Best msieve poly scores - Page 8

lorgix · 2017-11-07, 17:26

Just to fill in the gap in the table; here is a poly for the c149 cofactor of 6529^59-1.

Code:

R0: -149556254781551014261354736958
R1: 11730155871184751
A0: 42741186048570255159903007583861298925
A1: 4209644426190763741440777376759
A2: -2350189023711363370713455
A3: -131754643072821427
A4: 29177915758
A5: 240
skew 10400424.38, size 2.300e-014, alpha -6.436, combined = 7.109e-012 rroots = 5

Max0526 · 2017-11-07, 22:27

Quote:

Originally Posted by lorgix

Just to fill in the gap in the table; here is a poly for the c149 cofactor of 6529^59-1.

Code:

R0: -149556254781551014261354736958
R1: 11730155871184751
A0: 42741186048570255159903007583861298925
A1: 4209644426190763741440777376759
A2: -2350189023711363370713455
A3: -131754643072821427
A4: 29177915758
A5: 240
skew 10400424.38, size 2.300e-014, alpha -6.436, combined = 7.109e-012 rroots = 5

Not the best score for C149 though. The best one so far is reported here: http://www.mersenneforum.org/showpos...6&postcount=59 (149, 7.912e-12, Max0526, 03-17, (56*10^225-11)/9), originally from http://stdkmd.com/nrr/c.cgi?q=62221_225 and http://www.mersenneforum.org/showpos...0&postcount=46.

lorgix · 2017-11-08, 06:26

Is there a chart that is up to date, or will there be one?

VBCurtis · 2017-11-08, 15:17

Lorgix-
Perhaps at winter break. Once we had results for most size entries, the thread had served its purpose (providing target scores more realistic than msieve's estimates). So, I lost interest. I think Max posted a mostly-accurate chart a couple months ago, but I haven't updated in a year or more. Anyone can, the data is all in this thread!

lorgix · 2017-11-10, 12:50

Quote:

Originally Posted by fivemack

The 1.194e-13 178-digit result was for the cofactor of aliquot 8352 step 1764 (log=177.09)
The 1.800e-13 175-digit result was for the cofactor of aliquot 8352 step 1769 (log=174.78)
The 4.132e-13 169-digit result was for the cofactor of aliquot 3270 step 695 (log=168.73)
The 6.902e-13 167-digit result was for the cofactor of aliquot 8154 step 2379 (log=166.43)
The 2.051e-12 159-digit result was for the cofactor of aliquot 9436 step 1335 (log=158.04)

I have just found a 4.113e-13 for a2360.1675 (log=168.77)

My slight concern with tables of records is that, after a while, they will be recording mostly cases of people who've done too much polynomial selection: my 4.113 came from 9.6 million CPU-seconds of polsel, whilst it shouldn't take more than 35 million CPU-seconds to sieve on the same hardware. If I threw 128 CPU-weeks of msieve at some random C150 I could get an unassailably large entry in the table, but it would be strikingly useless.

That's a good point.
We already see the nice line that log(e-score) falls on when plotted against log(log(n)).
So I guess this kind of table is already "just for fun".

Quote:

Originally Posted by VBCurtis

Lorgix-
Perhaps at winter break. Once we had results for most size entries, the thread had served its purpose (providing target scores more realistic than msieve's estimates). So, I lost interest. I think Max posted a mostly-accurate chart a couple months ago, but I haven't updated in a year or more. Anyone can, the data is all in this thread!

I understand. We'll see who gets to it first.

I think I found a record;
547^97-1 (c151)

Code:

# norm 2.250160e-014 alpha -7.442564 e 6.423e-012 rroots 5
skew: 1576336.35
c0: -224892123760860127599655348688526705
c1: 2656807800756173903312494013874
c2: 1193061104863474198592747
c3: -2462504880291162356
c4: -482629274860
c5: 358800
Y0: -78619529763842715013646237482
Y1: 321544603785103

# norm 2.235836e-014 alpha -7.394872 e 6.391e-012 rroots 5
skew: 1513923.95
c0: 125503290026964733079026712104739775
c1: 2332576585400481642173343711810
c2: 1923497881118766401251019
c3: -2223929271425703796
c4: -668530730860
c5: 358800
Y0: -78619529797162453036273750754
Y1: 321544603785103

Posting both because they are close and in a league of their own.
For this number I found that reducing default stg1norm by a factor of about 10 (9.6~14.3 didn't make much difference) produced -nps hits at a nice rate. (Very much in line with what Curtis has suggested.) Probably because it was as loose as it could be without the GPU having to wait for the CPU very much.
This was done with a Tesla M2050 and a 3770K (just barely overclocked (3.9), and somewhat busy with other tasks at the same time (msieve having higher priority)).

lorgix · 2017-11-27, 09:41

337^107-1 (c169)

Code:

# norm 2.533366e-016 alpha -7.481840 e 4.446e-013 rroots 3
skew: 11205760.32
c0: -81690765888701212192779918819233510331056
c1: 30263975410997972781178589601887820
c2: -418559681536534102425790
c3: -617271737090854076550
c4: -8536570103769
c5: 1612620
Y0: -273396355344645938020790224019355
Y1: 41238030906080357

Max0526 · 2017-12-04, 17:32

Quote:

Originally Posted by lorgix

That's a good point.
We already see the nice line that log(e-score) falls on when plotted against log(log(n)).
So I guess this kind of table is already "just for fun".

I understand. We'll see who gets to it first.

I think I found a record;
547^97-1 (c151)

Code:

# norm 2.250160e-014 alpha -7.442564 e 6.423e-012 rroots 5
skew: 1576336.35
c0: -224892123760860127599655348688526705
c1: 2656807800756173903312494013874
c2: 1193061104863474198592747
c3: -2462504880291162356
c4: -482629274860
c5: 358800
Y0: -78619529763842715013646237482
Y1: 321544603785103

# norm 2.235836e-014 alpha -7.394872 e 6.391e-012 rroots 5
skew: 1513923.95
c0: 125503290026964733079026712104739775
c1: 2332576585400481642173343711810
c2: 1923497881118766401251019
c3: -2223929271425703796
c4: -668530730860
c5: 358800
Y0: -78619529797162453036273750754
Y1: 321544603785103

Posting both because they are close and in a league of their own.
For this number I found that reducing default stg1norm by a factor of about 10 (9.6~14.3 didn't make much difference) produced -nps hits at a nice rate. (Very much in line with what Curtis has suggested.) Probably because it was as loose as it could be without the GPU having to wait for the CPU very much.
This was done with a Tesla M2050 and a 3770K (just barely overclocked (3.9), and somewhat busy with other tasks at the same time (msieve having higher priority)).

CADO optimizes these polys a bit:

Code:

n: 1077716771708737183340860457188574541025402619131148434040051046114297574429192725186342646883142367258981585221196584917539077642715811408057416693097
Y0: -78619529722037093352526070937
Y1: 321544603785103
c0: -4936447502346974967017835975753520
c1: 2838431476497643514342070411984
c2: 191509051171781641885727
c3: -2652849866967553956
c4: -249382364860
c5: 358800
skew: 1590081.73888
# lognorm 46.88, E 39.42, alpha -7.46 (proj -2.67), 5 real roots
# MurphyE = 6.44466287e-12

Max0526 · 2017-12-25, 19:04

Lucas number cofactor
http://factordb.com/index.php?id=1100000000454879007
http://factordb.com/index.php?query=L3865B
Found by CADO-NFS, optimized by Msieve
By swellman's PM request

Code:

# norm 6.135551e-18 alpha -9.140155 e 4.696e-14 rroots 5 
skew: 155081925.86 
c0: 3258381761015893322703469355508610515519640008 
c1: 994948670578842945305212888379095752533 
c2: -11449251905133422400956610990120 
c3: -69609283675092247656887 
c4: 438761968162182 
c5: 1519560 
Y0: -510262266069797436279473833437161562 
Y1: 3739355384257464468353

VBCurtis · 2018-01-31, 03:58

Attached, a 2017 list of best polys:

Code:

Digits Score        Found by           Date     Number
145    1.394e-11    fivemack           04-16    ?
146    1.232e-11    Max0526            02-17    (136*10^208-1)/9
147    1.034e-11    Max0526            03-17    (10^215+54*10^107-1)/9
148    9.335e-12    Max0526            02-17    (59*10^229+13)/9
149    7.912e-12    Max0526            03-17    (56*10^225-11)/9
150    6.74e-12      chris2be8          02-14    572000:3135
151    6.423e-12    lorgix             11-17    547^97-1
152    5.193e-12    fivemack           04-16    ?
153    4.496e-12    RichS              09-17    434040:3059
154    3.821e-12    Max0526            02-17    (17*10^222-53)/9
155    3.33e-12      Max0526            02-17    10^229-11
156    2.893e-12    Max0526            03-17    (65*10^224-11)/9
157    2.776e-12    Max0526            02-17    (49*10^243-31)/9
158    2.430e-12    Max0526            02-17    (86*10^228-23)/9
159    2.051e-12    fivemack           04-16    9436:1335
160    1.656e-12    Max0526            02-17    (13*10^242-31)/9
161    1.635e-12    unconnected        02-16    933436:12438
162    1.238e-12    Max0526            02-17    (7*10^247-367)/9
163    1.096e-12    Max0526            03-17    (46*10^232+71)/9
164    9.26e-13     VBCurtis           01-16    13*2^797-1
165    8.531e-13    Max0526 	       09-17    (34*10^236-7)/9
166    7.11e-13     chris2be8          04-14    3408:1399
167    6.902e-13    fivemack           04-16    8154:2379
168    5.32e-13     sashamkrt          10-13    xyyx130_119
169    4.446e-13    lorgix             11-17    337^107-1
170    3.910e-13    RichD              02-16    P226+1 (wblipp)
171    3.730e-13    Max0526            02-17    (4*10^261-13)/9
172    3.338e-13    Max0526            12-16    13*2^905-1
173    2.579e-13    Max0526            08-17    xyyx130_129
174    2.010e-13    Max0526            03-17    (10^283+179)/9
175    1.800e-13    fivemack           04-16    8352:1769
176    1.631e-13    Max0526            03-17    (10^294+59)/3
177    1.448e-13    Max0526            03-17    (10^275+45*10^137-1)/9
178    1.194e-13    fivemack           04-16    8352:1764
180	 8.726e-14   Max0526		10-17	3408:1671
181	 7.874e-14   wombatman	       08-17	?
182	 6.732e-14   E. Branger	       08-17	(7*10^294-61)/9
184    4.696e-14    Max0526       	12-17	L3865B (msieve rootopt)
186    3.386e-14    firejuggler+CADO   04-17    122!+1
190    2.221e-14    frmky+CADO         09-16    7,401-
191    1.581e-14    Max0526            07-16    10,429-
192    1.698e-14    Max0526            09-16    5,485+
194    1.221e-14    fivemack           04-16    ?
195    8.044e-15    Max0526            03-17    148^83+83^148
196    8.164e-15    frmky+CADO         09-16    5,1085L
197    7.240e-15    Max0526            09-16    10,325+
198    7.15e-15     Gimarel            01-16    E-M
199    5.111e-15    Max0526            10-16    10,515+
204	2.611e-15   E. Branger		08-17   (64*10^335-1)/9
207_5  1.627e-15    wombatman+CADO     04-17    Cofactor of Home Primes Base 2 4496 index number 310
207_6  1.568e-15    firejuggler+CADO   04-17    Cofactor of Home Primes Base 2 4496 index number 310
208_5	1.439e-15   E. Branger		08-17   (64*10^329-1)/9
210_6	1.204e-15   R. Propper		09-13	RSA-210
212_6	9.426e-16   Shi Bai et al	07-12   RSA-704 (CADO)
216_6  4.66e-16     frmky              08-13    3,766+
216_5  3.61e-16     Gimarel            08-13    3,766+
221_6  3.36e-16     Gimarel            12-13    3,697+

I ignored some polys that were not used in factorizations; for instance, a previous record that Max ran through CADO to add a percent or two of score. I don't think applying current tools to previously run jobs adds to the table.
One should assume all Max0526 records are CADO unless otherwise marked, likewise all other users are msieve unless otherwise marked.
I look forward to trying a CADO search on deg 6; we simply need a worthy candidate! Wombatman, did your C207 ever get queued on Greg's 16e?

Dubslow · 2018-01-31, 04:03

Quote:

Originally Posted by VBCurtis

Attached, a 2017 list of best polys:

Code:

Digits Score        Found by           Date     Number
145    1.394e-11    fivemack           04-16    ?
146    1.232e-11    Max0526            02-17    (136*10^208-1)/9
147    1.034e-11    Max0526            03-17    (10^215+54*10^107-1)/9
148    9.335e-12    Max0526            02-17    (59*10^229+13)/9
149    7.912e-12    Max0526            03-17    (56*10^225-11)/9
150    6.74e-12      chris2be8          02-14    572000:3135
151    6.423e-12    lorgix             11-17    547^97-1
152    5.193e-12    fivemack           04-16    ?
153    4.496e-12    RichS              09-17    434040:3059
154    3.821e-12    Max0526            02-17    (17*10^222-53)/9
155    3.33e-12      Max0526            02-17    10^229-11
156    2.893e-12    Max0526            03-17    (65*10^224-11)/9
157    2.776e-12    Max0526            02-17    (49*10^243-31)/9
158    2.430e-12    Max0526            02-17    (86*10^228-23)/9
159    2.051e-12    fivemack           04-16    9436:1335
160    1.656e-12    Max0526            02-17    (13*10^242-31)/9
161    1.635e-12    unconnected        02-16    933436:12438
162    1.238e-12    Max0526            02-17    (7*10^247-367)/9
163    1.096e-12    Max0526            03-17    (46*10^232+71)/9
164    9.26e-13     VBCurtis           01-16    13*2^797-1
165    8.531e-13    Max0526 	       09-17    (34*10^236-7)/9
166    7.11e-13     chris2be8          04-14    3408:1399
167    6.902e-13    fivemack           04-16    8154:2379
168    5.32e-13     sashamkrt          10-13    xyyx130_119
169    4.446e-13    lorgix             11-17    337^107-1
170    3.910e-13    RichD              02-16    P226+1 (wblipp)
171    3.730e-13    Max0526            02-17    (4*10^261-13)/9
172    3.338e-13    Max0526            12-16    13*2^905-1
173    2.579e-13    Max0526            08-17    xyyx130_129
174    2.010e-13    Max0526            03-17    (10^283+179)/9
175    1.800e-13    fivemack           04-16    8352:1769
176    1.631e-13    Max0526            03-17    (10^294+59)/3
177    1.448e-13    Max0526            03-17    (10^275+45*10^137-1)/9
178    1.194e-13    fivemack           04-16    8352:1764
180	 8.726e-14   Max0526		10-17	3408:1671
181	 7.874e-14   wombatman	       08-17	?
182	 6.732e-14   E. Branger	       08-17	(7*10^294-61)/9
184    4.696e-14    Max0526       	12-17	L3865B (msieve rootopt)
186    3.386e-14    firejuggler+CADO   04-17    122!+1
190    2.221e-14    frmky+CADO         09-16    7,401-
191    1.581e-14    Max0526            07-16    10,429-
192    1.698e-14    Max0526            09-16    5,485+
194    1.221e-14    fivemack           04-16    ?
195    8.044e-15    Max0526            03-17    148^83+83^148
196    8.164e-15    frmky+CADO         09-16    5,1085L
197    7.240e-15    Max0526            09-16    10,325+
198    7.15e-15     Gimarel            01-16    E-M
199    5.111e-15    Max0526            10-16    10,515+
204	2.611e-15   E. Branger		08-17   (64*10^335-1)/9
207_5  1.627e-15    wombatman+CADO     04-17    Cofactor of Home Primes Base 2 4496 index number 310
207_6  1.568e-15    firejuggler+CADO   04-17    Cofactor of Home Primes Base 2 4496 index number 310
208_5	1.439e-15   E. Branger		08-17   (64*10^329-1)/9
210_6	1.204e-15   R. Propper		09-13	RSA-210
212_6	9.426e-16   Shi Bai et al	07-12   RSA-704 (CADO)
216_6  4.66e-16     frmky              08-13    3,766+
216_5  3.61e-16     Gimarel            08-13    3,766+
221_6  3.36e-16     Gimarel            12-13    3,697+

I ignored some polys that were not used in factorizations; for instance, a previous record that Max ran through CADO to add a percent or two of score. I don't think applying current tools to previously run jobs adds to the table.
One should assume all Max0526 records are CADO unless otherwise marked, likewise all other users are msieve unless otherwise marked.
I look forward to trying a CADO search on deg 6; we simply need a worthy candidate! Wombatman, did your C207 ever get queued on Greg's 16e?

Surely msieve and CADO should be kept in different ~~tables~~ columns? The last few posts here alone show they remain not directly comparable. For the same poly, you and Max have scores that differ by two orders of magnitude.

VBCurtis · 2018-01-31, 05:00

As far as I know, all CADO scores in the table are those computed using the regular Murphy E-score, not the one the current CADO tools report. I think Max has been aiming msieve at CADO polys to read E-scores, but perhaps he has a more elegant way.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
msieve parallel poly selection with MPI	drone84	Msieve	4	2017-06-28 09:18
msieve poly select: choosing Stage1norm	VBCurtis	Msieve	0	2016-04-11 21:33
Resume msieve poly search job?	Andi47	Msieve	1	2011-03-28 04:30
Msieve / lattice siever with degree 7/8 poly	Batalov	Msieve	54	2010-01-13 19:45
Different msieve 1.39 poly selection outputs...	Jeff Gilchrist	Msieve	5	2008-12-29 23:07

2017-11-08, 06:26	#80
lorgix Sep 2010 Scandinavia 3·5·41 Posts	Is there a chart that is up to date, or will there be one?

2017-11-08, 15:17	#81
VBCurtis "Curtis" Feb 2005 Riverside, CA 1001011111101₂ Posts	Lorgix- Perhaps at winter break. Once we had results for most size entries, the thread had served its purpose (providing target scores more realistic than msieve's estimates). So, I lost interest. I think Max posted a mostly-accurate chart a couple months ago, but I haven't updated in a year or more. Anyone can, the data is all in this thread!

2018-01-31, 05:00	#88
VBCurtis "Curtis" Feb 2005 Riverside, CA 4,861 Posts	As far as I know, all CADO scores in the table are those computed using the regular Murphy E-score, not the one the current CADO tools report. I think Max has been aiming msieve at CADO polys to read E-scores, but perhaps he has a more elegant way.