Polynomial Request Thread - Page 56

VBCurtis · 2016-05-17, 23:27

I meant only to answer Sean's query about the value of using msieve vs CADO. For anyone lacking a GPU, CADO is the clear choice for serious-effort composites (I dunno how low, best guess is 135+ digits). For feeding NFS@home, a GPU-day on msieve + a few core-days on CADO for a few different numbers will ensure the 14e queue has food, as well as helping us gather examples to compare CADO to msieve-GPU poly quality.

I'll do a GPU-day on this C173 starting at A1 = 100k to help the comparison. I'm personally more interested in comparing the software results than I am feeding the queue, else I'd knock out a few more polys with a day each. I might get to that in June, but my GPU is occupied most of the rest of May.

Dubslow · 2016-05-18, 05:09

My go at the C173 is taking longer than I thought. Call it 3 CPU-days (ETA 1). In sum it is the CADO-recommended amount (i.e. the amount it would do by default before moving to sieving).

XYYXF · 2016-05-19, 01:03

There's also another GNFS task, C188_140_113:

Code:

51928438462607404737043189983177088594484655472144336181756537025533635370879714014750880494339341212074869710914584937712249921999507347297047835234867294560314897640299447799219467293637

About 50% of t60 is done on it; a factor still may fall out, but the chance is quite low.

Dubslow · 2016-05-19, 04:26

Quote:

Originally Posted by XYYXF

C173_139_59

Code:

34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721

Thanks in advance :)

Here's the top 5 from CADO, its default amount of poly select, which was around ~3 CPU-days of work.

Code:

n: 344791381226851287279185036034642635800505109743063349757290593755437104984269862201468187980362970210149169351930
33172364267380987223293686993196656448672279611520337389721
skew: 1335808.0
c0: 387791869088524171451616910842298125552
c1: -1627706106635766854677420302311924
c2: 713730807612923775332352144
c3: 2041990797545923660967
c4: -104446892249868
c5: 36412740
Y0: -1232221203350897612542968561876865
Y1: 18890353014002536254449
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.25e-13
# lognorm 53.56

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 2481152.0
c0: -881531622281417491395870708712728361326
c1: -12392525690562916109927028699694411
c2: -8312029032340810486750452390
c3: 4323698392044887335255
c4: 487091542375152
c5: 10904400
Y0: -2069250787590067620642783482094530
Y1: 7491126098956512078473
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.05e-13
# lognorm 54.65

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 1853952.0
c0: -139380690982113285720506233886229813276
c1: 5197873714511436671788252636824292
c2: 1983142702530420904278165027
c3: -2526062094908554469239
c4: 59653677406176
c5: 108390240
Y0: -1307220951121355119121205412023852
Y1: 13517108705672089306643
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.04e-13
# lognorm 53.83

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 1455616.0
c0: -115745058707370098538432558550505215680
c1: 298652439254901617888471729514716
c2: 1016773653184537392958501959
c3: -282009201658895889647
c4: -44645499217632
c5: 35390880
Y0: -1878354179709467208731760585017919
Y1: 38210866208909943630899
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.02e-13
# lognorm 52.45

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 2962432.0
c0: 27034531063894665041390513650106337674400
c1: 143089369195626432079051495049020
c2: -16601414383359955900319047221
c3: 1924689305709482140505
c4: 783667781984436
c5: 67606560
Y0: -1250664399385069472739216286946479
Y1: 3695455190218983607253
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.02e-13
# lognorm 55.03

If in a few days no one has started C188 then I'll take a look at running it through CADO. I want to finish my current aliquot reservation first.

Dubslow · 2016-05-19, 05:52

VBCurtis, would you be willing to post your post-sizeopt hits again?

VBCurtis · 2016-05-19, 06:20

Quote:

Originally Posted by XYYXF

C173_139_59

Code:

34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721

Thanks in advance :)

3 good hits: 2.284, 2.277, and 2.259:

Code:

# norm 8.340529e-17 alpha -8.102789 e 2.259e-13 rroots 3
skew: 22231851.29
c0: 341468976065539083006521353956210974903800
c1: 265933747287181821140876904379636886
c2: 1504349170112038840531037885
c3: -1199027668924516073000
c4: -98766825250716
c5: 186480
Y0: -2840434825591921734390591425784537
Y1: 37459280589289771

# norm 8.321140e-17 alpha -6.359217 e 2.277e-13 rroots 5
skew: 17585569.89
c0: -89128313116298989163444393858408494846559
c1: 75202604379131164073290965726848862
c2: -10206874843434891360498546029
c3: -317099171577993101178
c4: 25847512339768
c5: 631176
Y0: -2225774614421631677857620554320578
Y1: 16708833215071681

# norm 8.453764e-17 alpha -7.682575 e 2.284e-13 rroots 5
skew: 20319586.54
c0: -946671312506868304798583076639702524932320
c1: 164282522294623228910796877746727332
c2: 25232278688149571650909987124
c3: -2924808965524900601507
c4: -78496319901694
c5: 591480
Y0: -2254879132167183735483368748054947
Y1: 7530111939454049

VBCurtis · 2016-05-19, 06:24

Quote:

Originally Posted by Dubslow

VBCurtis, would you be willing to post your post-sizeopt hits again?

Sure, they're attached. Two runs, so the scores reset after ~250 hits. 16 hr and 7 hr, I think. coeff's 100k to 675k.

Dubslow · 2016-05-19, 07:29

The first 8 alone got up to 2.17e-13. I think it'll be tomorrow though before the other ~380 finish.

Dubslow · 2016-05-19, 21:31

Here's the top 5 from CADO rootopt on your sizeopted hits:

Code:

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 6637568.0
c0: -10889105601629125038806527496775758297015
c1: 11394399399846453857523054958424094
c2: -786715094480263088740519037
c3: -703239076628310093690
c4: 7667352370888
c5: 631176
Y0: -2225774614517886687789347578920984
Y1: 16708833215071681
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.3e-13
# lognorm 53.03

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 8491008.0
c0: -146500114627489758079463551405859004491616
c1: 17347541839693874654249464144579488
c2: 12731068356335916002076064142
c3: -85971437136839853161
c4: -118392230007954
c5: 248040
Y0: -2682915137335644264087417105448787
Y1: 7534975867913461
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.17e-13
# lognorm 54.05

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 4622336.0
c0: -14433906057375928283259782953527905013333
c1: 5317830654603905183592922583396638
c2: 4152384915058422060601512491
c3: -858815766179496366872
c4: -99566523285516
c5: 186480
Y0: -2840434825624049697788571708706504
Y1: 37459280589289771
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.15e-13
# lognorm 53.42

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 7096320.0
c0: -10404303178741585574018368104632579899608
c1: 6365249504589167973034103288255794
c2: 6245522594066143920887484511
c3: -319958453839703630624
c4: -59071893514338
c5: 232200
Y0: -2718559468207889648042986596515533
Y1: 52790030796386939
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.15e-13
# lognorm 53.26

n: 34479138122685128727918503603464263580050510974306334975729059375543710498426986220146818798036297021014916935193033172364267380987223293686993196656448672279611520337389721
skew: 11898880.0
c0: -495972291318941666734522590963240671574985
c1: 182713847873464853237864878583336211
c2: 21717241076344588983037967951
c3: -3047098926687697538859
c4: -77335904161894
c5: 591480
Y0: -2254879132164229092750901586670474
Y1: 7530111939454049
# MurphyE (Bf=1.00e+07,Bg=5.00e+06,area=1.00e+16) = 2.15e-13
# lognorm 54.87

It seems that one day of GPU-Msieve is appreciably better than 2 days of CPU-CADO for stage 1 and sizeopted hits.

I think we'd need to do some trial sieving to see if the comparable scores we got after rootopt are really comparable polys.

VBCurtis · 2016-05-19, 23:49

Well, the scores for my best (2.284), your best (2.25), and the best hybrid GPU/CADO (2.30) are within 2%. I agree about test-sieving, but I don't think "appreciably better" is accurate. 2% is noise, as test-sieving often shows same-score polys with 5% sec/rel differences. I don't have time this weekend to test-sieve, so I hope you do!

wombatman · 2016-05-20, 00:33

If one of you wouldn't mind teaching me how, I'd be willing to do some test-sieving. I've been meaning to learn how to properly do it anyway.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
GIMPS wiki account request thread	ixfd64	mersennewiki	169	2018-09-21 05:43
Polynomial Discriminant is n^k for an n-1 degree polynomial	carpetpool	Miscellaneous Math	14	2017-02-18 19:46
Lost Prime Raider password request thread	cheesehead	Forum Feedback	6	2009-07-28 13:02
Polynomial	R.D. Silverman	NFSNET Discussion	13	2005-09-16 20:07
Deutscher Thread (german thread)	TauCeti	NFSNET Discussion	0	2003-12-11 22:12

2016-05-17, 23:27	#606
VBCurtis "Curtis" Feb 2005 Riverside, CA 2·2,437 Posts	I meant only to answer Sean's query about the value of using msieve vs CADO. For anyone lacking a GPU, CADO is the clear choice for serious-effort composites (I dunno how low, best guess is 135+ digits). For feeding NFS@home, a GPU-day on msieve + a few core-days on CADO for a few different numbers will ensure the 14e queue has food, as well as helping us gather examples to compare CADO to msieve-GPU poly quality. I'll do a GPU-day on this C173 starting at A1 = 100k to help the comparison. I'm personally more interested in comparing the software results than I am feeding the queue, else I'd knock out a few more polys with a day each. I might get to that in June, but my GPU is occupied most of the rest of May.

2016-05-18, 05:09	#607
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3×29×83 Posts	My go at the C173 is taking longer than I thought. Call it 3 CPU-days (ETA 1). In sum it is the CADO-recommended amount (i.e. the amount it would do by default before moving to sieving).

2016-05-19, 01:03	#608
XYYXF Jan 2005 Minsk, Belarus 2⁴×5² Posts	There's also another GNFS task, C188_140_113: Code: 51928438462607404737043189983177088594484655472144336181756537025533635370879714014750880494339341212074869710914584937712249921999507347297047835234867294560314897640299447799219467293637 About 50% of t60 is done on it; a factor still may fall out, but the chance is quite low.

2016-05-19, 05:52	#610
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	VBCurtis, would you be willing to post your post-sizeopt hits again?

2016-05-19, 07:29	#613
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 3·29·83 Posts	The first 8 alone got up to 2.17e-13. I think it'll be tomorrow though before the other ~380 finish.

2016-05-19, 23:49	#615
VBCurtis "Curtis" Feb 2005 Riverside, CA 11412₈ Posts	Well, the scores for my best (2.284), your best (2.25), and the best hybrid GPU/CADO (2.30) are within 2%. I agree about test-sieving, but I don't think "appreciably better" is accurate. 2% is noise, as test-sieving often shows same-score polys with 5% sec/rel differences. I don't have time this weekend to test-sieve, so I hope you do!

2016-05-20, 00:33	#616
wombatman I moo ablest echo power! May 2013 13·137 Posts	If one of you wouldn't mind teaching me how, I'd be willing to do some test-sieving. I've been meaning to learn how to properly do it anyway.