Register FAQ Search Today's Posts Mark Forums Read

2015-03-27, 15:27   #122
xilman
Bamboozled!

"𒉺𒌌𒇷𒆷𒀭"
May 2003
Down not across

97·113 Posts

Quote:
 Originally Posted by xilman It's time to wrap up the sieving phase now. Tom and I have arranged to transfer all the relations on my systems to his so that he can run the linear algebra and square root phases. That will happen on Friday evening, UK time. Accordingly, please upload all your data by 2015-03-27 12:00 UTC at the latest, to give me chance to unpack and check the relations that afternoon. It would be best from my POV if the transfers happen between 00:00 and 08:00 UTC if possible, please, so that they don't count against my "free" data limit.
I allowed a 2-hour grace period. Removing all the duplicates took 50 minutes.

The final state of play here is

Code:
Found 540343842 unique, 93906329 duplicate (14.8% of total), and 0 bad relations.
Largest dimension used: 1842 of 3000
Average dimension used: 1649.0 of 3000
Terminating program at Fri Mar 27 15:02:50 2015
To that duplicate total should be added another 18704982 from earlier runs. Tom will doubtless find millions more when he merges in his data set. The unique relations take up 63G of disk space. The uploaded files come to 40G, saving me a significant amount of data transfer cost.

Any remaining relations you may have will probably be ignored. Feel free to send them in over the next few hours but whether Tom uses them is entirely up to him. I'll disable the anon-ftp service some time this weekend.

Thanks all!

Paul

 2015-03-28, 08:33 #123 fivemack (loop (#_fork))     Feb 2006 Cambridge, England 192916 Posts The 63GB of relations data has been successfully transferred to my computer, using a 150Mbit-per-second transfer link comprising a Hitachi 4TB USB3 hard disc drive and a Claud Butler hybrid commuting bicycle, powered by tasty monkfish curry. The last batch of my local sieving should finish Sunday morning and I'll start the linear algebra then. Last fiddled with by fivemack on 2015-03-28 at 08:33
 2015-03-28, 08:59 #124 xilman Bamboozled!     "𒉺𒌌𒇷𒆷𒀭" May 2003 Down not across 97×113 Posts The NAS has been shutdown here and everything deleted, freeing up 257G. All the files are still on the back-up server put they'll probably get weeded out before too much longer because its disk is now 60% full and I'd prefer not to have important backups fail on me.
2015-03-29, 11:11   #125
fivemack
(loop (#_fork))

Feb 2006
Cambridge, England

144518 Posts

Uniqued-within-themselves relation counts from my side:

Code:
  114111349 200-248.u
5262696 248-252.partial.u
12248201 382-388.u
12203201 388-394.u
23702457 410-422.u
15785956 422-430.u
15649041 430-438.u
19365333 440-450.u
218328234 total
I'm not sure there's much point in doing a unique-among-all-218M followed by a merge, I'll just append all 218M to the 540M I collected and let msieve sort it out:

Code:
Sun Mar 29 03:08:51 2015  found 143702717 hash collisions in 758568029 relations
Sun Mar 29 03:09:09 2015  added 1219004 free relations
Sun Mar 29 03:09:09 2015  commencing duplicate removal, pass 2
Sun Mar 29 03:15:17 2015  found 71113210 duplicates and 688673823 unique relations
Code:
Sun Mar 29 07:38:35 2015  commencing 2-way merge
Sun Mar 29 07:39:57 2015  reduce to 122026455 relation sets and 78650729 unique ideals
Sun Mar 29 07:39:57 2015  commencing full merge
Sun Mar 29 07:54:49 2015  memory use: 6533.1 MB
Sun Mar 29 07:54:57 2015  found 53090605 cycles, need 48358929
Sun Mar 29 07:55:09 2015  weight of 48358929 cycles is about 5803766052 (120.01/cycle)

Sun Mar 29 11:26:47 2015  matrix is 48357161 x 48358929 (22297.4 MB) with weight 6307619226 (130.43/col)
Sun Mar 29 11:26:47 2015  sparse part has weight 5313189641 (109.87/col)
Sun Mar 29 11:46:25 2015  filtering completed in 3 passes
Sun Mar 29 11:46:33 2015  matrix is 48292695 x 48292895 (22280.7 MB) with weight 6302508887 (130.51/col)
Sun Mar 29 11:46:33 2015  sparse part has weight 5309538775 (109.94/col)
Sun Mar 29 11:51:29 2015  matrix starts at (0, 0)
Sun Mar 29 11:51:34 2015  matrix is 48292695 x 48292895 (22280.7 MB) with weight 6302508887 (130.51/col)
Sun Mar 29 11:51:34 2015  sparse part has weight 5309538775 (109.94/col)
Sun Mar 29 11:51:34 2015  saving the first 48 matrix rows for later
Sun Mar 29 11:51:42 2015  matrix includes 64 packed rows
Sun Mar 29 11:51:47 2015  matrix is 48292647 x 48292895 (21411.1 MB) with weight 5439342793 (112.63/col)
Sun Mar 29 11:51:47 2015  sparse part has weight 5129863044 (106.22/col)

Sun Mar 29 12:01:26 2015  linear algebra at 0.0%, ETA 1928h54m
The job is running with 26GB resident memory use and, all going well, should finish around Midsummer's Day.
Attached Files
 M991-nc1-log.txt (18.5 KB, 106 views)

Last fiddled with by fivemack on 2015-03-29 at 11:12

2015-04-27, 17:19   #126
R.D. Silverman

Nov 2003

22×5×373 Posts

Quote:
 Originally Posted by fivemack Uniqued-within-themselves relation counts from my side: The job is running with 26GB resident memory use and, all going well, should finish around Midsummer's Day.

Woe! Woe! Woe!

It is a sad state we are in. Woe is us....

Finishing the M991 sieving means that for the first time in a very long time, there
are no Cunningham composites from among the first 5 holes in any table
being sieved.

I can not recall the last time this happened.

BTW, Are Bai, Verser, and Texas State ever going to finish their reserved numbers?

2015-04-27, 18:06   #127
jyb

Aug 2005
Seattle, WA

178210 Posts

Quote:
 Originally Posted by R.D. Silverman Woe! Woe! Woe! It is a sad state we are in. Woe is us.... Finishing the M991 sieving means that for the first time in a very long time, there are no Cunningham composites from among the first 5 holes in any table being sieved. I can not recall the last time this happened.

2015-04-27, 18:12   #128
R.D. Silverman

Nov 2003

22·5·373 Posts

Quote:
 Originally Posted by jyb What about 10,359-?
Oops.... Clearly missed that.......

BTW, I wish this venue supported mpegs or avi's. I would have loved to include
Zero Mostel's lament at the phony funeral from A Funny Thing Happened on the Way
to the Forum.. His intonation was perfect.

2015-04-28, 13:46   #129
bdodson

Jun 2005
lehigh.edu

210 Posts

Quote:
 Originally Posted by jyb What about 10,359-?
RE: regular Cunninghams from first five holes being sieved

Another less visible feature of the first five, 10,359- C208 was tested
to 24.4t55=c t65 before being selected by NFS@Home; and
there's also the reserved number 2,2218M C206 tested to
20.6t55=c .8t65. Just counting Lehigh curves.

I'm just finishing a run through c200-c220 plus smaller first five holes,
with twelve more at .8t65 and thirteen more at/above t65. Looks
like another twenty-five at 3t60 =c 15t55, with Sam having
done another t60. I'm not claiming that we've reliably removed
p60's, especially among the first five above C280; but I haven't
found any below p60 this year, with three in p63-p67. -Bruce

2015-04-28, 16:23   #130
R.D. Silverman

Nov 2003

22·5·373 Posts

Quote:
 Originally Posted by bdodson RE: regular Cunninghams from first five holes being sieved Another less visible feature of the first five, 10,359- C208 was tested to 24.4t55=c t65 before being selected by NFS@Home; and there's also the reserved number 2,2218M C206 tested to 20.6t55=c .8t65. Just counting Lehigh curves. I'm just finishing a run through c200-c220 plus smaller first five holes, with twelve more at .8t65 and thirteen more at/above t65. Looks like another twenty-five at 3t60 =c 15t55, with Sam having done another t60. I'm not claiming that we've reliably removed p60's, especially among the first five above C280; but I haven't found any below p60 this year, with three in p63-p67. -Bruce
Nice work. Excellent.

However, it doesn't really matter if we miss a few p60's. What matters is whether the
composite can be done faster (less work) with NFS than running additional ECM
(which has a low probability of success).

2015-04-29, 16:40   #131
bdodson

Jun 2005
lehigh.edu

210 Posts

Quote:
 Originally Posted by R.D. Silverman Nice work. Excellent. However, it doesn't really matter if we miss a few p60's. What matters is whether the composite can be done faster (less work) with NFS than running additional ECM (which has a low probability of success).
I did 3t60s for the two current Cunningham gnfs records
Quote:
 6319 C221 3,697+ NFS@Home 6269 C216 3,766+ NFS@Home
which factored as
Quote:
 C216 = p66*p75*p76 C221 = p67*p76*p78, with p67=2430596119059914710337969915407030131984125918638366960356417238051 -vs- p66=303889341986146630791713973167874707042199651755239385807424842909
for the smallest factor. If I were doing the pretests now, I would have run the additional
2t60 = 10t55 on each, to 5t60 = t65, but still would have needed luck to get either one.
To get halfway up on p70-p65, for p67.5, maybe two more t65s, and still only 62%
chance of success; maybe one of them would have fallen. An additional 60t55, each,
from the 3t60 I did. Probably still better to run those curves on less tested numbers?
-Bruce

2015-04-29, 16:58   #132
R.D. Silverman

Nov 2003

1D2416 Posts

Quote:
 Originally Posted by bdodson I Probably still better to run those curves on less tested numbers? -Bruce
That would be my opinion.

 Similar Threads Thread Thread Starter Forum Replies Last Post fivemack Factoring 9 2015-07-04 20:12 jasonp Factoring 85 2014-12-01 20:01 fivemack ElevenSmooth 53 2008-04-28 08:53 MooooMoo Twin Prime Search 38 2006-06-08 04:12 TauCeti NFSNET Discussion 0 2003-12-11 22:12

All times are UTC. The time now is 21:58.

Wed Oct 27 21:58:07 UTC 2021 up 96 days, 16:27, 0 users, load averages: 0.97, 1.13, 1.09