SNFS targets which need more ECM
 

Here we'll collect potential SNFS targets which need some more ECM work. Usually ECM is stopped when the expected factor size is about 22% of SNFS difficulty.

There's a small table with work-to-do information (r means reserved):
[CODE] Composite | 300M |
-------------|--------|
C256_127_121 | done | fivemack completed 10240@3e8 4/Feb/17 0145
C278_135_128 | done | fivemack completed 20920@3e8 6/Mar/17 0349
C237_139_126 | r(2)/30k | fivemack reserved 20/May/17 for 30720@3e8
C277_148_81 | factor found | fivemack reserved 1/May/17 for 25600@3e8; completed 3/5/17
C228_150_142 | 0/20k | fivemack reserved 3/5/17 for 25600@3e8; completed 1/6/17
[/CODE]

As soon as ECM is done on some target, it will be moved to the [url=http://www.mersenneforum.org/showthread.php?t=19352]nearby thread[/url].

Any ECM help will be greatly appreciated :-)

I will run some ECM on C237_142_49 @B1=11e7. Targeting 5000 curves.

I must withdraw my reservation. Smoked one computer (it was old) and repairs make no economic sense. Looking to replace it but that might take a couple of months.

I'll do the 5k curves. It'll take a little bit since I'm doing GPU+CPU, but I'll update when they finish. :smile:

Stage 1 on GPU is finished. Stage 2 will take ~60 hours to complete, assuming no factor is found. :smile:

C237_142_49 held up. No factor found after 5632 curves at B1=110e6.

OK, I'm doing some test sieving now and will feed it to nfs@home

May I request that

C273_150_142
C277_148_81
C278_135_128

be added to this list. They're all of a size where nfs@home 15e will take two months to sieve, so probably do want substantial (say 20k @ 3e8, which is about half a t60 or four weeks on one GPU) ECM. Probably a job for yoyo@home.

I'll reserve C284_135_127 and run it for four weeks on my one GPU.

OK :)

Please note that C273_150_142 and C277_148_81 are still waiting for t50 in yoyo's queue: [url]http://www.rechenkraft.net/yoyo/download/download/stats/ecm/xy/wu_status[/url]

4160 curves done at 3e8 on C284_135_127 with no factor found. 16224 to go. Probably another three weeks.

Request that C175_133_61 be added to the list. I will be running 6000 curves @B1=11e7 on it for eventual queuing to NFS@Home (14e).

Andrey: why isn't the work defined by the top post on this list being queued up on yoyo@home?

Completing 7610@43M on numbers which are 300-digit SNFS difficulty and unlikely to be sieved this decade is not obviously a higher priority than getting tractable numbers ready for processing. Running the appropriate ECM on a single individual's resources is going to take a lot longer than sieving the number on nfs@home resources.

There are few t50 tasks remaining. Meanwhile I feed yoyo@home with some heavier tasks to get ready GNFS targets (currently there are two C178's and four C181's, fitting 15e I hope). I'll take into account SNFS/GNFS difficulty to build a t55 queue for yoyo@home.

[QUOTE=swellman;437344]Request that C175_133_61 be added to the list. I will be running 6000 curves @B1=11e7 on it for eventual queuing to NFS@Home (14e).[/QUOTE]

ECM completed with no factors found. I'll request it for NFS@Home sieving in 14e queue, along with a few other stragglers.

Can we add the following to the table? Please note I am not reserving any of them, just noting them for preprocessing prior to queuing with NFS@Home.

C208_148_104 7k curves @11e7
C178_140_114 full t55
C194_137_54 6k curves @11e7

All have (or will shortly have) completed t50.

Fivemack's new ECM toy may change the above values, or perhaps the entire table.

May I also request the following be added to the table, each for 8000 curves @B1=11e7

C212_138_53
C226_137_55
C216_131_67

Thanks.

I've got some business travel coming up, so it is my intent to concentrate on performing ECM for the foreseeable future. ECM on things like this table...

Can we add the following to the table, with a target of 9000 curves each at B1=11e7?

C211_121_95
C229_123_88
C227_124_85
C201_125_82
C234_138_82

These are the last of my nominees for this thread until a lot of the backlog is chewed up. Thanks.

Reserving C254_125_111 (initially running 4992@3e8)

[QUOTE=swellman;438067]Can we add the following to the table? Please note I am not reserving any of them, just noting them for preprocessing prior to queuing with NFS@Home.

C208_148_104 7k curves @11e7
C178_140_114 full t55
C194_137_54 6k curves @11e7

All have (or will shortly have) completed t50.

Fivemack's new ECM toy may change the above values, or perhaps the entire table.[/QUOTE]

There will be quite a lot of changes. For the C208 which is a S238 according to yafu it recommends [[1, 300], [3, 800], [11, 200], [43, 2700]] which is a lot less.

[QUOTE=henryzz;438416]There will be quite a lot of changes. For the C208 which is a S238 according to yafu it recommends [[1, 300], [3, 800], [11, 200], [43, 2700]] which is a lot less.[/QUOTE]

I've always felt the traditional rules of thumb required a LOT of ECM. If Baysian analysis tells us to reduce it (with props to RDS et al), I'm all for it.

Reserving C178_140_114 for ECM to t55. Thank you.

I'll probably enqueue all t55 tasks to yoyo in September.

[QUOTE=XYYXF;439256]I'll probably enqueue all t55 tasks to yoyo in September.[/QUOTE]

Nice. Though a full t55 is painfully slow on my hardware I will finish C178_140_114. ETA is 23 more days barring a hit. It's the the last C17X to be wrangled.

Once all the t50 jobs are completed for this project, what is the next step? There are probably several possible paths, none being better than the others. Just curious.

[QUOTE=swellman;439385]Once all the t50 jobs are completed for this project, what is the next step?[/QUOTE][QUOTE=XYYXF;439256]I'll probably enqueue all t55 tasks to yoyo in September.[/QUOTE]:)

Ah, so you mean ALL t55 tasks! I wrongly assumed you were referring to the table in the OP.

Factor on!:bow wave:

Request C239_150_41 and C224_136_106 for 8k curves @11e7 be added to the table. Thanks.

Added.

All B1=110M tasks (except C178_140_114 which is reserved) will be done by yoyo very soon :)

C178_140_114
 

Completed ECM to a full t55 with no factors found. I'll post it in the NFS@Home thread as a suggested 15e SNFS candidate.

C187_142_67:[code]7128360149039274062977936286065067193399002753241614879146717358452822451341395399593821536547721254531403479426890031997079861053249682361463634809511739986354847541601144102179145030773[/code]
and C202_134_87:[code]1224207612787772158727935698179132543695520938138955754343136473295556393880530309878400848872624432573326818957063965590308948460703802360355065604787211104932741174202689365592600945143938887378429707[/code]
both survived t55 and need 5000 curves @300M.

I'll take those two, it will take about five days and I can start in about a week.

5040@3e8 completed on C187_142_67; no factor found. About 5000 thread-hours.

5040@3e8 completed on C202_134_87; no factor found. 5361 thread-hours.

I will do test sieving on both of these and queue up appropriately.

I think C187_142_67 is a GNFS number: the SNFS difficulty is something like 268, the Murphy E score for the SNFS polynomial is 2.29e-14 and I've already found a 1.83e-14 GNFS polynomial after half a CPU-hour.

I am tempted at least to throw 1500 CPU-hours at the polynomial search and see what comes out; is that OK? The 15e queue is long enough that a few days delay is immaterial.

[QUOTE=axn;451261]What is the SNFS poly that you used? The obvious deg-5 poly gives me 1.967e-014 (msieve 1.52 SVN 883).[/QUOTE]

SNFS difficulty 260 seems a bit high for a degree-5 polynomial. I used degree 6 and multiplied through by 67^2*142, giving

[code]
$ cat C187_142_67.snfs
n: 7128360149039274062977936286065067193399002753241614879146717358452822451341395399593821536547721254531403479426890031997079861053249682361463634809511739986354847541601144102179145030773
skew: 0.1078
c6: 637438
c0: 1
Y0: 473338454265244805675008
Y1: -66956888672235945457062019127709902451882721
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
alambda: 2.6
rlambda: 2.6
alim: 240000000
rlim: 240000000
[/code]

which has yield 1.10rel/Q and 0.63s/rel at Q=240M.

OK, my prejudices were wrong;

[code]
n: 7128360149039274062977936286065067193399002753241614879146717358452822451341395399593821536547721254531403479426890031997079861053249682361463634809511739986354847541601144102179145030773
skew: 0.7405
c5: 20164
c0: 4489
Y0: -9544396591804396261630861312
Y1: 1349256365417755877454657051946796697206085388680241
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
alambda: 2.6
rlambda: 2.6
alim: 240000000
rlim: 240000000
[/code]

sieves at Q=240M with yield ~1.48 and time ~0.52s/rel - much better than the sextic, and competitive with the best GNFS polynomial I have.

(though, after more extensive trial sieving, the yield drops off to ~1.35, time ~0.54, and the GNFS polynomial is 1.52, 0.51)

I have moved the GNFS discussion to the GNFS thread, and am trial-sieving a bunch of polynomials tonight ready for queuing the best tomorrow.

I will run 10240 curves at 3e8 on C256_127_121 starting tomorrow, it should take about six days

( started 26/1/17 0854 )

I will run at least 20480 curves at B1=3e8 on C278_135_128

(first 8000 started 3/2/17 1545)

10240 curves at B1=3e8 complete on C256_127_121, no factor found.

I am on vacation for the next two weeks and probably without Internet access; could someone else do the polynomial setup and trial sieving and push the job onto the appropriate NFS@home queue?

I'll work on generating some polynomials and trial-sieving in YAFU. I'll put what I come up with in this thread so others can evaluate it before it gets submitted to NFS@Home. :smile:

YAFU gives the following SNFS polynomial as the best from trial-sieving:

[CODE]n: 2279019635262178260223610372866161113233537351181247262995589386985378698045180235228538838348697209388279801210996958280092611824762735859504092308376304672536049578066139601245777058442691694530960357274040042752387753146406894539941430545737889457466479
# 127^121+121^127, difficulty: 266.60, anorm: 2.48e+38, rnorm: 5.33e+49
# scaled difficulty: 268.49, suggest sieving rational side
# size = 2.087e-13, alpha = 1.491, combined = 3.768e-14, rroots = 0
type: snfs
size: 266
skew: 1.0081
c6: 121
c0: 127
Y1: -1191446152405248657777607437681912764659201
Y0: 54763699237492901685126120802225273763666521
m: 1662494550986228882566228817664053580038096714821209997934679199277061987019250219666195095292608720990844899512073487570488175183098414434723678817318610879312388031432395464414980456036350664791631044752565052965047488994184947124894919850559221309624285
rlim: 79600000
alim: 79600000
lpbr: 32
lpba: 32
mfbr: 64
mfba: 64
rlambda: 2.6
alambda: 2.6[/CODE]

snfspoly gives the same polynomial as Yafu for C256_127_121.

Seems like we've got a poly then. I'll post it in the NFS@Home thread.

Back from vacation.

My work machine got rebooted while I was away, so only 3000 curves were run from that batch (no factor).

10240 more curves started on GPU for C278_135_128, I'll do the rest of the job on GPU

[b]26/2[/b] 10240 curves started on stage 2, 7680 more started on GPU.

20920@3e8 completed on C278_135_128, no factor found.

Starting 25600@3e8 on C277_148_81

C277_148_81 done
 

[code]
Input number is 8251773826007569935997508400258020387415704842496291384921804552005221900662216913425918766759959411986740389289775723397876489045257317273581231513418658422520832086671482588420883344056709719929170923380392719222127153334982010212372015900798480010407898902278648943275730607 (277 digits)
Using B1=300000000-300000000, B2=3299108824840, polynomial Dickson(30), sigma=3:3380894970
Step 1 took 0ms
Step 2 took 1030674ms
********** Factor found in step 2: 9686520884177541535661024084340500289983192555244349
Found prime factor of 52 digits: 9686520884177541535661024084340500289983192555244349
Prime cofactor 851882107587920364660641481266438338230403696173481878562521588790426709670625553787663550972720378015286810501557748588179170143272281107487622422170172133317187820399465875718881126383728000476177931894438551879974742116443 has 225 digits
[/code]

Also found with sigma=3:4245679786

25600@3e8 stage-1 curves completed on C228_150_142.

Owing to a scheduling problem, I won't start stage-2 until tomorrow, and am not quite sure how long it will take.

Starting 30720@3e8 stage-1 curves on C237_139_126

Dear XYYXF, can we have some more candidates please?

Stage 2 now started on C228_150_142; on 12 threads on an i7-5820K it should take about eleven days and be finished in the evening of June 1st.

Stage 2 completed on 25600 curves on C228_150_142; no factor found.

First 12800 of 30720 curves on C237_139_126 completed both stages; no factor found.

Proceeding.

30720 curves on C237_139_126 completed, no factors found

[QUOTE=fivemack;460202]Stage 2 completed on 25600 curves on C228_150_142; no factor found.
[/QUOTE]

I started test sieving C228_150_142, just out of idle curiosity. It seems to be a relatively fast sieving 15/33 job, though I'm still gathering data.

How much more ECM is warranted before starting SNFS (assuming that is practical on 15e)?

[QUOTE=swellman;462121]I started test sieving C228_150_142, just out of idle curiosity. It seems to be a relatively fast sieving 15/33 job, though I'm still gathering data.

How much more ECM is warranted before starting SNFS (assuming that is practical on 15e)?[/QUOTE]

I think we have done enough ECM and can proceed to SNFS at once; I'll be interested to see your timing information once it's available.

[QUOTE=fivemack;462137]I think we have done enough ECM and can proceed to SNFS at once; I'll be interested to see your timing information once it's available.[/QUOTE]

Should have results by Friday as I am traveling right now. Yields on 15e were 1.8+ for Q <200M but actual timing data on a modern (though non-bleeding edge) processor using the full range of Q won't be available until I'm home.

[QUOTE=swellman;462121]I started test sieving C228_150_142, just out of idle curiosity. It seems to be a relatively fast sieving 15/33 job, though I'm still gathering data.

How much more ECM is warranted before starting SNFS (assuming that is practical on 15e)?[/QUOTE]

Using SNFS-281 (not sure if that's right or not) for difficulty and B1=260e6 with 25600 curves, Bayesian tool says no more ECM.

C228_150_142 test sieving results
 

C228_150_142 using the 15e siever, running one case at a time on a lightly loaded i7-3740QM (2.7 GHz) with 16Gb RAM in Win64 environment. Appears to be feasible for 15/33.

[code]
n: 338313268707086634350944196310206935873788335179366657051362293687646025379031678776100858721017586250223875732855601898290781411089750754177921272802024755914362548300092366882262859585028219144728898823695818541135039489183653
# 150^142+142^150, difficulty: 280.89, anorm: 6.00e+037, rnorm: -3.92e+052
# scaled difficulty: 283.36, suggest sieving rational side
# size = 6.177e-014, alpha = 0.000, combined = 1.426e-014, rroots = 0
type: snfs
size: 280
skew: 1.0627
c6: 25
c0: 36
Y1: -38237431646084858983458149164083507643015743502
Y0: 200678255510666758709703572094440460205078125
rlim: 500000000
alim: 500000000
lpbr: 33
lpba: 33
mfbr: 66
mfba: 66
rlambda: 3.0
alambda: 3.0
[/code]


Test sieving with Q in blocks of 5000
[code]
Q= 20M 9464 rels 0.25046 sec/rel
Q=100M 9185 rels 0.27238 sec/rel
Q=200M 8307 rels 0.33120 sec/rel
Q=300M 7999 rels 0.39008 sec/rel
Q=400M 7337 rels 0.39881 sec/rel
Q=500M 6589 rels 0.49123 sec/rel
Q=600M 5966 rels 0.53501 sec/rel
Q=700M 6007 rels 0.56676 sec/rel
[/code]

Suggesting a range of 20M-650M+ for Q.

C278_135_128 For 15e?
 

C278_135_128 seems suitable for 15e, though I'm not sure where it stands wrt ECM.

[code]
n: 33927504534476723516408816689812909770719932521618781773756995289471261780068216894657109048688644848466558715329703788016114249127096969960226894868006429336428695102577256482033601100297718050792231930859693133081358214479422873421598023360475055929769632226103912177557671593
# 135^128+128^135, difficulty: 286.77, anorm: 1.81e+039, rnorm: -8.22e+052
# scaled difficulty: 289.05, suggest sieving rational side
# size = 2.717e-014, alpha = 0.000, combined = 7.892e-015, rroots = 0
type: snfs
size: 286
skew: 2.4183
c6: 1
c0: 200
Y1: -13098464280290651262885471684574127197265625000
Y0: 2923003274661805836407369665432566039311865085952
rlim: 500000000
alim: 500000000
lpbr: 33
lpba: 33
mfbr: 66
mfba: 66
rlambda: 3.0
alambda: 3.0
[/code]


Test sieving with Q in blocks of 2000
[code]
Q= 20M 3218 0.24388 sec/rel
Q=100M 3057 0.26519 sec/rel
Q=200M 2838 0.31295 sec/rel
Q=300M 2623 0.37733 sec/rel
Q=400M 2195 0.43902 sec/rel
Q=500M 2564 0.55437 sec/rel
Q=600M 2088 0.56867 sec/rel
Q=700M 2051 0.62015 sec/rel
Q=800M 1924 0.68246 sec/rel
[/code]

FWIW.

[QUOTE=swellman;462654]C278_135_128 seems suitable for 15e, though I'm not sure where it stands wrt ECM.[/quote]

I'd done 20920 curves at B1=3e8 which was enough according to the Bayesian analysis. Job queued up.

[QUOTE=fivemack;462137]I think we have done enough ECM and can proceed to SNFS at once; I'll be interested to see your timing information once it's available.[/QUOTE]

Do you still want to queue [url=http://www.mersenneforum.org/showpost.php?p=462424&postcount=55]C228_150_142[/url] as well? It appears ready for 15e, though back to back 33-bit jobs may smother the fire.

Thank you for enqueing C278_135_128. Seems to be near the limits of 15e.