20220717, 01:41  #78 
Sep 2008
Kansas
111001011001_{2} Posts 
Taking 143__553_17m1.

20220717, 18:12  #79 
Jul 2003
So Cal
7×353 Posts 
C193_3408_1720 is done.

20220717, 20:32  #80 
"Serge"
Mar 2008
Phi(4,2^7658614+1)/2
26D3_{16} Posts 

20220718, 23:07  #81 
"Oliver"
Sep 2017
Porta Westfalica, DE
2^{2}×7×41 Posts 
Taking 7489_71m1.

20220719, 05:59  #82 
Aug 2020
79*6581e4;3*2539e3
259_{16} Posts 
5p2_950M is finally done...
Code:
Mon Jul 18 21:28:49 2022 p72 factor: 827973016668650713388064731991152830057402447576623382393481045697281201 Mon Jul 18 21:28:49 2022 p120 factor: 175729134090666610433244159968147764083858403497721040118931072065628823538382099068657470560738914506844459897080986001 Last fiddled with by bur on 20220719 at 06:01 
20220721, 12:21  #83 
Aug 2020
79*6581e4;3*2539e3
601 Posts 
And taking 9p8_277.

20220721, 20:36  #84 
"Oliver"
Sep 2017
Porta Westfalica, DE
10001111100_{2} Posts 
7489_71m1 can be moved to postprocessing. With TD 120, it build a matrix of size 27M.
Do we have a guide on how to select the first TD value that should be tried in dependecy of the relation count? 
20220721, 21:40  #85  
Jun 2012
2×7×263 Posts 
Quote:
Seriously it’s roughly a function of siever used (14d, 15e, 16e_small) and lpb (31,32,33). I personally use historical data and experience. I would use TD=110 for a 31bit job on 14, and go with TD of 130+ for the biggest jobs. IIRC the biggest TD I can remember ever seeing is 140. In this case I think TD of 126 or 128 would probably build a matrix. But higher TD does not guarantee a huge reduction in solve time  sometimes it’s just a few hours to maybe half a day savings. The above is what has worked for me, obviously not the optimal solution backed with rigorous analysis! ETA: the number of rels is obviously a factor but I tend to use the same number of rels for a given lpb value. Last fiddled with by swellman on 20220721 at 21:45 

20220722, 16:29  #86 
Jul 2003
So Cal
7×353 Posts 
C213_146_132 is done. I used this as a test run on the NCSA Delta cluster on 4x A100 gpus. The 26M matrix took 3.5 hours to solve.

20220722, 19:43  #87  
"Curtis"
Feb 2005
Riverside, CA
1560_{16} Posts 
Quote:
It's a useful experiment to take a job from 15e or a big esmall job and filter with TD 120 then start the matrix step and note ETA, then repeat at 128 or 130. I doubt the ETA will change much, like on the order of 5% of expected time. So, i'd rather not filter multiple times... but I may be mistaken about the time savings. 

20220725, 11:16  #88 
Jun 2012
2·7·263 Posts 
7m6_323 Factored
Code:
prp77 factor: 18790799179598956677809813233818970955254712169449836196099826918661598374937 prp116 factor: 38383382908158359751904523723457649684098373309026463919143035424526538480306292084304646382999001824515465550318037 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
2022 “small” 15e post processing reservations and results  swellman  NFS@Home  140  20220927 05:26 
2022 14d post processing reservations and results  swellman  NFS@Home  108  20220926 23:03 
2021 14d post processing reservations and results  swellman  NFS@Home  279  20220101 04:22 
2021 16e post processing reservations and results  swellman  NFS@Home  37  20211207 02:52 
2019 15e post processing reservations and results  swellman  NFS@Home  220  20191228 16:40 