Poly select and planning for 2,2330M - Page 5

Max0526 · 2019-07-19, 03:26

Quote:

Originally Posted by VBCurtis

4.3M thread-seconds later:

Code:

n: 2019167005807456047760270226654711862551070622459320820768633837520963439033850414840732672810544781990446046$
skew: 646896.496
c0: -279567254431149509523770958289009390033170
c1: 25083343894490643237108315247666493509
c2: 768310355319881904792798032705966
c3: -412472666866271946113375269
c4: -1967356887803328785216
c5: 143279560013700
c6: -74178720
Y0: -8433804753284135081668111590062466
Y1: 6851718171937808451398387
# MurphyE (Bf=3.436e+10,Bg=3.436e+10,area=1.718e+16) = 2.57e-08

cownoise: skew 726799.19110 murphy-E 1.29348321e-15

Edit: Next attempt will be 1.05M to 11M with incr=4620, nq=46656

Edit2: deg 6 scores on the C207 had to be about 20% higher to get the same sieve performance as a deg 5; so this poly (before spin) may perform similar to the 1.05-1.1 deg 5 polys posted previously. This is *not* a sign to move all our search to deg 6! One hopes that if we keep finding 1.09's, Max will spin one to 1.20ish and that will sieve better than a 1.3x or low 1.4 deg 6.

This one spins up a bit:

Code:

Y0: -8433824184756870697292879755887998
Y1: 6851718171937808451398387
c0: 50157313234096582642596433071178509682080150
c1: -131070603403140917396598444903577641749
c2: -771635080061282454139704797834642
c3: 390143412665212478133766565
c4: 1969397541140278648016
c5: -144541785113220
c6: 74178720
skew: 856429.18891
# lognorm 59.44, E 49.58, alpha -9.87 (proj -2.99), 2 real roots
# MurphyE = 1.31319727e-15

Max0526 · 2019-07-22, 22:39

The three Gimarel's polys from posts 36 and 43 unfortunately don't spin up, sorry.
@swellman: Thank you for the PM.

Max0526 · 2019-07-23, 00:52

Quote:

Originally Posted by VBCurtis

3.9Megathreadseconds pass:

Code:

n: 2019167005807456047760270226654711862551070622459320820768633837520963439033850414840732672810544781990446046413975756717396304999711509128197909162150081255541065710228$
skew: 1978626.282
c0: -67099649020847517071718063977040322149214160
c1: -405395177732383477636411051561488679066
c2: -318115883936383226544178265024933
c3: 262115157317139500440001570
c4: -12429068846689584746
c5: 6799591787372
c6: 4402860
Y0: -5977564625917306585710641319793239
Y1: 20598621408885914689078117
# MurphyE (Bf=3.436e+10,Bg=3.436e+10,area=1.718e+16) = 2.67e-08

cownoise: skew 2139407.13605 Murphy-E 1.38866049e-15

This hit came from the best-scoring poly after size-opt.
I'm taking a break from this poly select for some time; hopefully we'll have another good deg 5 candidate or two, and Max will stop by to try some spin.

This would be our best in degree 6 now (there are too many above 1.3 for this c6):

Code:

Y0: -11945075559091849769723496148553586
Y1: 20598621408885914689078117
c0: 13501574082332817040603148668654622503347773320
c1: -10156881097460087806945005627043407706466
c2: -507376313380945306542548696254673
c3: 510623497111647728073268516
c4: -199120054352596466
c5: 6623191339726
c6: 1100715
skew: 5701047.29482
# lognorm 59.44, E 49.91, alpha -9.53 (proj -1.26), 4 real roots
 # MurphyE = 1.39786197e-15

and the runners-up from other families:

Code:

Y0: -5969974157519860579463204168147441
Y1: 41197242817771829378156234
c0: 4948432546369891708467662655357568688304015
c1: -46198486877243578879304102264648068968
c2: -35388008013238892347531774203699
c3: 514868288492261548046580388
c4: 36267187927953793976
c5: 132273253955296
c6: 70445760
skew: 757624.78917
# lognorm 58.42, E 49.79, alpha -8.63 (proj -2.41), 4 real roots
# MurphyE = 1.34461710e-15

Code:

Y0: -5972537779545924884861748074276793
Y1: 20598621408885914689078117
c0: 3375393520583204260150787167163655625836943330
c1: -5078440548730043903472502813521703853233
c2: -507376313380945306542548696254673
c3: 1021246994223295456146537032
c4: -796480217410385864
c5: 52985530717808
c6: 17611440
skew: 2848696.89912
# lognorm 60.13, E 50.37, alpha -9.76 (proj -2.88), 4 real roots
# MurphyE = 1.34452946e-15

Gimarel · 2019-07-30, 14:32

Code:

# norm 1.547898e-20 alpha -7.709420 e 1.130004e-15 rroots 5
skew: 224966051.12
c0: -46523542304709920815932116270534204096078149944102
c1: 1108746025557925802550149908112273592528253
c2: 1247251566549430335138785462051674
c3: -54819856537027331130006773
c4: -55070735672396492
c5: 372372000
Y0: -16108126957193005421825842611692543447840
Y1: 161831307530189843309

Gimarel · 2019-08-01, 15:11

Code:

# norm 1.528182e-20 alpha -7.971412 e 1.105489e-15 rroots 5
skew: 177022801.68
c0: 18141221687315584816029160568403275969140566329420
c1: 5254194658276600631654506562253982994736
c2: -9469476947909136029668922395444289
c3: 49282705511088036410465805
c4: 365507147213333748
c5: 125405280
Y0: -20025219395652712795847739261487881046851
Y1: 194209346463474127813

VBCurtis · 2019-08-05, 08:18

Looks like we're at 1.39 for deg 6, and 1.18 (with a couple of 1.17's and 1.16's) for deg 5.

I should have some time to begin test-sieving in a week or two, at least to try to determine what degree we should focus a last burst of poly select on.

2330L may be in postprocessing before the end of August. I suspect other contributors share my interest in doing some other personal work before doing another team-CADO-sieve; I'd be willing to start the next CADO team sieve around the time 2330L yields factors (2, perhaps 3 months after 2330L sieve finishes? Say, mid-Nov?).

RDS was pretty adamant that this number should be handled by NFS@home; given the schedule above, I think we wouldn't be able to finish CADO sieving before Spring begins, so I think I agree with him. Does anyone with a 48GB+ machine want to sign up for solving a ~50M matrix for 2330M? It might even fit on a 32GB machine, if that machine isn't used for "regular" work.
If so, we should aim to complete poly select and test-sieving in the next 4-6 weeks?

Max0526 · 2019-08-06, 19:11

Quote:

Originally Posted by Gimarel

Code:

# norm 1.547898e-20 alpha -7.709420 e 1.130004e-15 rroots 5
skew: 224966051.12
c0: -46523542304709920815932116270534204096078149944102
c1: 1108746025557925802550149908112273592528253
c2: 1247251566549430335138785462051674
c3: -54819856537027331130006773
c4: -55070735672396492
c5: 372372000
Y0: -16108126957193005421825842611692543447840
Y1: 161831307530189843309

Quote:

Originally Posted by Gimarel

Code:

# norm 1.528182e-20 alpha -7.971412 e 1.105489e-15 rroots 5
skew: 177022801.68
c0: 18141221687315584816029160568403275969140566329420
c1: 5254194658276600631654506562253982994736
c2: -9469476947909136029668922395444289
c3: 49282705511088036410465805
c4: 365507147213333748
c5: 125405280
Y0: -20025219395652712795847739261487881046851
Y1: 194209346463474127813

I can't spin any of the two up, sorry.

SethTro · 2019-08-07, 05:10

Quote:

Originally Posted by VBCurtis

Looks like we're at 1.39 for deg 6, and 1.18 (with a couple of 1.17's and 1.16's) for deg 5.

I should have some time to begin test-sieving in a week or two, at least to try to determine what degree we should focus a last burst of poly select on.

2330L may be in postprocessing before the end of August. I suspect other contributors share my interest in doing some other personal work before doing another team-CADO-sieve; I'd be willing to start the next CADO team sieve around the time 2330L yields factors (2, perhaps 3 months after 2330L sieve finishes? Say, mid-Nov?).

RDS was pretty adamant that this number should be handled by NFS@home; given the schedule above, I think we wouldn't be able to finish CADO sieving before Spring begins, so I think I agree with him. Does anyone with a 48GB+ machine want to sign up for solving a ~50M matrix for 2330M? It might even fit on a 32GB machine, if that machine isn't used for "regular" work.
If so, we should aim to complete poly select and test-sieving in the next 4-6 weeks?

I have a 64GB machine that might do the job. How long is the matrix expected to take (on 32 threads)?

VBCurtis · 2019-08-07, 06:52

Quote:

Originally Posted by SethTro

I have a 64GB machine that might do the job. How long is the matrix expected to take (on 32 threads)?

I solved a C206 matrix from NFS@home in about 7 weeks on 10 threads of the machine that's hosting 2330L. msieve's matrix-solving process doesn't benefit much from hyperthreading, and is slowed enough by splitting over two sockets that I stick to solving on one socket (I have not achieved joy on the MPI flavor of msieve, which *does* split nicely over multiple sockets; I compiled it but my first try to run it, on that C206 job, produced errors and I haven't tried again since).

Anyway, that C206 matrix was around 40M dimensions; data is very thin up at C210, but a matrix of 50 to 55M would take 50% to 80% longer than my C206 did. So, 3ish months on 8 cores, maybe 2 months on all 16 cores (or 6 weeks if you find MPI success).

fivemack · 2019-08-07, 22:27

pineapple (i9/7940X, 14 cores, 64GB memory) stands ready to assist.

It has run a 47.62M matrix at density 114 in just under four weeks.

(VBITS=256 is very effective on Skylake hardware when you build with a modern gcc)

VBCurtis · 2019-08-08, 02:26

Excellent! Perhaps someone previously in contact with frmky can ask if early Sept is acceptable to queue this number; we should be finished with poly select within a month.

2019-07-22, 22:39	#46
Max0526 "Max" Jun 2016 Toronto 713₁₀ Posts	C210 polys The three Gimarel's polys from posts 36 and 43 unfortunately don't spin up, sorry. @swellman: Thank you for the PM.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Poly select and test-sieving for RSA232	VBCurtis	Operation Kibibit	25	2020-01-07 01:57
YAFU Poly Select Deadline	amphoria	YAFU	22	2016-09-17 09:47
msieve poly select: choosing Stage1norm	VBCurtis	Msieve	0	2016-04-11 21:33
Starting NFS skipping poly select	jux	YAFU	5	2016-01-02 01:01
big job planning	henryzz	Cunningham Tables	16	2010-08-07 05:08

2019-08-05, 08:18	#50
VBCurtis "Curtis" Feb 2005 Riverside, CA 4606₁₀ Posts	Looks like we're at 1.39 for deg 6, and 1.18 (with a couple of 1.17's and 1.16's) for deg 5. I should have some time to begin test-sieving in a week or two, at least to try to determine what degree we should focus a last burst of poly select on. 2330L may be in postprocessing before the end of August. I suspect other contributors share my interest in doing some other personal work before doing another team-CADO-sieve; I'd be willing to start the next CADO team sieve around the time 2330L yields factors (2, perhaps 3 months after 2330L sieve finishes? Say, mid-Nov?). RDS was pretty adamant that this number should be handled by NFS@home; given the schedule above, I think we wouldn't be able to finish CADO sieving before Spring begins, so I think I agree with him. Does anyone with a 48GB+ machine want to sign up for solving a ~50M matrix for 2330M? It might even fit on a 32GB machine, if that machine isn't used for "regular" work. If so, we should aim to complete poly select and test-sieving in the next 4-6 weeks?

2019-08-07, 22:27	#54
fivemack (loop (#_fork)) Feb 2006 Cambridge, England 7·911 Posts	pineapple (i9/7940X, 14 cores, 64GB memory) stands ready to assist. It has run a 47.62M matrix at density 114 in just under four weeks. (VBITS=256 is very effective on Skylake hardware when you build with a modern gcc)

2019-08-08, 02:26	#55
VBCurtis "Curtis" Feb 2005 Riverside, CA 2·7²·47 Posts	Excellent! Perhaps someone previously in contact with frmky can ask if early Sept is acceptable to queue this number; we should be finished with poly select within a month.