2020 15e post processing reservations and results

swellman · 2020-01-01, 13:31

A new thread for 2020.

fivemack · 2020-01-06, 22:04

Code:

UID        PID  PPID  C       SZ      RSS PSR STIME TTY    STAT   TIME CMD
pineapp+ 30651 29686 99 11505538 45052856  12 08:54 pts/11 Rl+ 9518:25 /home/nfsworld/msieve-svn-20190823/msieve-MP-V256-SKL -v -nc2 -t 14

linear algebra completed 839221 of 51257400 dimensions (1.6%, ETA 699h55m)

ETA 5 February. I think this is about as big as we can sensibly go at 15e:

Code:

992294513 relations
991199455 usable
679822896 unique
511124251 after singleton removal

That's from 100M-850M, so yield of 1.3r/Q, 0.9 unique r/Q.

Did manage to build a matrix from 629723855 unique, but it was a bit cumbersome:

Code:

Fri Jan  3 00:14:09 2020  weight of 60006039 cycles is about 6961061087 (116.01/cycle)

In a fit of optimism I'm doing a polynomial search for L4235A log(N)=200.07

VBCurtis · 2020-01-06, 22:15

Dropping 14% from the matrix size from just 50M more unique relations sounds pretty good! I wonder if 25M more unique relations would have meaningfully helped the matrix; 50M is for me kind of a hard line where I really want more relations to not face something bigger, but you're already stretching the tools, as you noted.

I'd like to do a GNFS-200 on 15e, just because it's a Big Round Number. Perhaps a hybrid of offline 16e (or CADO) up to Q=100M, combined with 15e queue from 100M up.

VBCurtis · 2020-01-16, 05:39

On L4365, I'm getting "filtering wants 1M more relations" after one pass of in-memory singleton removal with TD set to 90. I have 197.2M unique relations and 75.3M duplicates.

May I have another 10MQ? Or 20?

swellman · 2020-01-16, 15:17

Quote:

Originally Posted by VBCurtis

On L4365, I'm getting "filtering wants 1M more relations" after one pass of in-memory singleton removal with TD set to 90. I have 197.2M unique relations and 75.3M duplicates.

May I have another 10MQ? Or 20?

Added 20M to the Q range.

VBCurtis · 2020-01-21, 23:11

Quote:

Originally Posted by swellman

Added 20M to the Q range.

That was sufficient- TD 100 failed in final merge, but TD 84 produced a 29.9M matrix, which will take 2 weeks to solve. 205.6M unique relations.

Batalov · 2020-01-22, 02:25

I can LA 33331_267.

swellman · 2020-01-27, 21:25

Code:

p74 factor: 63146266403491097505456958543771519866531286235829049932705522770396114797
p159 factor: 502359711161757421016410192688166297917171056378426903635199335536365634437389348511892855963909336883092645357270193940562520137791470938242506382255079415667

289.6M raw / 238.2 unique relations. Details at https://pastebin.com/S2iiYT7X.

VBCurtis · 2020-02-03, 17:55

L4365b is complete:

Code:

p86 factor: 25822548244299880006391249505592569731337769069672690296006992583112443319353229235941
Mon Feb  3 09:48:13 2020  p125 factor: 76097052239250743628646066953269717664413797368576928648747487396621406664383446306987339367155281257014361063842022979551961

Sieving this quartic job from Q=40-440M yielded 209M unique relations, and a 29.2M matrix at TD 90 which took ~260hr to solve on 10 cores of Ivy Bridge.

Log at https://pastebin.com/LW2vgMeK

I can do the matrix on L4635A when ready.

fivemack · 2020-02-13, 07:35

Code:

Wed Feb 12 23:55:48 2020  p77 factor: 27699165354693819016801075486232972683723699059340461464932398144896384968753
Wed Feb 12 23:55:48 2020  p123 factor: 138045480967182100682657375185450702724737432877108829119835497270093124360933783400049080678451873650763452274375127767069

991199455 legible input relations; 679822896 after duplicate removal, 511124251 after initial singleton removal. Matrix built at density 132, failed at 134.

37.4 days on 14 cores i9/7940X for 51.26M matrix (brief pause in the middle due to electrical works, after which to my surprise the -ncr ran at 2/3 the speed of the previous half's work)

fivemack · 2020-02-17, 07:10

ETA for F1645 is 11 March: 46.9M matrix

2020-01-01, 13:31	#1
swellman Jun 2012 3×1,283 Posts	2020 15e post processing reservations and results A new thread for 2020.

2020-01-06, 22:04	#2
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2·7·461 Posts	F1459 progressing Code: UID PID PPID C SZ RSS PSR STIME TTY STAT TIME CMD pineapp+ 30651 29686 99 11505538 45052856 12 08:54 pts/11 Rl+ 9518:25 /home/nfsworld/msieve-svn-20190823/msieve-MP-V256-SKL -v -nc2 -t 14 linear algebra completed 839221 of 51257400 dimensions (1.6%, ETA 699h55m) ETA 5 February. I think this is about as big as we can sensibly go at 15e: Code: 992294513 relations 991199455 usable 679822896 unique 511124251 after singleton removal That's from 100M-850M, so yield of 1.3r/Q, 0.9 unique r/Q. Did manage to build a matrix from 629723855 unique, but it was a bit cumbersome: Code: Fri Jan 3 00:14:09 2020 weight of 60006039 cycles is about 6961061087 (116.01/cycle) In a fit of optimism I'm doing a polynomial search for L4235A log(N)=200.07 Last fiddled with by fivemack on 2020-01-07 at 09:46

2020-02-03, 17:55	#9
VBCurtis "Curtis" Feb 2005 Riverside, CA 13010₈ Posts	L4365b is complete: Code: p86 factor: 25822548244299880006391249505592569731337769069672690296006992583112443319353229235941 Mon Feb 3 09:48:13 2020 p125 factor: 76097052239250743628646066953269717664413797368576928648747487396621406664383446306987339367155281257014361063842022979551961 Sieving this quartic job from Q=40-440M yielded 209M unique relations, and a 29.2M matrix at TD 90 which took ~260hr to solve on 10 cores of Ivy Bridge. Log at https://pastebin.com/LW2vgMeK I can do the matrix on L4635A when ready.

2020-02-13, 07:35	#10
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2×7×461 Posts	F1459 done Code: Wed Feb 12 23:55:48 2020 p77 factor: 27699165354693819016801075486232972683723699059340461464932398144896384968753 Wed Feb 12 23:55:48 2020 p123 factor: 138045480967182100682657375185450702724737432877108829119835497270093124360933783400049080678451873650763452274375127767069 991199455 legible input relations; 679822896 after duplicate removal, 511124251 after initial singleton removal. Matrix built at density 132, failed at 134. 37.4 days on 14 cores i9/7940X for 51.26M matrix (brief pause in the middle due to electrical works, after which to my surprise the -ncr ran at 2/3 the speed of the previous half's work)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
2020 14e post processing reservations and results	swellman	NFS@Home	544	2020-12-31 15:14
2019 14e post processing reservations and results	swellman	NFS@Home	862	2019-12-31 10:51
2019 15e post processing reservations and results	swellman	NFS@Home	220	2019-12-28 16:40
2018 15e post-processing reservations and results	fivemack	NFS@Home	221	2019-01-04 13:08
2018 14e post processing reservations and results	pinhodecarlos	NFS@Home	551	2019-01-04 13:06

2020-01-06, 22:15	#3
VBCurtis "Curtis" Feb 2005 Riverside, CA 2³·3·5·47 Posts	Dropping 14% from the matrix size from just 50M more unique relations sounds pretty good! I wonder if 25M more unique relations would have meaningfully helped the matrix; 50M is for me kind of a hard line where I really want more relations to not face something bigger, but you're already stretching the tools, as you noted. I'd like to do a GNFS-200 on 15e, just because it's a Big Round Number. Perhaps a hybrid of offline 16e (or CADO) up to Q=100M, combined with 15e queue from 100M up.

2020-01-16, 05:39	#4
VBCurtis "Curtis" Feb 2005 Riverside, CA 2³·3·5·47 Posts	On L4365, I'm getting "filtering wants 1M more relations" after one pass of in-memory singleton removal with TD set to 90. I have 197.2M unique relations and 75.3M duplicates. May I have another 10MQ? Or 20?

2020-01-22, 02:25	#7
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 89·113 Posts	I can LA 33331_267.

2020-02-17, 07:10	#11
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 1936₁₆ Posts	ETA for F1645 is 11 March: 46.9M matrix