2020 16e post processing reservations and results

swellman · 2020-09-14, 17:07

Currently downloading C191_M61_k32 to see it’s ready for SNFS.

Lucas(1510) aim for 360Mrels can be moved to sieving status at any time - I’m reluctant to move it without understanding the desired Q range.

swellman · 2020-09-15, 23:02

I bumped Q up to 160M for Lucas(1510).

fivemack · 2020-10-04, 10:11

I'm doing an initial filtering run on L1510 today.

fivemack · 2020-10-04, 20:24

That's interesting; the status page https://escatter11.fullerton.edu/nfs/crunching_fs.php says there are 390 million relations, I only see 311875466 relations of which about 249650000 unique and about 180400000 survive singleton removal, and that's not then enough for a matrix. There's plenty of free space on the processing disc so I don't think that's the issue.

I don't know if this is something frmky can usefully look at; I suspect it's easier just to throw more computrons at the problem, so I've extended the range by 20MQ.

frmky · 2020-10-04, 22:45

The script just counts the number of lines. A simple
zcat L1510.dat.gz | wc -l
agrees with the webpage. Sometimes the project gets odd garbage in returned files as we use a forgiving validation check. Likely that.

Edit: Doesn't seem to be that. I just did a quick remdups and got
Found 305246658 unique, 88901570 duplicate (22.6% of total), and 57712 bad relations.
Largest dimension used: 1000 of 1000
Average dimension used: 931.4 of 1000
*** Some redundant relations may have been retained (increase DIM)
*** 47650 (quasi-unique) relations were not hashed

So seems like >300M unique ones there.

fivemack · 2020-10-05, 11:36

Quote:

Originally Posted by frmky

The script just counts the number of lines. A simple
zcat L1510.dat.gz | wc -l
agrees with the webpage

Thanks for looking at that; my script was clearly misbehaving somewhere. Rerunning, I'm seeing 405321355 relations of which 405266903 are in the correct form, 312.9M unique, singleton removal running happily and I'll expect at least 255M after that pass.

(and now I have a 31M density-90 matrix and am trying to optimise the density)

fivemack · 2020-10-10, 10:16

L1510 linalg started, should be done 16 October evening

fivemack · 2020-10-16, 08:30

Code:

Fri Oct 16 06:41:07 2020  p93 factor: 768278079468147872006161615793478445262891405249833298882663471356153278846031859145179203921
Fri Oct 16 06:41:07 2020  p112 factor: 3998301491227939181761854883689691516394446820135772189170217349181287770175361143916814142602476105295172951001

133 hours on 20 cores Skylake Xeon 4114 for a 29.38M matrix at density 106

Log attached and at https://pastebin.com/3e7jQc6b

swellman · 2020-10-21, 13:34

Code:

p84 factor: 548495967839253555860650689244596928240665623636601859336571489439754186678079415649
p107 factor: 50144064156698295308606156707972010557985318544494830853525957506091319040013078433900267278158370866053089

Details at https://pastebin.com/6kNmDaEu

jyb · 2020-11-02, 19:12

Taking 3p2_1578M.

jyb · 2020-11-03, 23:08

Quote:

Originally Posted by jyb

Taking 3p2_1578M.

Now in linear algebra.

2020-09-14, 17:07	#1
swellman Jun 2012 2⁵·3·31 Posts	2020 16e post processing reservations and results Currently downloading C191_M61_k32 to see it’s ready for SNFS. Lucas(1510) aim for 360Mrels can be moved to sieving status at any time - I’m reluctant to move it without understanding the desired Q range.

2020-10-04, 22:45	#5
frmky Jul 2003 So Cal 100000100010₂ Posts	The script just counts the number of lines. A simple zcat L1510.dat.gz \| wc -l agrees with the webpage. Sometimes the project gets odd garbage in returned files as we use a forgiving validation check. Likely that. Edit: Doesn't seem to be that. I just did a quick remdups and got Found 305246658 unique, 88901570 duplicate (22.6% of total), and 57712 bad relations. Largest dimension used: 1000 of 1000 Average dimension used: 931.4 of 1000 * Some redundant relations may have been retained (increase DIM) * 47650 (quasi-unique) relations were not hashed So seems like >300M unique ones there. Last fiddled with by frmky on 2020-10-05 at 04:44

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
2020 “small” 15e post processing reservations and results	pinhodecarlos	NFS@Home	116	2020-12-31 16:44
2020 14e post processing reservations and results	swellman	NFS@Home	544	2020-12-31 15:14
2020 15e post processing reservations and results	swellman	NFS@Home	112	2020-12-29 22:58
2019 15e post processing reservations and results	swellman	NFS@Home	220	2019-12-28 16:40
2018 14e post processing reservations and results	pinhodecarlos	NFS@Home	551	2019-01-04 13:06

2020-09-15, 23:02	#2
swellman Jun 2012 2⁵×3×31 Posts	I bumped Q up to 160M for Lucas(1510).

2020-10-04, 10:11	#3
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 1100011101111₂ Posts	I'm doing an initial filtering run on L1510 today.

2020-10-04, 20:24	#4
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 13×491 Posts	That's interesting; the status page https://escatter11.fullerton.edu/nfs/crunching_fs.php says there are 390 million relations, I only see 311875466 relations of which about 249650000 unique and about 180400000 survive singleton removal, and that's not then enough for a matrix. There's plenty of free space on the processing disc so I don't think that's the issue. I don't know if this is something frmky can usefully look at; I suspect it's easier just to throw more computrons at the problem, so I've extended the range by 20MQ.

2020-10-10, 10:16	#7
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 13×491 Posts	L1510 linalg started, should be done 16 October evening

2020-11-02, 19:12	#10
jyb Aug 2005 Seattle, WA 11010100100₂ Posts	Taking 3p2_1578M.