mersenneforum.org Reserved for MF - Sequence 276
 Register FAQ Search Today's Posts Mark Forums Read

 2022-08-07, 19:15 #155 ryanp     Jun 2012 Boulder, CO 2×211 Posts I'm sieving more now. I'm wondering, though, the best way to determine "I am at a good stopping point; a valid matrix can be formed". For example, the last bit of my log from filtering with target_density=130 and ~1.7B uniques was: Code: Thu Aug 4 06:02:08 2022 commencing in-memory singleton removal Thu Aug 4 06:02:31 2022 begin with 382020342 relations and 381478357 unique ideals Thu Aug 4 06:05:00 2022 reduce to 381901947 relations and 375910448 ideals in 6 passes Thu Aug 4 06:05:00 2022 max relations containing the same ideal: 120 Thu Aug 4 06:07:21 2022 removing 608320 relations and 495379 ideals in 112941 cliques Thu Aug 4 06:07:29 2022 commencing in-memory singleton removal Thu Aug 4 06:07:52 2022 begin with 381293627 relations and 375910448 unique ideals Thu Aug 4 06:09:55 2022 reduce to 381292945 relations and 375414387 ideals in 5 passes Thu Aug 4 06:09:55 2022 max relations containing the same ideal: 119 Thu Aug 4 06:11:02 2022 relations with 0 large ideals: 1988 Thu Aug 4 06:11:02 2022 relations with 1 large ideals: 69122 Thu Aug 4 06:11:02 2022 relations with 2 large ideals: 1154553 Thu Aug 4 06:11:02 2022 relations with 3 large ideals: 8235700 Thu Aug 4 06:11:02 2022 relations with 4 large ideals: 32013767 Thu Aug 4 06:11:02 2022 relations with 5 large ideals: 73520066 Thu Aug 4 06:11:02 2022 relations with 6 large ideals: 103511912 Thu Aug 4 06:11:02 2022 relations with 7+ large ideals: 162785837 Thu Aug 4 06:11:02 2022 commencing 2-way merge Thu Aug 4 06:14:39 2022 reduce to 260519145 relation sets and 254639763 unique ideals Thu Aug 4 06:14:39 2022 commencing full merge Thu Aug 4 07:30:29 2022 memory use: 28478.0 MB Thu Aug 4 07:30:56 2022 found 113365496 cycles, need 112775963 Thu Aug 4 07:31:42 2022 weight of 112775963 cycles is about 14661120944 (130.00/cycle) Thu Aug 4 07:31:43 2022 distribution of cycle lengths: Thu Aug 4 07:31:43 2022 1 relations: 2999784 Thu Aug 4 07:31:43 2022 2 relations: 6074372 Thu Aug 4 07:31:43 2022 3 relations: 7951682 Thu Aug 4 07:31:43 2022 4 relations: 8612278 Thu Aug 4 07:31:43 2022 5 relations: 9269128 Thu Aug 4 07:31:43 2022 6 relations: 9154906 Thu Aug 4 07:31:43 2022 7 relations: 9041748 Thu Aug 4 07:31:43 2022 8 relations: 8585973 Thu Aug 4 07:31:43 2022 9 relations: 8046841 Thu Aug 4 07:31:43 2022 10+ relations: 43039251 Thu Aug 4 07:31:43 2022 heaviest cycle: 27 relations Thu Aug 4 07:32:05 2022 commencing cycle optimization Thu Aug 4 07:36:05 2022 start with 969975472 relations Thu Aug 4 08:08:48 2022 pruned 43823274 relations Thu Aug 4 08:08:50 2022 memory use: 25937.9 MB Thu Aug 4 08:08:51 2022 distribution of cycle lengths: Thu Aug 4 08:08:51 2022 1 relations: 2999784 Thu Aug 4 08:08:51 2022 2 relations: 6256591 Thu Aug 4 08:08:51 2022 3 relations: 8324812 Thu Aug 4 08:08:51 2022 4 relations: 9020005 Thu Aug 4 08:08:51 2022 5 relations: 9762136 Thu Aug 4 08:08:51 2022 6 relations: 9623615 Thu Aug 4 08:08:51 2022 7 relations: 9515903 Thu Aug 4 08:08:51 2022 8 relations: 8987594 Thu Aug 4 08:08:51 2022 9 relations: 8382709 Thu Aug 4 08:08:51 2022 10+ relations: 39902814 Thu Aug 4 08:08:51 2022 heaviest cycle: 27 relations Thu Aug 4 08:13:14 2022 RelProcTime: 126525 Thu Aug 4 08:14:10 2022 elapsed time 35:09:42 So it certainly seemed like a matrix could be formed at this point. Should I be targeting a certain number of unique rels?
2022-08-08, 08:22   #156
bur

Aug 2020
79*6581e-4;3*2539e-3

23×3×52 Posts

Quote:
 Originally Posted by frmky After filtering, the matrix has more rows than columns. The system is overdetermined so there's no solution other than the trivial one. So we need more sieving for msieve.
If this situation requires more sieving, I'd assume we have more distinct prime factors (or ideals actually?) than relations? But then why is it called overdetermined? That to me sounds like having more equations than unknowns, i.e. more relations than factors, but that's apparently not the case.

Last fiddled with by bur on 2022-08-08 at 08:23

2022-08-08, 11:13   #157
charybdis

Apr 2020

32·5·19 Posts

Quote:
 Originally Posted by bur If this situation requires more sieving, I'd assume we have more distinct prime factors (or ideals actually?) than relations? But then why is it called overdetermined? That to me sounds like having more equations than unknowns, i.e. more relations than factors, but that's apparently not the case.
You're right about more equations than unknowns, but the equations correspond to ideals, not relations. The equations encode "this ideal must appear an even number of times in the desired set of relations". In other words, if we find a set of relations that satisfies all the equations, then every ideal appears an even number of times in it, so multiplying the relations produces squares on both the rational and algebraic sides, as desired.

 2022-08-10, 05:27 #158 frmky     Jul 2003 So Cal 3·823 Posts A little more sieving was all it needed. LA is running now on four A100 GPUS. Code: linear algebra completed 514317 of 109926366 dimensions (0.5%, ETA 70h 7m)
2022-08-10, 12:47   #159
ryanp

Jun 2012
Boulder, CO

1A616 Posts

Quote:
 Originally Posted by frmky A little more sieving was all it needed. LA is running now on four A100 GPUS. Code: linear algebra completed 514317 of 109926366 dimensions (0.5%, ETA 70h 7m)
Great news! Thanks again for your help on this.

 2022-08-13, 04:13 #160 frmky     Jul 2003 So Cal 3·823 Posts I was able to move it to an 8x A100 system for most of the LA. It's done and posted to factordb.
2022-08-13, 14:24   #161
sweety439

"99(4^34019)99 palind"
Nov 2016
(P^81993)SZ base 36

357710 Posts

Quote:
 Originally Posted by frmky I was able to move it to an 8x A100 system for most of the LA. It's done and posted to factordb.
http://factordb.com/sequences.php?se...20&fr=0&to=100

Now https://oeis.org/A008892 can be extended to n=2145 and https://oeis.org/A347769 can be extended to n=2708, can someone update them?

2022-08-13, 14:31   #162
mathwiz

Mar 2019

1001011102 Posts

Quote:
 Originally Posted by sweety439 http://factordb.com/sequences.php?se...20&fr=0&to=100 Now https://oeis.org/A008892 can be extended to n=2145 and https://oeis.org/A347769 can be extended to n=2708, can someone update them?
Why the hell do you keep asking others to do this work? Submit the changes to OEIS yourself!

2022-08-13, 14:34   #163
charybdis

Apr 2020

32×5×19 Posts

Quote:
 Originally Posted by mathwiz Why the hell do you keep asking others to do this work? Submit the changes to OEIS yourself!
Sweety is unsurprisingly banned from OEIS. You reap what you sow.

 2022-08-15, 11:34 #164 swellman     Jun 2012 25×5×23 Posts Currently at i2145 a C209. Is anyone working this series, say ECM?
2022-08-15, 15:46   #165
ryanp

Jun 2012
Boulder, CO

2×211 Posts

Quote:
 Originally Posted by swellman Currently at i2145 a C209. Is anyone working this series, say ECM?
I've run 27K curves at B1=26e7 with GMP-ECM, no luck so far.

 Similar Threads Thread Thread Starter Forum Replies Last Post RichD Aliquot Sequences 549 2022-08-24 20:00 schickel Aliquot Sequences 3149 2021-12-01 00:06 devarajkandadai Miscellaneous Math 3 2020-12-01 22:08 sweety439 sweety439 17 2017-06-13 03:49 petrw1 Lone Mersenne Hunters 82 2010-01-11 01:57

All times are UTC. The time now is 00:06.

Mon Sep 26 00:06:22 UTC 2022 up 38 days, 21:34, 0 users, load averages: 2.02, 1.53, 1.33