Queue management for 16e queue

VBCurtis · 2020-09-04, 22:33

Post jobs here you wish to be run on the 16e siever = f_small queue.

Guidelines:
1. Greg has requested a maximum lim of 225M to keep memory use comfortably below 1GB/client process and to keep our jobs from overlapping the size of his. Edit: 200/250M should be acceptable, averaging to 225M.
2. The more jobs posted to 16e rather than 15e, the faster 15e will move along.

Many jobs are roughly the same speed on 15e and 16e; for instance, GNFS 184-187. Post where you like, but consider posting test-sieve and suggested Q-range for both 15e and 16e if your job is near the crossover; both so others can learn where the crossover is, and so the queue managers can keep the wait as short as possible on all queues.

Note: Greg has also requested a lim cap of 200M on 15e.

fivemack · 2020-10-04, 20:29

Would anyone be up for say 30k curves at B1=260M on F1565

Code:

898577600957458082817012215882165508563668653664358493738893522466409221942746665068993028034375228782419911153278586365513991645861651135664033105687400260471070947660690029466178073221007827086032497462411095795403425738745977703024509992798576951889223184161

It seems at the difficult end of fs problems (261-digit quartic, I'm seeing yield of about 1.5 relations per Q with 3x33lpr 2x31lpa) and I think it does need that degree of ECM before it's sensible to throw volunteer cycles at the SNFS.

swellman · 2020-10-04, 21:43

Quote:

Originally Posted by fivemack

Would anyone be up for say 30k curves at B1=260M on F1565

Code:

898577600957458082817012215882165508563668653664358493738893522466409221942746665068993028034375228782419911153278586365513991645861651135664033105687400260471070947660690029466178073221007827086032497462411095795403425738745977703024509992798576951889223184161

It seems at the difficult end of fs problems (261-digit quartic, I'm seeing yield of about 1.5 relations per Q with 3x33lpr 2x31lpa) and I think it does need that degree of ECM before it's sensible to throw volunteer cycles at the SNFS.

That’s a good sized ECM. I’ll ask Yoyo@Home if he’s willing to carry it.

swellman · 2020-10-05, 09:53

Quote:

Originally Posted by fivemack

Would anyone be up for say 30k curves at B1=260M on F1565

Code:

898577600957458082817012215882165508563668653664358493738893522466409221942746665068993028034375228782419911153278586365513991645861651135664033105687400260471070947660690029466178073221007827086032497462411095795403425738745977703024509992798576951889223184161

It seems at the difficult end of fs problems (261-digit quartic, I'm seeing yield of about 1.5 relations per Q with 3x33lpr 2x31lpa) and I think it does need that degree of ECM before it's sensible to throw volunteer cycles at the SNFS.

Yoyo@Home is running this job now. Under the Aliquot tab.

fivemack · 2020-10-09, 10:05

OK, my initial relation-sampling was over too small a region; I'm seeing roughly one relation per Q for

Code:

n: 898577600957458082817012215882165508563668653664358493738893522466409221942746665068993028034375228782419911153278586365513991645861651135664033105687400260471070947660690029466178073221007827086032497462411095795403425738745977703024509992798576951889223184161
skew: 1
c4: 1
c3: -1
c2: 1
c1: -1
c0: 1
Y1: -71558092601766452430641106302905217344934236440122960529002115744
Y0: 187341518601536966291015050946540312701895836604078191803255601777
lpbr: 33
lpba: 31
mfbr: 96
mfba: 62
rlambda: 3.4
alambda: 2.4
rlim: 250000000
alim: 200000000

in the Q=150M..400M range, which would imply sieving 150M..700M to get something plausible, and it's sieving very slowly on my hardware (absolute best time is 2.5 seconds per relation) - I think this is too big for _fs. Thank you for the ECM, but I think this one will have to wait for cheaper computrons.

fivemack · 2020-10-09, 10:11

So, the easiest outstanding Fibonacci quartic is L2995A; difficulty 250.37

http://factordb.com/index.php?showid...00000216955784

Code:

6647246946274833393456661338737441255614261080716309840406229789042569778456310323482125160454068850539445673923995827291479838370880449615345378166944942195594414495045570103808609316829203645670851879601136941

Would it be possible to queue up 20k curves at B1=260M on that one?

An alternative would be to binary-search the limits of _fs feasibility, with L4605A difficulty 256.64

http://factordb.com/index.php?id=1100000000216955924

Code:

3584242181859180897028227800643356671652242043498330593404045120001651398447679822689230799275312050386811076406419865592578911649557946131137999151560109118895314557059384056761129147893233379009026747238392618096742921

25k curves at B1=260M would be nice for that one, I'll do another thorough trial-sieve which will take ~48 hours..

swellman · 2020-10-10, 14:49

Quote:

Originally Posted by fivemack

So, the easiest outstanding Fibonacci quartic is L2995A; difficulty 250.37

http://factordb.com/index.php?showid...00000216955784

Code:

6647246946274833393456661338737441255614261080716309840406229789042569778456310323482125160454068850539445673923995827291479838370880449615345378166944942195594414495045570103808609316829203645670851879601136941

Would it be possible to queue up 20k curves at B1=260M on that one?

An alternative would be to binary-search the limits of _fs feasibility, with L4605A difficulty 256.64

http://factordb.com/index.php?id=1100000000216955924

Code:

3584242181859180897028227800643356671652242043498330593404045120001651398447679822689230799275312050386811076406419865592578911649557946131137999151560109118895314557059384056761129147893233379009026747238392618096742921

25k curves at B1=260M would be nice for that one, I'll do another thorough trial-sieve which will take ~48 hours..

Yoyo has agreed to run these jobs but he asked me to manage the process. Keep in mind that work is limited to 20,000 curves per job for B1 =< 260M, 12,000 curves for B1 = 850M.

I will enqueue both jobs with yoyo in the next day or two.

fivemack · 2020-10-11, 09:28

Thank you so much!

VBCurtis · 2020-10-29, 00:55

QUEUED AS C196_3366_2224

Aliquot 3366 index 2224, a C196, is ready for 16e-small:

Code:

n: 1398516542119091547927880887727114963659389509804105922692120268647401334715710534035357328326678677296622476655524803463753551464210096969080964891534065442657109752753516921412646600842893476177
type: gnfs
skew: 125001952.15
c0: -24623140536189314142763591449244095867976653480
c1: 1591604144502435503327811672315382999292
c2: 45151115187254052425427334115817
c3: -227121374773875476420597
c4: -2975093123155512
c5: 6032880
Y0: -67399847496998630851429653894088679427
Y1: 125291089571292601013129
rlim: 134000000
alim: 250000000
lpbr: 32
lpba: 33
mfbr: 63
mfba: 95
rlambda: 2.6
alambda: 3.6

Test-sieving 1kq blocks indicates a Q-range of 30-245M should be sufficient for ~625M relations:

Code:

Q=20M	4467 rels, 0.596 sec
Q=50M	3766 rels, 0.645 sec	
Q=80M	2982 rels, 0.711 sec
Q=110M	3129 rels, 0.713 sec
Q=140M	2717 rels, 0.765 sec
Q=170M	2011 rels, 0.807 sec
Q=200M	2548 rels, 0.854 sec
Q=230M	2257 rels, 0.877 sec

I would like to handle the matrix.

charybdis · 2020-10-29, 01:13

Quote:

Originally Posted by VBCurtis

Code:

n: 13985165421190915479278808877271149636593895098041059226921202686474013347157105340353$

Might want to fix that line...

Curtis: Thanks! Fixed.

jyb · 2020-10-30, 21:15

QUEUED AS 8p3_834L

SNFS-251.06 (quartic) C216 HCN (8+3,834L), ECM to t60.

Code:

n: 335966767032973712039032374362950903677945846587153282355179398197279228448619141924385189638636246189124256780938091105048640848316159613076786121923701877546449489912905969884545889712234051981784709587026105258473
skew: 1.22474
c4: 4
c3: -12
c2: 18
c1: -18
c0: 9
Y1: 834385168331080533771857328695283
Y0: -411376139330301510538742295639337626245683966408394965837152256
rlim: 225000000
alim: 225000000
lpbr: 32
lpba: 32
mfbr: 96
mfba: 64
rlambda: 3.6
alambda: 2.6

Trial sieving 2K blocks.

Code:

 Q		Yield
---		-----
50M		3083
90M		3649
130M		3730
170M		3497
210M		4019
250M		3774
290M		3954
330M		3735

Recommend sieving special Q on rational side, 50M - 300M.

2020-09-04, 22:33	#1
VBCurtis "Curtis" Feb 2005 Riverside, CA 5989₁₀ Posts	Queue management for 16e queue Post jobs here you wish to be run on the 16e siever = f_small queue. Guidelines: 1. Greg has requested a maximum lim of 225M to keep memory use comfortably below 1GB/client process and to keep our jobs from overlapping the size of his. Edit: 200/250M should be acceptable, averaging to 225M. 2. The more jobs posted to 16e rather than 15e, the faster 15e will move along. Many jobs are roughly the same speed on 15e and 16e; for instance, GNFS 184-187. Post where you like, but consider posting test-sieve and suggested Q-range for both 15e and 16e if your job is near the crossover; both so others can learn where the crossover is, and so the queue managers can keep the wait as short as possible on all queues. Note: Greg has also requested a lim cap of 200M on 15e. Last fiddled with by VBCurtis on 2020-09-05 at 18:05

2020-10-04, 20:29	#2
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 1936₁₆ Posts	Would anyone be up for say 30k curves at B1=260M on F1565 Code: 898577600957458082817012215882165508563668653664358493738893522466409221942746665068993028034375228782419911153278586365513991645861651135664033105687400260471070947660690029466178073221007827086032497462411095795403425738745977703024509992798576951889223184161 It seems at the difficult end of fs problems (261-digit quartic, I'm seeing yield of about 1.5 relations per Q with 3x33lpr 2x31lpa) and I think it does need that degree of ECM before it's sensible to throw volunteer cycles at the SNFS.

2020-10-09, 10:11	#6
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2×7×461 Posts	So, the easiest outstanding Fibonacci quartic is L2995A; difficulty 250.37 http://factordb.com/index.php?showid...00000216955784 Code: 6647246946274833393456661338737441255614261080716309840406229789042569778456310323482125160454068850539445673923995827291479838370880449615345378166944942195594414495045570103808609316829203645670851879601136941 Would it be possible to queue up 20k curves at B1=260M on that one? An alternative would be to binary-search the limits of _fs feasibility, with L4605A difficulty 256.64 http://factordb.com/index.php?id=1100000000216955924 Code: 3584242181859180897028227800643356671652242043498330593404045120001651398447679822689230799275312050386811076406419865592578911649557946131137999151560109118895314557059384056761129147893233379009026747238392618096742921 25k curves at B1=260M would be nice for that one, I'll do another thorough trial-sieve which will take ~48 hours..

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Queue management for 14e queue	VBCurtis	NFS@Home	170	2023-01-02 15:27
Queue management for e_small and 15e queues	VBCurtis	NFS@Home	254	2022-01-02 01:59
Run down the queue on MPRIME without quitting GIMPS	Rodrigo	Software	7	2018-05-25 13:26
Improving the queue management.	debrouxl	NFS@Home	10	2018-05-06 21:05
split a prime95 queue & client installation	joblack	Information & Answers	1	2009-01-06 08:45

2020-10-11, 09:28	#8
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 2×7×461 Posts	Thank you so much!