Fast Breeding (guru management) - Page 105

swellman · 2017-09-21, 10:47

Quote:

Originally Posted by fivemack

...if you don't mind, could you try post-processing when it hits the Q=400M mark, and we can add more relations if that doesn't work.

Of course. Anyone know a good minimum TD to aim for? Or is it build a matrix anyway possible?

RichD · 2017-09-21, 12:03

QUEUED C177 from the OPN t550 file.
Sieve on the algebraic side.

Code:

n: 332835016354511782859920847764611369215654651692152387544372086667611336781169647953439747356831034116086752413757313198324771557842674176530558333493181202822353107340319678877
# 4287755796749^17-1, difficulty: 227.38, skewness: 127.46, alpha: 0.00
# cost: 1.40523e+18, est. time: 669.16 GHz days (not accurate yet!)
lss: 0
skew: 127.459
c6: 1
c0: -4287755796749
Y1: -1
Y0: 78829746184630742457434595059507577749
m: 78829746184630742457434595059507577749
rlim: 200000000
alim: 200000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.7
alambda: 2.7
type: snfs

Trial sieving 5K blocks.

Code:

fivemack · 2017-09-21, 12:14

Quote:

Originally Posted by swellman

Of course. Anyone know a good minimum TD to aim for? Or is it build a matrix anyway possible?

I think it would be interesting to know both 'how many relations to build with td=70' and 'how many relations to build with td=120' (if I'm going to pick a reasonable oversieving point at random)

VBCurtis · 2017-09-21, 16:33

Quote:

Originally Posted by swellman

Of course. Anyone know a good minimum TD to aim for? Or is it build a matrix anyway possible?

This job will stretch 14e more than any other NFS@home has done, at least that I've seen. I think you'll need 700-750M relations at 33LP, so Q may have to reach 550 or 600M (by that time, yield will be dropping rather quickly, above 2 * alim and all...). If you can't build a good matrix (say, TD 110 or higher), we can sieve low Q values off-grid using 15e or even 16e to get a few million more relations.

I hope this works!

swellman · 2017-09-21, 23:02

C213_142_112 is ready for SNFS using 14e.

Code:

n: 455715217613905419326896862948405969104460570801874971719843970220193901956710849582700079225285279197142578692909920761615400944852069264077050644230406950814525311956771920417257379928316326381911686147209565113
# 142^112+112^142, difficulty: 261.86, anorm: 9.94e+038, rnorm: -2.13e+049
# scaled difficulty: 263.58, suggest sieving rational side
# size = 3.024e-013, alpha = 0.000, combined = 4.961e-014, rroots = 0
type: snfs
size: 261
skew: 2.1646
c6: 49
c0: 5041
Y1: -14475468573095697959140800125364450189377536
Y0: 149248036287203113136096720283566231
rlim: 268000000
alim: 268000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving with Q in blocks of 5k.

Code:

Q=20M     rels=6797
Q=80M     rels=5510
Q=150M    rels=4901
Q=250M    rels=4511
Q=450M    rels=3969

Suggesting a sieving range of 20M-520M with a target # rels = 460M.

VBCurtis · 2017-09-22, 05:23

For 202_137_75, might we sieve the -a side from 20M up rather than the -r side from 400M up? I'm getting timings twice as good for Q=20M -a versus Q=400M -r:
Q = 20M -a 0.160 sec/rel yield 1.75/Q
Q = 120M -a 0.229 sec/rel yield 1.32/Q

Q = 400M -r 0.405 sec/rel yield 1.00/Q
(tests done on 2kQ regions)
Perhaps this job could be re-queued on the -a side from 20M up to generate the remaining relations after -r finishes to Q=400M? perhaps as 202_137_75a, and Sean can just zcat together the two dat files before filtering?

Does sieving both sides produce more duplicates than normal?

swellman · 2017-09-22, 09:19

Just keep in mind I'm using a Win 7 box. Copy command - no zcat. Not sure what the limit on file size in DOS environment. Merging two huge files on my machine could be problematic, though in all fairness I've not tried that for some years.

unconnected · 2017-09-22, 14:58

C157 from aliquot sequence 11040. I'll run LA for it.

Code:

# lognorm 47.75, E 41.44, alpha -6.31 (proj -1.64), 5 real roots
# MurphyE = 2.41811536e-12
n: 2153646797243598406022781698871921179468736333856465460905533539179197067080286019717757270291317968880659999793269997464862726882530680275363360113121869979
Y0: -3412887084652747994920013434801
Y1: 726799483512671712557
c0: -155739768072620872585403642819221342740
c1: 105731537875959470823347887885348
c2: 16051986632892334694632495
c3: -2782308567581898834
c4: -134653299134
c5: 4680
skew: 11561243.62085
rlim: 35200000
alim: 35200000
lpbr: 29
lpba: 29
mfbr: 58
mfba: 58
rlambda: 2.6
alambda: 2.6
type: gnfs

Suggesting sieving range 20M-60M.

VBCurtis · 2017-09-22, 15:06

Quote:

Originally Posted by fivemack

Sorry, that was my fault. I thought it was going very slowly for a 33LP.

I have submitted 143_57 as 14e/33, and this time actually changed the LPB values in the input file rather than just the number in the description. Initially sieving 200MQ; let's look at the yield when that's done and see how much further VBCurtis wants to go.

Looks like the initial 200MQ will produce 560M relations; I think 720-750M is enough for this job, so how about another 70MQ (to 290 total)?

fivemack · 2017-09-22, 17:11

Quote:

Originally Posted by VBCurtis

Looks like the initial 200MQ will produce 560M relations; I think 720-750M is enough for this job, so how about another 70MQ (to 290 total)?

Your wish is my command. Done.

swellman · 2017-09-23, 21:50

C222_143_73 is ready for SNFS. It is notionally best sieved on the rational side, but the algebraic side is almost as good. Perhaps this should be sieved on both sides simultaneously? Is that command lss:0? For the gatekeeper to decide.

Code:

n: 103501114207510940834382339780477032273818392297712699737243103372773427310309763554999518496440606907949943972568789048639960888063252135338756392008546273739323139843951769793341393703783471868302578674124875908308512349
# 143^73+73^143, difficulty: 268.32, anorm: 2.04e+038, rnorm: 2.43e+050
# scaled difficulty: 270.33, suggest sieving rational side
# size = 4.244e-013, alpha = 0.000, combined = 6.046e-014, rroots = 0
type: snfs
size: 268
skew: 4.6749
c6: 1
c0: 10439
Y1: -73119371471655725294164801
Y0: 524503804723748275248688345080154231098133441
rlim: 268000000
alim: 268000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving on the -r side with Q in blocks of 10K

Code:

total yield: 15159, q=20010017 (0.93006 sec/rel)
total yield: 14175, q=50010001 (0.97618 sec/rel)
total yield: 12850, q=80010001 (1.09179 sec/rel)
total yield: 12401, q=150010001 (1.34504 sec/rel)
total yield: 10877, q=250010011 (1.61802 sec/rel)
total yield: 9785, q=350010013 (1.75480 sec/rel)

Suggesting a sieving range for Q of 20M-480M with target # rels 500M

Test sieving on the -a side with Q in blocks of 10K

Code:

total yield: 17378, q=20010017 (0.80164 sec/rel)
total yield: 13575, q=50010001 (0.98051 sec/rel)
total yield: 12230, q=80010001 (1.12366 sec/rel)
total yield: 11622, q=150010001 (1.32670 sec/rel)
total yield: 7822, q=250010011 (1.82551 sec/rel)
total yield: 9263, q=350010013 (1.80202 sec/rel)

Suggesting a sieving range of 20M-500M with target # rels 500M.

A combined sieving on the -r and -a sides would have duplicates, but I am unsure of the percentage. Is a sieving range of 20M-350M a feasible place to start?

Whatever is decided, I will perform the postprocessing.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
System management notes	kriesel	kriesel	7	2020-10-21 18:52
Improving the queue management.	debrouxl	NFS@Home	10	2018-05-06 21:05
Script-based Primenet assignment management	ewmayer	Software	3	2017-05-25 04:02
Do normal adults give themselves an allowance? (...to fast or not to fast - there is no question!)	jasong	jasong	35	2016-12-11 00:57
Power Management settings	PrimeCroat	Hardware	3	2004-02-17 19:11

2017-09-22, 05:23	#1150
VBCurtis "Curtis" Feb 2005 Riverside, CA 2·2,437 Posts	For 202_137_75, might we sieve the -a side from 20M up rather than the -r side from 400M up? I'm getting timings twice as good for Q=20M -a versus Q=400M -r: Q = 20M -a 0.160 sec/rel yield 1.75/Q Q = 120M -a 0.229 sec/rel yield 1.32/Q Q = 400M -r 0.405 sec/rel yield 1.00/Q (tests done on 2kQ regions) Perhaps this job could be re-queued on the -a side from 20M up to generate the remaining relations after -r finishes to Q=400M? perhaps as 202_137_75a, and Sean can just zcat together the two dat files before filtering? Does sieving both sides produce more duplicates than normal?

2017-09-22, 09:19	#1151
swellman Jun 2012 11×281 Posts	Just keep in mind I'm using a Win 7 box. Copy command - no zcat. Not sure what the limit on file size in DOS environment. Merging two huge files on my machine could be problematic, though in all fairness I've not tried that for some years.