340750730957357536373
c2: -4334253142523712812747088913
c3: -2990011334403374889
c4: 100730842476
c5: 36
Y0: -5625506082442257542922649046990
Y1: 57665872119821731
lpbr: 30
lpba: 30
mfbr: 60
mfba: 60
alambda: 2.6
rlambda: 2.6
alim: 40000000
rlim: 40000000

Sieved Q=20M..60M, 2746 thread-hours on 40-thread machine.

Quote:

EDIT:
fivemack, I hope you haven't started line 88 (5, -8) yet. If you haven't, could you please hold on for a bit, just to give me a chance to look at all available SNFS polys and their spins for that c209? It would be a perfect training ground for the definitely possible (just checked) polys 5-8 and the newly baked spinner.

I'm afraid I'm already several thread-weeks into the SNFS run, I don't think it's worth restarting now even for the 20% increase in E value. I expect factors in about a week.

EdH · 2021-06-13, 14:33

Quote:

Originally Posted by EdH

My Bash script failed (miserably) to spin the tested SNFS polynomial on one of my machines. I'm trying the same machine on the latest GNFS requested poly (in the msieve thread) to see if that returns anything at least equal to the initial one.

As to the c312, it is at t50. Moving to the c326. . .

My GNFS spin appears to be running properly on the number from another thread, so I guess I need to approach SNFS spinning in a different manner.

As for line 161, the c326 shed a small prime to become a c298 which has had t50 ECM. I'm curently ECMing the c328 a bit. . .

Max0526 · 2021-06-13, 15:17

Quote:

Originally Posted by EdH

My GNFS spin appears to be running properly on the number from another thread, so I guess I need to approach SNFS spinning in a different manner.

As for line 161, the c326 shed a small prime to become a c298 which has had t50 ECM. I'm curently ECMing the c328 a bit. . .

EdH, please read through the explanation of what happens in CADO with SNFS polys:

Quote:

Originally Posted by charybdis

CADO's sizeopt and rootopt are designed for GNFS, so it makes sense that they would produce odd results when fed SNFS polynomials. SNFS polys have amazing size properties; this is why SNFS is so much faster than GNFS. The size properties of SNFS polynomials are so good that we don't usually care about the root properties at all (it's common to see positive alpha values for SNFS, indicating the algebraic norms are *less* likely to be smooth than random numbers of the same size!), and any attempt to run rootopt will likely ruin the size properties. CADO won't realise that the resulting polynomial is bad, because it doesn't know that it was given an SNFS polynomial!

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
factoring 2ⁿ-2 equivalent to factoring 2ⁿ-1(I think)	baih	Miscellaneous Math	9	2020-09-21 07:11
OpenCL GPU P-1 Factoring and ECM Factoring	xx005fs	GPU Computing	3	2018-10-27 14:49

2021-06-12, 20:36	#245
firejuggler "Vincent" Apr 2010 Over the rainbow 2²·7·103 Posts	i'm jumping around but (10,9)c204= 184374369527696042002774666493*P175

2021-06-13, 05:34	#250
bur Aug 2020 796581e-4;32539e-3 659 Posts	The sieving goes on... 21,608,085/22,162,816 I was hoping the estimation was good, because I got a similar value of 17,000,000 with the function that yafu uses. Well, as long as it won't take longer than a comparable GNFS run, I'm fine. :D The rels/q are already down to 1100 rels / 10000 q though.