Fast Breeding (guru management) - Page 91

swellman · 2017-04-05, 17:20

Quote:

Originally Posted by VBCurtis

I use alim/4 for GNFS and alim/6 for SNFS, subject to a minimum starting value of 5M and maximum starting value of 25M. For small (say, a core-week or less) projects, I use alim/3 for GNFS and alim/4 for SNFS.

BOINC seems to start at 20M pretty regularly, and I don't think it matters a whole lot whether one chooses 15M or 20M or 25M to start when alim/rlim are 100M+, so 20M has become a sort of de facto standard for all but the largest projects.

In the past, it was believed that the faster sec/rel times at small Q were illusions because of higher duplicate rates, but I think that has been debunked in the alim/5 to alim/2 region.

Thank you for the information. Yes, the higher duplicate rate at low Q was one of the warnings I'd read, but if it's a myth then so be it.

I'm curious - is there a quick rule of thumb for modeling the decreasing yield curve over a large range of Q for an accurate estimate of the sieving range? Or is that more art? I'm good at modeling such things but it can be time consuming, so if there's a quick-but-good-enough method then I'm all ears!

RichD · 2017-04-05, 19:56

P21.14943_13M.C206 is ready.

Code:

n: 89565020195209009565467987087523777190165220963732300333632805305657311963519775871355884908105015698333537785557847431074271876592471825692366680584078161538467475749662494429667299122846947799699591673579
# 149431854123332538041^13-1, difficulty: 242.09, skewness: 1.00, alpha: 3.10
# cost: 4.51385e+18, est. time: 2149.45 GHz days (not accurate yet!)
skew: 1.000
c6: 1
c5: 1
c4: -5
c3: -4
c2: 6
c1: 3
c0: -1
Y1: -149431854123332538041
Y0: 22329879026736935651195230967634712117682
m: 23565865290286569388823858250238266971263782880320796060116123560770899985421941989827510445870492077635584653040018904191599893740867088229038460402604998044704586084004555217759935764696304299260298567603
type: snfs
rlim: 120000000
alim: 120000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Trial sieve 5K blocks.

Code:

 Q   Yield
20M  12416
50M  10525
80M   8962

swellman · 2017-04-05, 20:07

C209_122_107 is nearing completion of t55 by yoyo@Home.

Code:

n: 45593358642996478364538112111174968061772662806395817785620327515809295803708534368841966179066170869564093890043051507627396767585413436241604695929929618429387862507149811619032959373562170721399964592363601
# 122^107+107^122, difficulty: 247.58, anorm: 2.36e+039, rnorm: -1.26e+047
# scaled difficulty: 248.87, suggest sieving rational side
# size = 8.302e-013, alpha = 0.000, combined = 1.075e-013, rroots = 0
type: snfs
size: 247
skew: 10.5727
c6: 1
c0: 1396778
Y1: -38696844624861790832365403138487376998001
Y0: 35848992283832616457430560986334756864
rlim: 250000000
alim: 250000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving gives the following yields

Code:

1.90 @Q=40M
1.64 @Q=100M
1.21 @Q=180M
1.22 @Q=250M
1.01 @Q=370M

Which leads me to believe the sieving range is 40-450M.

VBCurtis · 2017-04-08, 05:34

13*2^793-1:

Code:

# 13*2^793-1 difficulty: 240 
n: 245459994333326919443410901346097158174068960121046154069468560261450530982369518932964157573944452706561200577507849282111043339664498382619421553516587020088145546714743668816094603954681579
m: 5444517870735015415413993718908291383296
type: snfs
skew: 1.72
c6: 26
c0: -1
rlim: 110000000
alim: 110000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.7
alambda: 2.7

ECM'ed to half a t55.
Q of 20M to 120M should be sufficient for me to run the LA, same range as was run on 13*2^792-1. Avg yield is near 3.4, and 340M rels is enough to build a matrix for these numbers.
I'm not yet fully certain I have skew calculations correct; I did 26^(1/6) for this skew. Is that right, or should skew be the reciprocal of that?

fivemack · 2017-04-08, 10:22

Quote:

I'm not yet fully certain I have skew calculations correct; I did 26^(1/6) for this skew. Is that right, or should skew be the reciprocal of that?

Skew should be the reciprocal of that - to remember that, remember that GNFS polynomials have large skew, and have the constant term much larger than the leading coefficient.

Skew inverted, polynomial queued

fivemack · 2017-04-08, 10:32

Queued C209_122_107, 13_2_793m and C206_149xx041_13

VBCurtis · 2017-04-09, 19:29

Quote:

Originally Posted by fivemack

Skew should be the reciprocal of that - to remember that, remember that GNFS polynomials have large skew, and have the constant term much larger than the leading coefficient.

Skew inverted, polynomial queued

Thanks! Looks like inverting skew improved yield by 25%. Good thing I didn't even edit skew from 1.0x for the first couple years I factored these numbers; I would have edited it the wrong direction!

Downloading relations now.

Max0526 · 2017-04-10, 03:23

@VBCurtis @fivemack
You can get a bit higher E score if you set skew=0.75 instead of skew=0.58:

Code:

Msieve v. 1.53 (SVN unknown)
random seeds: 2db15a70 021c94d2
factoring 245459994333326919443410901346097158174068960121046154069468560261450530982369518932964157573944452706561200577507849282111043339664498382619421553516587020088145546714743668816094603954681579 (192 digits)
no P-1/P+1/ECM available, skipping
commencing number field sieve (192-digit input)
R0: -5444517870735015415413993718908291383296
R1: 1
A0: -1
A1: 0
A2: 0
A3: 0
A4: 0
A5: 0
A6: 26
skew 0.75, size 8.995e-012, alpha 0.805, combined = 5.873e-013 rroots = 2

Obtained from http://myfactors.mooo.com/ (Optimal Skew)

swellman · 2017-04-12, 17:25

Any more candidates for 14e? The queue is once again running dry.

I should have one posted later today but more are needed.

swellman · 2017-04-12, 18:46

C209_127_91 has nearly completed a full t55 (85.6% complete at the time of this posting) by yoyo@Home.

Code:

n: 26940995360358453137720369146025885289223140375197877977981639330991363717529250240230037034188430266450967981323191907750900799884059432559083059910290611501303668882237162459786057340251559518931079556646633
# 127^91+91^127, difficulty: 250.76, anorm: 2.15e+038, rnorm: 1.34e+047
# scaled difficulty: 252.22, suggest sieving rational side
# size = 1.141e-012, alpha = 0.000, combined = 1.346e-013, rroots = 0
type: snfs
size: 250
skew: 1.0571
c6: 91
c0: 127
Y1: -36062498658084837781704007862143
Y0: 137996870875659993023030601717979081222891
rlim: 240000000
alim: 240000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Some test sieving on blocks of 1000 Q, hope this is in a more user friendly form

Code:

yield: 2836, q=20001001 (0.49325 sec/rel)
yield: 1602, q=140001013 (0.65194 sec/rel)
yield: 1465, q=260001013 (0.88608 sec/rel)

RichD · 2017-04-12, 21:31

86353_47M.C185 is ready.

Code:

n: 28448940404495058815836669391775762163760283231206204352120421742419083638829799923679271303255141623796529822942824468873818526386945481411116502483777554312314040915730295977513616383
# 86353^47-1, difficulty: 236.94, skewness: 6.65, alpha: 0.00
# cost: 3.01467e+18, est. time: 1435.56 GHz days (not accurate yet!)
skew: 6.648
c6: 1
c0: -86353
Y1: -1
Y0: 3091857291722705477037327063091586156161
m: 3091857291722705477037327063091586156161
type: snfs
rlim: 134000000
alim: 134000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Test sieving 5K blocks.

Code:

 Q    Yield
20M   29018
60M   21919
100M  20312

2017-04-08, 05:34	#994
VBCurtis "Curtis" Feb 2005 Riverside, CA 4873₁₀ Posts	132^793-1: Code: # 132^793-1 difficulty: 240 n: 245459994333326919443410901346097158174068960121046154069468560261450530982369518932964157573944452706561200577507849282111043339664498382619421553516587020088145546714743668816094603954681579 m: 5444517870735015415413993718908291383296 type: snfs skew: 1.72 c6: 26 c0: -1 rlim: 110000000 alim: 110000000 lpbr: 32 lpba: 32 mfbr: 64 mfba: 64 rlambda: 2.7 alambda: 2.7 ECM'ed to half a t55. Q of 20M to 120M should be sufficient for me to run the LA, same range as was run on 132^792-1. Avg yield is near 3.4, and 340M rels is enough to build a matrix for these numbers. I'm not yet fully certain I have skew calculations correct; I did 26^(1/6) for this skew. Is that right, or should skew be the reciprocal of that? Last fiddled with by VBCurtis on 2017-04-08 at 05:34*

2017-04-08, 10:32	#996
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 6424₁₀ Posts	Queued C209_122_107, 13_2_793m and C206_149xx041_13 Last fiddled with by fivemack on 2017-04-08 at 10:34

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

2017-04-12, 17:25	#999
swellman Jun 2012 11×281 Posts	Any more candidates for 14e? The queue is once again running dry. I should have one posted later today but more are needed.