20210305, 19:45  #936 
"Curtis"
Feb 2005
Riverside, CA
142D_{16} Posts 
Is there a sense that the filtering bug was squashed, and only the size of the dataset is in the way of filtering jobs this size?
Seems 33/35 is better than 33/36 for this job, to shrink the dataset by 25% (say, 2.1B uniques rather than 2.8B). I suppose it only matters in principle, since matrix size also stands in the way of trying an eventougher job. Bravo for slaying this beast, all the same! Well, almost done.. 
20210307, 05:23  #937 
Jul 2003
So Cal
8D0_{16} Posts 
I think so... Certainly hope so! Filtering with lots of excess relations is still hard so things could go awry there, but with a reasonable excess the remaining limit is 2^32 total, not unique, relations. With prior duplicate and perhaps singleton removal, this permits large runs.
The size of the dataset isn't much of an issue. The filtering running time is dwarfed by both the sieving and LA running times anyway, and someone completing numbers this size likely has access to a computer with sufficient storage and memory. Code:
200G Mar 6 01:55 msieve.dat.gz reading ideals above 720000 commencing singleton removal, initial pass memory use: 41024.0 MB reading all ideals from disk memory use: 87634.0 MB keeping 1372156493 ideals with weight <= 200, target excess is 10188461 Last fiddled with by frmky on 20210307 at 05:28 
20210309, 11:38  #938 
Oct 2007
Manchester, UK
2·683 Posts 
Finally finished my ECM work on the t2200 file!
Composites larger than 2^1018 have had 224 curves @ B1=50e3, B2=13.7e6 (25 digit level). Composites less than 2^1018 have had 1152 curves @ B1=3e6, B2=14e9 (40 digit level). The total haul is 815 factors, these have been submitted to factordb, and attached to this post. 
20210325, 12:10  #939 
Apr 2006
103 Posts 
Thanks! That is a lot of factors, with an explicit lower bound on the ECM effort for these numbers.

20210326, 16:59  #940 
May 2003
3×97 Posts 
That's a great idea. What is the current ECM status on the t2200 composites?

20210329, 18:31  #941 
Jun 2012
Boulder, CO
2^{2}×3×29 Posts 
Here's the latest on the #1 MWRB, with frmky lending a helping hand for the LA:
Code:
linear algebra completed 20216008 of 109441779 dimensions (18.5%, ETA 854h19m) 
20210505, 18:32  #942 
Jun 2012
Boulder, CO
2^{2}·3·29 Posts 
Thanks in large part to frmky's help with the filtering fix and running LA, the #1 MWRB has fallen.
http://factordb.com/index.php?query=...1454629%5E171 
20210505, 18:49  #943 
Jul 2003
So Cal
2^{4}·3·47 Posts 
Here's the log for posterity.

20210505, 19:50  #944 
"Carlos Pinho"
Oct 2011
Milton Keynes, UK
2·3·5·167 Posts 
Congrats, well done gents!

20210505, 20:19  #945 
Apr 2006
103 Posts 
Ryan, frmky, thank you.
Now I explore the subtrees unlocked by the new factors for lower bounds on many parameters of an OPN. 
20210505, 21:36  #946  
Apr 2020
1001010101_{2} Posts 
Quote:
Glad to see this completed successfully without any further bugs. I presume this means the 4G relation limit is now the limiting factor in the size of jobs that can be run with msieve? 

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Passive Pascal  Xyzzy  GPU Computing  1  20170517 20:22 
Tesla P100 — 5.4 DP TeraFLOPS — Pascal  Mark Rose  GPU Computing  52  20160702 12:11 
Nvidia Pascal, a third of DP  firejuggler  GPU Computing  12  20160223 06:55 
Calculating perfect numbers in Pascal  Elhueno  Homework Help  5  20080612 16:37 
Factorization attempt to a c163  a new Odd Perfect Number roadblock  jchein1  Factoring  30  20050530 14:43 