20211120, 18:57  #298  
Apr 2020
593 Posts 
Quote:
I remember reading somewhere that subsequent changes to CADO mean that this doesn't work anymore. CADO's filtering is more efficient but its matrices are denser, so even if you could get a 60M CADOproduced matrix to run in msieve I don't know if it would be much faster than a 72M msieveproduced matrix. 

20211120, 19:28  #299 
"Oliver"
Sep 2017
Porta Westfalica, DE
2^{3}×3×5×7 Posts 
Unfortunately, I do not know how to do this other than setting up a HTTP server manually on that port afterwards.

20211120, 20:26  #300  
"Ed Hall"
Dec 2009
Adirondack Mtns
2·3^{2}·233 Posts 
Quote:
Thanks! 

20211121, 18:48  #301 
"Oliver"
Sep 2017
Porta Westfalica, DE
840_{10} Posts 
We are getting there! This was with nearly 1.3B relations and TD 124 (and 16 threads because I forgot to remove it).
Code:
found 346536029 duplicates and 950738834 unique relations ... Linear algebra completed 4299 of 69265750 dimensions (0.0%, ETA 3137h10m) 
20211121, 20:47  #302 
"Ed Hall"
Dec 2009
Adirondack Mtns
2×3^{2}×233 Posts 
Might we want to change this:
Code:
echo n "<p>Based on 1.3B goal: ${relspdone:0:2}.${relspdone:2}% finished – ETA: " >> $HTML_FILE printf "%(%Y%m%d %H:%M:%S)T</p>\n" $tleft >> $HTML_FILE Code:
echo n "<p>Based on 1.3B goal: ${relspdone:0:3}.${relspdone:3}% finished" >> $HTML_FILE 
20211121, 21:05  #303 
"Oliver"
Sep 2017
Porta Westfalica, DE
2^{3}·3·5·7 Posts 
Yes, I saw the problem. My ideas were:
Last fiddled with by kruoli on 20211121 at 21:13 Reason: Correcting automatic correction. 
20211122, 02:37  #304 
"Curtis"
Feb 2005
Riverside, CA
7·11·67 Posts 
I vote to stop sieving Thursday. That should be 1.4xG relations, and pretty far down the "more relations makes a smaller matrix" curve. The matrix should be 60M or so by then.
You can use TD of 130; I believe that's what Greg uses on the big NFS@home jobs, even when they are very oversieved. I wonder if the large lim choices brought us a larger matrix than expected for the number of unique relations gathered. 
20211122, 07:07  #305 
"Oliver"
Sep 2017
Porta Westfalica, DE
2^{3}×3×5×7 Posts 

20211122, 19:07  #306 
"Oliver"
Sep 2017
Porta Westfalica, DE
2^{3}·3·5·7 Posts 

20211123, 11:38  #307 
Just call me Henry
"David"
Sep 2007
Liverpool (GMT/BST)
3×5×397 Posts 
Is the plan to do this matrix on a CPU or to pass it off to a GPU? I get the impression frin https://mersenneforum.org/showthread.php?t=27042 that a GPU would do it far faster although it would need to be a high memory GPU.

20211123, 12:35  #308 
"Oliver"
Sep 2017
Porta Westfalica, DE
1101001000_{2} Posts 
My plan was to run it on a CPU since I do not have any working GPU in the moment (beside some Quadro 600 and a GT 1030), and especially nothing near the calibre I would need for this matrix. I want to save up for a "big" card but I will not buy one in the next few months. Maybe late next year. Those high memory cards are extraordinarily expensive.
If anyone else is eager do run this on a GPU, I will put the data up for download. The data should be exchangeable via Internet in a day or less I hope. 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Using 16e on smaller numbers  fivemack  Factoring  3  20170919 08:52 
NFS on smaller numbers?  skan  YAFU  6  20130226 13:57 
Bernoulli(200) c204  akruppa  Factoring  114  20120820 14:01 
checking smaller number  fortega  Data  2  20050616 22:48 
Factoring Smaller Numbers  marc  Factoring  6  20041009 14:17 