Msieve 1.40 Beta Feedback - Page 3

Jeff Gilchrist · 2009-03-03, 14:30

I just finished poly selection with a C132, and happened to just finish the same one with 1.39 the day before.

1.39:

Code:

polynomial selection complete
R0: -19601538359178012858243772
R1:  362245613431421
A0: -3596818452703299422588803854324375
A1:  25369161362208464044358530617
A2:  29821113782020323028439
A3: -132269498744319265
A4: -48248765752
A5:  47280
skew 544564.83, size 8.312120e-14, alpha -7.928360, combined = 1.168045e-12
elapsed time 19:31:12

1.40beta1:

Code:

polynomial selection complete
R0: -20308800184976938479023733
R1:  405438325069519
A0:  56288955419506072631918641873120
A1:  1857942727866668410885482904
A2: -2888189578940218801522
A3: -26365842636068533
A4:  16177108056
A5:  39600
skew 436881.66, size 1.151133e-12, alpha -6.381464, combined = 6.518216e-11
elapsed time 19:28:31

Now if I re-run poly selection with 1.40beta1 using the old msieve.dat.p file, it will finish right away and report the same polynomial but give the new score so to compare:
skew 544564.83, size 1.168045e-12, alpha -7.928360, combined = 6.420400e-11

So it looks like 1.40 generated a better one according to the Murphy E value anyways. The alpha is quite a bit different so I'm not sure what difference there would in real sieving.

jasonp · 2009-03-03, 16:21

The two should sieve about he same; note that the combined score in v1.39 is actually the size score in v1.40. v1.39 did not actually optimize polynomials for high Murphy score, so it isn't surprising that v1.40 finds a (slightly) better Murphy score.

Jeff Gilchrist · 2009-03-03, 19:06

So I just finished 5 poly selections with 1.40beta1 all C132 or C133s. I did notice some errors in the output, not sure if this happened in 1.39 or not:

Quote:

p 12254665 15931064 26923503 35000553 lattice 3333581
batch 5000 13593649
integrator failed 2.829177e-16 6.510742e-20
error: size score computation failed
integrator failed 1.070448e-16 6.296326e-20
error: size score computation failed
integrator failed 4.835671e-16 6.072324e-20
error: size score computation failed

If you want the output or any other files, let me know. They all had those error outputs to varying degrees.

The 5 polys that were generated are:

Code:

polynomial selection complete
R0: -28049983285732204469831666
R1:  415918189585193
A0:  10681504745584543706654880359695
A1:  320692898755356107511046719
A2: -896465223768151866998
A3: -25246432839495259
A4:  18894920523
A5:  41160
skew 265051.76, size 1.158302e-12, alpha -6.308461, combined = 6.611170e-11

polynomial selection complete
R0: -25043072078991536197773317
R1:  500810759980091
A0:  116567090883408195054754470259200
A1:  35581374347897846358206836992
A2:  81764510047522465945392
A3: -123415064274124832
A4: -89599631627
A5:  49980
skew 858939.42, size 1.121560e-12, alpha -8.249332, combined = 6.359523e-11

polynomial selection complete
R0: -20308800184976938479023733
R1:  405438325069519
A0:  56288955419506072631918641873120
A1:  1857942727866668410885482904
A2: -2888189578940218801522
A3: -26365842636068533
A4:  16177108056
A5:  39600
skew 436881.66, size 1.151133e-12, alpha -6.381464, combined = 6.518216e-11

polynomial selection complete
R0: -38905099305711493030228534
R1:  528922922629967
A0:  914938046534427904296708222045531
A1:  8843544226278046696751092452
A2: -30075340366464522371749
A3: -84098666213673742
A4:  96276901808
A5:  58800
skew 583076.85, size 7.491921e-13, alpha -7.032207, combined = 5.058169e-11

polynomial selection complete
R0: -33741066621966721400837752
R1:  469234914160613
A0: -45201965182259963415813457444917
A1:  3035342643500588236781756415
A2: -6466045427748810203567
A3: -53271320134610357
A4:  20397515266
A5:  43440
skew 471899.30, size 9.406818e-13, alpha -6.800495, combined = 5.823381e-11

10metreh · 2009-03-03, 20:00

Could this be the first completed GNFS with v1.40? Here is the log:

Code:

Tue Mar 03 19:22:20 2009  
Tue Mar 03 19:22:22 2009  
Tue Mar 03 19:22:22 2009  Msieve v. 1.40
Tue Mar 03 19:22:22 2009  random seeds: 0010f447 0021e88e
Tue Mar 03 19:22:22 2009  factoring 3046940677438483872492570557601426992994034892890836252574868172241908349534364214600428141457053 (97 digits)
Tue Mar 03 19:22:24 2009  searching for 15-digit factors
Tue Mar 03 19:22:27 2009  commencing number field sieve (97-digit input)
Tue Mar 03 19:22:27 2009  R0: -6102425988474888054
Tue Mar 03 19:22:27 2009  R1:  9564234367
Tue Mar 03 19:22:27 2009  A0:  407934978221462444764895
Tue Mar 03 19:22:27 2009  A1: -154209112521686219328
Tue Mar 03 19:22:27 2009  A2: -7261393706314151
Tue Mar 03 19:22:27 2009  A3:  1675925509656
Tue Mar 03 19:22:27 2009  A4:  25858416
Tue Mar 03 19:22:27 2009  A5:  360
Tue Mar 03 19:22:27 2009  skew 10703.18, size 3.876917e-009, alpha -6.133547, combined = 5.255935e-009
Tue Mar 03 19:22:27 2009  
Tue Mar 03 19:22:27 2009  commencing relation filtering
Tue Mar 03 19:22:27 2009  commencing duplicate removal, pass 1
Tue Mar 03 19:22:55 2009  error -15 reading relation 301787
Tue Mar 03 19:23:02 2009  error -11 reading relation 601316
Tue Mar 03 19:23:13 2009  error -15 reading relation 1034599
Tue Mar 03 19:23:15 2009  error -4 reading relation 1066768
Tue Mar 03 19:23:18 2009  error -15 reading relation 1216882
Tue Mar 03 19:23:53 2009  found 163099 hash collisions in 1950145 relations
Tue Mar 03 19:24:13 2009  added 214 free relations
Tue Mar 03 19:24:16 2009  commencing duplicate removal, pass 2
Tue Mar 03 19:25:13 2009  found 162468 duplicates and 1787891 unique relations
Tue Mar 03 19:25:18 2009  memory use: 36.2 MB
Tue Mar 03 19:25:21 2009  reading rational ideals above 983040
Tue Mar 03 19:25:21 2009  reading algebraic ideals above 983040
Tue Mar 03 19:25:21 2009  commencing singleton removal, pass 1
Tue Mar 03 19:28:57 2009  relations with 0 large ideals: 33799
Tue Mar 03 19:29:03 2009  relations with 1 large ideals: 266492
Tue Mar 03 19:29:03 2009  relations with 2 large ideals: 664421
Tue Mar 03 19:29:03 2009  relations with 3 large ideals: 598607
Tue Mar 03 19:29:03 2009  relations with 4 large ideals: 191596
Tue Mar 03 19:29:03 2009  relations with 5 large ideals: 18349
Tue Mar 03 19:29:03 2009  relations with 6 large ideals: 14623
Tue Mar 03 19:29:03 2009  relations with 7+ large ideals: 4
Tue Mar 03 19:29:03 2009  1787891 relations and about 1807219 large ideals
Tue Mar 03 19:29:04 2009  commencing singleton removal, pass 2
Tue Mar 03 19:32:22 2009  found 904409 singletons
Tue Mar 03 19:32:26 2009  current dataset: 883482 relations and about 705265 large ideals
Tue Mar 03 19:32:29 2009  commencing singleton removal, pass 3
Tue Mar 03 19:34:32 2009  found 165920 singletons
Tue Mar 03 19:34:54 2009  current dataset: 717562 relations and about 527758 large ideals
Tue Mar 03 19:35:06 2009  commencing singleton removal, final pass
Tue Mar 03 19:35:56 2009  memory use: 11.4 MB
Tue Mar 03 19:35:57 2009  commencing in-memory singleton removal
Tue Mar 03 19:35:57 2009  begin with 717562 relations and 534400 unique ideals
Tue Mar 03 19:35:59 2009  reduce to 636505 relations and 451610 ideals in 11 passes
Tue Mar 03 19:35:59 2009  max relations containing the same ideal: 49
Tue Mar 03 19:36:01 2009  reading rational ideals above 491520
Tue Mar 03 19:36:01 2009  reading algebraic ideals above 491520
Tue Mar 03 19:36:01 2009  commencing singleton removal, final pass
Tue Mar 03 19:36:31 2009  keeping 500441 ideals with weight <= 20, new excess is 106994
Tue Mar 03 19:36:35 2009  memory use: 16.9 MB
Tue Mar 03 19:36:36 2009  commencing in-memory singleton removal
Tue Mar 03 19:36:36 2009  begin with 636719 relations and 500441 unique ideals
Tue Mar 03 19:36:38 2009  reduce to 633899 relations and 496548 ideals in 8 passes
Tue Mar 03 19:36:38 2009  max relations containing the same ideal: 20
Tue Mar 03 19:36:40 2009  relations with 0 large ideals: 15524
Tue Mar 03 19:36:40 2009  relations with 1 large ideals: 87834
Tue Mar 03 19:36:40 2009  relations with 2 large ideals: 194501
Tue Mar 03 19:36:40 2009  relations with 3 large ideals: 204156
Tue Mar 03 19:36:40 2009  relations with 4 large ideals: 103282
Tue Mar 03 19:36:40 2009  relations with 5 large ideals: 24457
Tue Mar 03 19:36:40 2009  relations with 6 large ideals: 3948
Tue Mar 03 19:36:40 2009  relations with 7+ large ideals: 197
Tue Mar 03 19:36:40 2009  commencing 2-way merge
Tue Mar 03 19:36:41 2009  reduce to 379283 relation sets and 241936 unique ideals
Tue Mar 03 19:36:41 2009  ignored 4 oversize relation sets
Tue Mar 03 19:36:41 2009  commencing full merge
Tue Mar 03 19:36:52 2009  memory use: 14.7 MB
Tue Mar 03 19:36:52 2009  found 159403 cycles, need 138136
Tue Mar 03 19:36:52 2009  weight of 138136 cycles is about 9739801 (70.51/cycle)
Tue Mar 03 19:36:52 2009  distribution of cycle lengths:
Tue Mar 03 19:36:52 2009  1 relations: 18829
Tue Mar 03 19:36:52 2009  2 relations: 13295
Tue Mar 03 19:36:52 2009  3 relations: 12413
Tue Mar 03 19:36:52 2009  4 relations: 11566
Tue Mar 03 19:36:52 2009  5 relations: 11072
Tue Mar 03 19:36:52 2009  6 relations: 10112
Tue Mar 03 19:36:52 2009  7 relations: 9288
Tue Mar 03 19:36:52 2009  8 relations: 8279
Tue Mar 03 19:36:52 2009  9 relations: 7578
Tue Mar 03 19:36:52 2009  10+ relations: 35704
Tue Mar 03 19:36:52 2009  heaviest cycle: 19 relations
Tue Mar 03 19:36:52 2009  commencing cycle optimization
Tue Mar 03 19:36:53 2009  start with 898319 relations
Tue Mar 03 19:36:59 2009  pruned 42732 relations
Tue Mar 03 19:36:59 2009  memory use: 22.2 MB
Tue Mar 03 19:36:59 2009  distribution of cycle lengths:
Tue Mar 03 19:36:59 2009  1 relations: 18829
Tue Mar 03 19:36:59 2009  2 relations: 13776
Tue Mar 03 19:36:59 2009  3 relations: 13127
Tue Mar 03 19:36:59 2009  4 relations: 12228
Tue Mar 03 19:36:59 2009  5 relations: 11740
Tue Mar 03 19:36:59 2009  6 relations: 10623
Tue Mar 03 19:36:59 2009  7 relations: 9891
Tue Mar 03 19:36:59 2009  8 relations: 8609
Tue Mar 03 19:36:59 2009  9 relations: 7800
Tue Mar 03 19:36:59 2009  10+ relations: 31513
Tue Mar 03 19:36:59 2009  heaviest cycle: 19 relations
Tue Mar 03 19:37:01 2009  RelProcTime: 233
Tue Mar 03 19:37:01 2009  elapsed time 00:03:58
Tue Mar 03 19:37:18 2009  
Tue Mar 03 19:37:18 2009  
Tue Mar 03 19:37:18 2009  Msieve v. 1.40
Tue Mar 03 19:37:18 2009  random seeds: 0010f447 0021e88e
Tue Mar 03 19:37:18 2009  factoring 3046940677438483872492570557601426992994034892890836252574868172241908349534364214600428141457053 (97 digits)
Tue Mar 03 19:37:20 2009  searching for 15-digit factors
Tue Mar 03 19:37:24 2009  commencing number field sieve (97-digit input)
Tue Mar 03 19:37:24 2009  R0: -6102425988474888054
Tue Mar 03 19:37:24 2009  R1:  9564234367
Tue Mar 03 19:37:24 2009  A0:  407934978221462444764895
Tue Mar 03 19:37:24 2009  A1: -154209112521686219328
Tue Mar 03 19:37:24 2009  A2: -7261393706314151
Tue Mar 03 19:37:24 2009  A3:  1675925509656
Tue Mar 03 19:37:24 2009  A4:  25858416
Tue Mar 03 19:37:24 2009  A5:  360
Tue Mar 03 19:37:24 2009  skew 10703.18, size 3.876917e-009, alpha -6.133547, combined = 5.255935e-009
Tue Mar 03 19:37:24 2009  
Tue Mar 03 19:37:24 2009  commencing linear algebra
Tue Mar 03 19:37:24 2009  read 138136 cycles
Tue Mar 03 19:37:25 2009  cycles contain 455006 unique relations
Tue Mar 03 19:38:38 2009  read 455006 relations
Tue Mar 03 19:38:43 2009  using 20 quadratic characters above 33553838
Tue Mar 03 19:38:51 2009  building initial matrix
Tue Mar 03 19:39:07 2009  memory use: 49.9 MB
Tue Mar 03 19:39:10 2009  read 138136 cycles
Tue Mar 03 19:39:16 2009  matrix is 137759 x 138136 (37.7 MB) with weight 12837439 (92.93/col)
Tue Mar 03 19:39:16 2009  sparse part has weight 8924927 (64.61/col)
Tue Mar 03 19:39:29 2009  filtering completed in 3 passes
Tue Mar 03 19:39:29 2009  matrix is 136529 x 136729 (37.4 MB) with weight 12701282 (92.89/col)
Tue Mar 03 19:39:29 2009  sparse part has weight 8834346 (64.61/col)
Tue Mar 03 19:39:38 2009  read 136729 cycles
Tue Mar 03 19:39:44 2009  matrix is 136529 x 136729 (37.4 MB) with weight 12701282 (92.89/col)
Tue Mar 03 19:39:44 2009  sparse part has weight 8834346 (64.61/col)
Tue Mar 03 19:39:44 2009  saving the first 48 matrix rows for later
Tue Mar 03 19:39:45 2009  matrix is 136481 x 136729 (35.9 MB) with weight 9993250 (73.09/col)
Tue Mar 03 19:39:45 2009  sparse part has weight 8603212 (62.92/col)
Tue Mar 03 19:39:45 2009  matrix includes 64 packed rows
Tue Mar 03 19:39:45 2009  using block size 10922 for processor cache size 256 kB
Tue Mar 03 19:39:48 2009  commencing Lanczos iteration
Tue Mar 03 19:39:48 2009  memory use: 34.5 MB
Tue Mar 03 19:47:39 2009  lanczos halted after 2160 iterations (dim = 136477)
Tue Mar 03 19:47:41 2009  recovered 29 nontrivial dependencies
Tue Mar 03 19:47:41 2009  BLanczosTime: 534
Tue Mar 03 19:47:41 2009  elapsed time 00:09:00
Tue Mar 03 19:47:43 2009  
Tue Mar 03 19:47:43 2009  
Tue Mar 03 19:47:43 2009  Msieve v. 1.40
Tue Mar 03 19:47:43 2009  random seeds: 0010f447 0021e88e
Tue Mar 03 19:47:43 2009  factoring 3046940677438483872492570557601426992994034892890836252574868172241908349534364214600428141457053 (97 digits)
Tue Mar 03 19:47:45 2009  searching for 15-digit factors
Tue Mar 03 19:47:48 2009  commencing number field sieve (97-digit input)
Tue Mar 03 19:47:49 2009  R0: -6102425988474888054
Tue Mar 03 19:47:49 2009  R1:  9564234367
Tue Mar 03 19:47:49 2009  A0:  407934978221462444764895
Tue Mar 03 19:47:49 2009  A1: -154209112521686219328
Tue Mar 03 19:47:49 2009  A2: -7261393706314151
Tue Mar 03 19:47:49 2009  A3:  1675925509656
Tue Mar 03 19:47:49 2009  A4:  25858416
Tue Mar 03 19:47:49 2009  A5:  360
Tue Mar 03 19:47:49 2009  skew 10703.18, size 3.876917e-009, alpha -6.133547, combined = 5.255935e-009
Tue Mar 03 19:47:49 2009  
Tue Mar 03 19:47:49 2009  commencing square root phase
Tue Mar 03 19:47:49 2009  reading relations for dependency 1
Tue Mar 03 19:47:50 2009  read 68182 cycles
Tue Mar 03 19:47:50 2009  cycles contain 279170 unique relations
Tue Mar 03 19:48:02 2009  read 279170 relations
Tue Mar 03 19:48:07 2009  multiplying 225510 relations
Tue Mar 03 19:49:20 2009  multiply complete, coefficients have about 8.20 million bits
Tue Mar 03 19:49:21 2009  initial square root is modulo 2679368959
Tue Mar 03 19:52:05 2009  sqrtTime: 240
Tue Mar 03 19:52:05 2009  prp35 factor: 79433306239198200389998055491829123
Tue Mar 03 19:52:05 2009  prp62 factor: 38358477340263354052027712003622188325699794811037315066562911
Tue Mar 03 19:52:05 2009  elapsed time 00:04:05

mdettweiler · 2009-03-03, 20:31

Quote:

Originally Posted by 10metreh

Could this be the first completed GNFS with v1.40? Here is the log:

Code:

log snipped

Sorry to bust your bubble, but I did two successful postprocessing jobs with v1.40beta1 last night, one on a C101 and the other on a C100. And I'm sure that Jason probably did a handful of them as part of his own internal alpha testing.

smh · 2009-03-03, 21:02

Quote:

Originally Posted by 10metreh

Could this be the first completed GNFS with v1.40?

Nope. I did at least half a dozen today, ranging from a c98 to a c106

CRGreathouse · 2009-03-03, 21:20

Quote:

Originally Posted by smh

Nope. I did at least half a dozen today, ranging from a c98 to a c106

But were any of them -1 mod 127? Huh?

Batalov · 2009-03-03, 22:31

The real test would have been to factor $2^{815}-3$
and have a factor $\equiv$ 85 (mod 324)...

frmky · 2009-03-04, 16:06

When doing linear algebra with multiple threads with the 64-bit Linux binary, the ETA gives the time required if there was only 1 thread running.

Oops, I see you released beta 2. I haven't checked to see if the bug is still there...

jasonp · 2009-03-04, 16:41

Quote:

Originally Posted by frmky

When doing linear algebra with multiple threads with the 64-bit Linux binary, the ETA gives the time required if there was only 1 thread running.

Oops, I see you released beta 2. I haven't checked to see if the bug is still there...

Beta 2 won't change that; it looks like getrusage() adds up all the time used in threads, like clock() does. A reading of the man page suggests that RUSAGE_SELF will only count the time used by the host process, but doesn't say anything about threads.

Jeff Gilchrist · 2009-03-04, 19:42

FYI, I just finished a C132 that was poly selected with 1.39, sieved with ggnfs, started post-processing with 1.39, and finished with 1.40beta1 and it all worked fine. This was done with the Windows 32bit binary.

2009-03-03, 16:21	#24
jasonp Tribal Bullet Oct 2004 3,541 Posts	The two should sieve about he same; note that the combined score in v1.39 is actually the size score in v1.40. v1.39 did not actually optimize polynomials for high Murphy score, so it isn't surprising that v1.40 finds a (slightly) better Murphy score. Last fiddled with by jasonp on 2009-03-03 at 16:23

2009-03-04, 16:06	#31
frmky Jul 2003 So Cal 2×3⁴×13 Posts	When doing linear algebra with multiple threads with the 64-bit Linux binary, the ETA gives the time required if there was only 1 thread running. Oops, I see you released beta 2. I haven't checked to see if the bug is still there... Last fiddled with by frmky on 2009-03-04 at 16:09

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Msieve 1.53 feedback	xilman	Msieve	149	2018-11-12 06:37
Msieve 1.51 feedback	xilman	Msieve	256	2014-01-26 22:06
Msieve v1.46 feedback	em99010pepe	Msieve	153	2010-12-12 14:21
Msieve 1.44 feedback	xilman	Msieve	111	2010-09-14 21:50
Msieve 1.43 feedback	Jeff Gilchrist	Msieve	47	2009-11-24 15:53

2009-03-03, 14:30	#23
Jeff Gilchrist Jun 2003 Ottawa, Canada 495₁₆ Posts	I just finished poly selection with a C132, and happened to just finish the same one with 1.39 the day before. 1.39: Code: polynomial selection complete R0: -19601538359178012858243772 R1: 362245613431421 A0: -3596818452703299422588803854324375 A1: 25369161362208464044358530617 A2: 29821113782020323028439 A3: -132269498744319265 A4: -48248765752 A5: 47280 skew 544564.83, size 8.312120e-14, alpha -7.928360, combined = 1.168045e-12 elapsed time 19:31:12 1.40beta1: Code: polynomial selection complete R0: -20308800184976938479023733 R1: 405438325069519 A0: 56288955419506072631918641873120 A1: 1857942727866668410885482904 A2: -2888189578940218801522 A3: -26365842636068533 A4: 16177108056 A5: 39600 skew 436881.66, size 1.151133e-12, alpha -6.381464, combined = 6.518216e-11 elapsed time 19:28:31 Now if I re-run poly selection with 1.40beta1 using the old msieve.dat.p file, it will finish right away and report the same polynomial but give the new score so to compare: skew 544564.83, size 1.168045e-12, alpha -7.928360, combined = 6.420400e-11 So it looks like 1.40 generated a better one according to the Murphy E value anyways. The alpha is quite a bit different so I'm not sure what difference there would in real sieving.

2009-03-03, 22:31	#30
Batalov "Serge" Mar 2008 Phi(4,2^7658614+1)/2 3⁶·13 Posts	The real test would have been to factor $2^{815}-3$ and have a factor $\equiv$ 85 (mod 324)...

2009-03-04, 19:42	#33
Jeff Gilchrist Jun 2003 Ottawa, Canada 3×17×23 Posts	FYI, I just finished a C132 that was poly selected with 1.39, sieved with ggnfs, started post-processing with 1.39, and finished with 1.40beta1 and it all worked fine. This was done with the Windows 32bit binary.