Fast Breeding (guru management) - Page 94

VBCurtis · 2017-04-28, 23:23

I have a ton of near-cunningham stuff ECM'ed to just over a t50; less than half of optimal, but enough to feed the queue in an emergency. I'll aim a dozen cores at ECM until the announcement, just in case.

RichD · 2017-04-30, 11:26

Another OPN from the t600 file.

Code:

n: 12504721975844856110697241554601196703762091371396736793144009607857269420234994875741440690988236657796360320862503133792893914145486370015708570016126408503686973395098094257
# 104281^47-1, difficulty: 240.87, skewness: 6.86, alpha: 0.00
# cost: 4.10446e+18, est. time: 1954.50 GHz days (not accurate yet!)
skew: 6.861
c6: 1
c0: -104281
Y1: -1
Y0: 13984324640414081187623635661438587861441
m: 13984324640414081187623635661438587861441
type: snfs
rlim: 134000000
alim: 134000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Trial sieving 5K blocks.

Code:

swellman · 2017-04-30, 17:23

C193_131_102 is ready for SNFS, to be sieved with 15e. I do not plan on suggesting any additional 15e candidates until that queue is considerably shorter.

Code:

n: 2740563180589160391621707648679867528583953126646972986009531107420925652080867961217598253167050281390162411820901451715308217666066971599869189147473353913754138736753526703892829478004132617
# 131^102+102^131, difficulty: 265.14, anorm: 2.02e+037, rnorm: 1.05e+050
# scaled difficulty: 267.25, suggest sieving rational side
# size = 3.437e-013, alpha = 0.000, combined = 5.296e-014, rroots = 0
type: snfs
size: 265
skew: 2.1616
c6: 1
c0: 102
Y1: -985398793384554108247251712700460611
Y0: 154597967077587985745381119925836646348488704
rlim: 134000000
alim: 134000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.7
alambda: 2.7

Test sieving, Q in blocks of 5000:

Code:

Q=20M   total yield: 6795, q=20005003 (1.06367 sec/rel)
Q=100M  total yield: 5735, q=100005001 (1.39100 sec/rel)
Q=200M  total yield: 4616, q=200005007 (1.76671 sec/rel)
Q=300M  total yield: 3803, q=300005039 (2.12377 sec/rel)

suggesting a range of 20M-270M for Q.

RichD · 2017-05-01, 02:23

OPN from the t600 file is ready.

Code:

n: 3582123724009125350979135532253668176917146315309069102903598732280603715909936623363796377970580002623282837233544467878001808051279400221966150804199215087404769081158554191159143409953966039
# 25125200742702138527^13-1, difficulty: 232.80, skewness: 1.00, alpha: 3.10
# cost: 2.17123e+18, est. time: 1033.92 GHz days (not accurate yet!)
skew: 1.000
c6: 1
c5: 1
c4: -5
c3: -4
c2: 6
c1: 3
c0: -1
Y1: -25125200742702138527
Y0: 631275712361080093443627372579097729730
m: 2942309812454780857174959580625797121082650591780520548459155615893205344204950445727368887008237009235463040328040082883725470745346026159304591613666306253853561846670793772820592220799691604
type: snfs
rlim: 80000000
alim: 80000000
lpbr: 31
lpba: 31
mfbr: 62
mfba: 62
rlambda: 2.6
alambda: 2.6

Trial sieving 5K blocks.

Code:

  Q   Yield
 20M 19,036
 60M 14,468
100M 12,859

wombatman · 2017-05-01, 13:09

Quick question: what's an appropriate alim/rlim for the C207? I want to make sure I test-sieve appropriate Q-ranges before posting it up.

VBCurtis · 2017-05-01, 14:53

I would try 536M on both sides first (power-of-two, rounded), and then something larger like 800M. I don't think we want to expect to use Q larger than 1.5 * rlim, which is why I fear 536M may not be enough (we may need Q above 1e9). The choice depends on how yields look, really- if sieving 100M to 800M is going to be enough, 536M would be quicker.

swellman · 2017-05-01, 16:25

Don't forget to check 3LPs on your best candidate(s). See http://www.mersenneforum.org/showpos...&postcount=952 for a refresher if you need it.

A rule of thumb I've come across on this forum is that a/rlim < 2^(lpb-4), which in this case is ~536M as VBCurtis points out. I do not know what happens if you increase a/rlim above this threshold.

wombatman · 2017-05-01, 17:38

Quote:

Originally Posted by VBCurtis

I would try 536M on both sides first (power-of-two, rounded), and then something larger like 800M. I don't think we want to expect to use Q larger than 1.5 * rlim, which is why I fear 536M may not be enough (we may need Q above 1e9). The choice depends on how yields look, really- if sieving 100M to 800M is going to be enough, 536M would be quicker.

Thanks. YAFU chose 100M, which I guessed was far too low, but I didn't know how much higher I would need to travel.

Quote:

Originally Posted by swellman

Don't forget to check 3LPs on your best candidate(s). See http://www.mersenneforum.org/showpos...&postcount=952 for a refresher if you need it.

A rule of thumb I've come across on this forum is that a/rlim < 2^(lpb-4), which in this case is ~536M as VBCurtis points out. I do not know what happens if you increase a/rlim above this threshold.

Thanks. Yeah, I've tried 3LPs and increasing the large prime bound to 33 (for algebraic only, thus far). Both 3LPs and increasing the bound led to increases in both yield and speed. Now that I have a proper limit, I can do better yield/speed checks on the two best polynomials, which are pretty close in terms of viability.

swellman · 2017-05-01, 21:43

C208_147_50 and C198_149_50 are ready for NFS. Both can be run on 14e, though the second (C198_149_50) may be better suited for 15e/31.

Code:

n: 1202488500242124574094685588829895536874010314117513572280895082444267256993786380990563199163017709568011379910135080421260357568548724831255581892733119854780946936542079537679417901789297803205092709289361
# 147^50+50^147, difficulty: 254.44, anorm: 1.19e+038, rnorm: 5.77e+047
# scaled difficulty: 256.05, suggest sieving rational side
# size = 1.111e-012, alpha = 0.000, combined = 1.300e-013, rroots = 0
type: snfs
size: 254
skew: 1.0662
c6: 49
c0: 72
Y1: -1526288802270065127
Y0: 596046447753906250000000000000000000000000
rlim: 240000000
alim: 240000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving with Q in blocks of 5000

Code:

Q=20M   total yield: 12538, q=20005003 (0.41425 sec/rel)
Q=100M  total yield: 10140, q=100005001 (0.54565 sec/rel)
Q=200M  total yield: 8314, q=200005007 (0.75086 sec/rel)
Q=300M  total yield: 7636, q=300005039 (0.85505 sec/rel)

suggesting a range of 20-300M for Q

Code:

n: 355865490586785959066984347712728983905427858154327229480903647077634391948558871491545608798250431936333826364100726589278827574595392064971455099743504501738466786994504247194028826825300041338041
# 149^50+50^149, difficulty: 254.85, anorm: 2.11e+039, rnorm: 9.34e+047
# scaled difficulty: 256.29, suggest sieving rational side
# size = 4.334e-013, alpha = 0.000, combined = 6.651e-014, rroots = 0
type: snfs
size: 254
skew: 10.1755
c6: 1
c0: 1110050
Y1: -242935032749128801
Y0: 2980232238769531250000000000000000000000000
rlim: 240000000
alim: 240000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving with Q in blocks of 5000

Code:

Q=20M   total yield: 8015, q=20005003 (0.59273 sec/rel)
Q=100M  total yield: 6700, q=100005001 (0.77428 sec/rel)
Q=200M  total yield: 5468, q=200005007 (1.11010 sec/rel)
Q=300M  total yield: 4908, q=300005039 (1.29205 sec/rel)

suggesting a range of 20-510M for Q - perhaps run as 15e?

fivemack · 2017-05-02, 10:20

C208_147_50 queued on 14e. C198_149_50 queued on 15e

(some timings for C198_149_50 on the two sievers)

Code:

14e
total yield: 13740, q=50010001 (0.28664 sec/rel) 
total yield: 13114, q=100010017 (0.34573 sec/rel) 
total yield: 11497, q=150010001 (0.41877 sec/rel) 
total yield: 10762, q=200010011 (0.49476 sec/rel) 
total yield: 9613, q=250010011 (0.57307 sec/rel) 

15e
total yield: 28944, q=50010001 (0.24353 sec/rel)
total yield: 27219, q=100010017 (0.28596 sec/rel) 
total yield: 24011, q=150010001 (0.33472 sec/rel) 
total yield: 22313, q=200010011 (0.37820 sec/rel) 
total yield: 20062, q=250010011 (0.42392 sec/rel)

swellman · 2017-05-04, 11:25

C213_142_74 is ready for NFS. Sieve using 14e.

Code:

n: 164301310715231570211687638819270453919811804856418781178265316562254413712075720905065881225107134715664949546263235080480716365263042915251485594025158238272607526746828387151574098692873919436153635237018303741
# 142^74+74^142, difficulty: 242.85, anorm: 1.76e+033, rnorm: -2.58e+054
# scaled difficulty: 246.38, suggest sieving rational side
# size = 3.236e-017, alpha = 0.021, combined = 1.178e-013, rroots = 1
type: snfs
size: 242
skew: 15.0713
c5: 1
c0: 777592
Y1: -665331702192301904671965482114744277708773138432
Y0: 5873205959385493353867330551
rlim: 240000000
alim: 240000000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.8
alambda: 2.8

Test sieving with Q in blocks of 5000

Code:

Q=20M   total yield: 6865, q=20005003 (1.07902 sec/rel)
Q=100M  total yield: 8128, q=100005001 (0.99405 sec/rel)
Q=200M  total yield: 8018, q=200005007 (1.22869 sec/rel)
Q=300M  total yield: 7196, q=300005039 (1.41274 sec/rel)

suggesting a sieving range for Q of 20-360M

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

2017-04-28, 23:23	#1024
VBCurtis "Curtis" Feb 2005 Riverside, CA 11·443 Posts	I have a ton of near-cunningham stuff ECM'ed to just over a t50; less than half of optimal, but enough to feed the queue in an emergency. I'll aim a dozen cores at ECM until the announcement, just in case.

2017-05-01, 13:09	#1028
wombatman I moo ablest echo power! May 2013 13×137 Posts	Quick question: what's an appropriate alim/rlim for the C207? I want to make sure I test-sieve appropriate Q-ranges before posting it up.

2017-05-01, 14:53	#1029
VBCurtis "Curtis" Feb 2005 Riverside, CA 11·443 Posts	I would try 536M on both sides first (power-of-two, rounded), and then something larger like 800M. I don't think we want to expect to use Q larger than 1.5 * rlim, which is why I fear 536M may not be enough (we may need Q above 1e9). The choice depends on how yields look, really- if sieving 100M to 800M is going to be enough, 536M would be quicker.

2017-05-01, 16:25	#1030
swellman Jun 2012 11×281 Posts	Don't forget to check 3LPs on your best candidate(s). See http://www.mersenneforum.org/showpos...&postcount=952 for a refresher if you need it. A rule of thumb I've come across on this forum is that a/rlim < 2^(lpb-4), which in this case is ~536M as VBCurtis points out. I do not know what happens if you increase a/rlim above this threshold.