Fast Breeding (guru management) - Page 107

fivemack · 2017-09-28, 16:26

Queued C226_141_59 on 15e and C213_142_112 and C157_11040_10066 on 14e

VBCurtis · 2017-09-28, 16:55

Quote:

Originally Posted by swellman

Suggesting a sieving range of 20M-255M with a target # relations=520M.

Note that when comparing 15e to 14e, 15e will produce fewer duplicate relations, so a 15e job will need fewer raw relations than the same job run on 14e. I bet 460-470M rels is enough here, while 500-520M would be enough for nearly any 15/32 job we could run.

450M 15e rels is often as good as 475M 14e rels; that is, the duplication rate is often 2/3rds. I'd like some confirmation of this; perhaps we could report raw vs unique relations for some time on the big 14e and all 15e jobs?

swellman · 2017-09-28, 17:25

Quote:

Originally Posted by VBCurtis

Note that when comparing 15e to 14e, 15e will produce fewer duplicate relations, so a 15e job will need fewer raw relations than the same job run on 14e. I bet 460-470M rels is enough here, while 500-520M would be enough for nearly any 15/32 job we could run.

450M 15e rels is often as good as 475M 14e rels; that is, the duplication rate is often 2/3rds. I'd like some confirmation of this; perhaps we could report raw vs unique relations for some time on the big 14e and all 15e jobs?

I bow to your experience on the dup rate - 460M+ for the upper end of Q range seems a good thing. No need for extra work! Didn't realize the effect was that big.

Also I will report raw/unique relations for all my jobs from here on out. (No more 15/31 left in my queue, and only seven 14/31 jobs in the pipe).

VBCurtis · 2017-09-28, 19:21

I used to report both raw and unique, then I started using remdups so now I must pay attention to how many rels I downloaded before remdups-ing the file. This discussion is a reminder to myself that duplication rates matter, and the 14e queue is in new territory for nearly everyone here so data is scarce.

Alfred · 2017-09-30, 07:38

567661^41-1

I'm factoring this number, too.
My ETA is 01.10.2017.

Shall I proceed or stop ?

PS: I overlooked that the number was queued on 27.09.2017 (?)

RichD · 2017-09-30, 08:11

Quote:

Originally Posted by Alfred

567661^41-1

I'm factoring this number, too.
My ETA is 01.10.2017.

Shall I proceed or stop ?

PS: I overlooked that the number was queued on 27.09.2017 (?)

Go ahead and finish it up. You will complete it before any post-processor will get to it Report the factors in the reservation thread so the entry can be removed from the queue.

swellman · 2017-09-30, 20:51

queued C215_147_53 is ready for SNFS as a 15e/32 job. Poly follows.

Code:

n: 10076208994859598682531631559632476483149792580846464207011654996964452486511085469244349665103204144457374964397184525525959111926947665668462480126249582933781635804479872947702733600393007111639304416333391508707
# 147^53+53^147, difficulty: 253.47, anorm: 9.36e+039, rnorm: -9.87e+047
# scaled difficulty: 254.81, suggest sieving rational side
# size = 3.242e-013, alpha = 0.000, combined = 5.409e-014, rroots = 0
type: snfs
size: 253
skew: 1.2556
c6: 148877
c0: 583443
Y1: 1526288802270065127
Y0: -241335311011519234780052665404754645838881
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 5k

Code:

Q=20M    rels=14562
Q=80M    rels=10995
Q=150M   rels=9660
Q=200M   rels=9485
Q=250M   rels=8655

Suggesting a sieving range for Q of 20M-240M with a target # rels=470M.

swellman · 2017-10-02, 10:40

C224_122_119 is ready for SNFS as a 14e/32 job. Poly follows, sieve on the rational side.

Code:

n: 13307877300925386685345938840329367328291391900290267316907953921852040681608789787186937952056280570334468243036501611555199039785149719157777200413911511801319547784384281194429131286718282393950678213312628628743260334049
type: snfs
size: 254
skew: 10.954
c6: 1
c0: 1727642
Y1: 324294234694341316421188266002423799213601
Y0: -533576401152564663352396469720606521163776
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:

Q=20M    rels=16088
Q=80M    rels=12470
Q=150M   rels=11521
Q=250M   rels=9877
Q=350M   rels=8915
Q=450M   rels=7653

Suggesting a sieving range for Q of 20M-490M with a target # rels=480M

swellman · 2017-10-02, 17:51

C227_135_76 is ready for SNFS as a 14e/32 job.

Code:

n: 71214437902112375896060657437728460097095202290204741636386555474729246128282657237134079404186626444618087412420357520216260587061572146926249192161085361782512091881680170848064795781912935659844494761770661911855612843638311
# 135^76+76^135, difficulty: 253.91, anorm: 2.32e+031, rnorm: 3.71e+056
# scaled difficulty: 258.11, suggest sieving rational side
# size = 2.283e-017, alpha = 0.000, combined = 8.004e-014, rroots = 1
type: snfs
size: 253
skew: 2.6673
c5: 1
c0: 135
Y1: -90158468827845266682220458984375
Y0: 605294867755151978358116750555481286863651157835776
rlim: 268000000
alim: 268000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving on -r side with Q in blocks of 2K

Code:

Q=20M   rels=2276
Q=80M   rels=2531
Q=150M  rels=2841
Q=250M  rels=2978
Q=350M  rels=2448

Suggesting a sieving range for Q of 20M-380M with a target # rels=480M.

###############################

C227_140_61 is ready for SNFS as a 14e/32 job.

Code:

n: 10635572682307117167317953922926507675605717270731798946034434606605400168679923192857937714017057118456560683668735671068538016904190817593187769145392166670529533862425660587135557161772941230606817283786036463345241928211437
# 140^61+61^140, difficulty: 249.95, anorm: 2.37e+031, rnorm: 5.95e+055
# scaled difficulty: 254.01, suggest sieving rational side
# size = 3.290e-017, alpha = 1.145, combined = 1.057e-013, rroots = 1
type: snfs
size: 249
skew: 2.6867
c5: 1
c0: 140
Y1: -56693912375296000000000000
Y0: 97551820892064552722266841889541788960852432978481
rlim: 268000000
alim: 268000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 62
rlambda: 2.8
alambda: 2.8

Test sieving on the -r side with Q in blocks of 2K

Code:

Q=20M    rels=2756
Q=80M    rels=2936
Q=150M   rels=3213
Q=250M   rels=3354
Q=350M   rels=2836

Suggesting a sieving range for Q of 20M-330M with target # rels=480M.

VBCurtis · 2017-10-02, 21:36

Sean-
Next time you test-sieve a candidate that stretches 14e (say, Q over 400M looks likely), try a test with rlim (assume sieving r side) of standard 268M, but alim (non-sieving side) of 400M or even 500M. This should improve yield a fair bit, while reducing the usual speed penalty from exceeding 268M factor-base bound.

This is likely to improve yield, but also result in a flatter sec/rel & yield curve at higher Q. sec/rel may get worse at lower Q, but hopefully the reduction in Q-range compensates. Worth a test-sieve, at least.

swellman · 2017-10-02, 21:41

Quote:

Originally Posted by VBCurtis

Sean-
Next time you test-sieve a candidate that stretches 14e (say, Q over 400M looks likely), try a test with rlim (assume sieving r side) of standard 268M, but alim (non-sieving side) of 400M or even 500M. This should improve yield a fair bit, while reducing the usual speed penalty from exceeding 268M factor-base bound.

This is likely to improve yield, but also result in a flatter sec/rel & yield curve at higher Q. sec/rel may get worse at lower Q, but hopefully the reduction in Q-range compensates. Worth a test-sieve, at least.

Tip noted. Never tried it but will from now on. I do attempt 3 LPs, poly deg, etc on all such jobs but never a lopsided r/alim. Thank you.

ETA: I can go back and try this technique on my recently proposed 14/32 jobs.

2017-09-30, 07:38	#1171
Alfred May 2013 Germany 5×17 Posts	567661^41-1 I'm factoring this number, too. My ETA is 01.10.2017. Shall I proceed or stop ? PS: I overlooked that the number was queued on 27.09.2017 (?) Last fiddled with by Alfred on 2017-09-30 at 07:40

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

2017-09-28, 16:26	#1167
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 1100100011000₂ Posts	Queued C226_141_59 on 15e and C213_142_112 and C157_11040_10066 on 14e

2017-09-28, 19:21	#1170
VBCurtis "Curtis" Feb 2005 Riverside, CA 2×2,437 Posts	I used to report both raw and unique, then I started using remdups so now I must pay attention to how many rels I downloaded before remdups-ing the file. This discussion is a reminder to myself that duplication rates matter, and the 14e queue is in new territory for nearly everyone here so data is scarce.

2017-10-02, 21:36	#1176
VBCurtis "Curtis" Feb 2005 Riverside, CA 2·2,437 Posts	Sean- Next time you test-sieve a candidate that stretches 14e (say, Q over 400M looks likely), try a test with rlim (assume sieving r side) of standard 268M, but alim (non-sieving side) of 400M or even 500M. This should improve yield a fair bit, while reducing the usual speed penalty from exceeding 268M factor-base bound. This is likely to improve yield, but also result in a flatter sec/rel & yield curve at higher Q. sec/rel may get worse at lower Q, but hopefully the reduction in Q-range compensates. Worth a test-sieve, at least.