mersenneforum.org My luck
 Register FAQ Search Today's Posts Mark Forums Read

 2018-01-12, 15:25 #1 vasyannyasha     "Vasiliy" Apr 2017 Ukraine 3F16 Posts My luck I didn't find 12-digit factor after 1000 curves with B1=11000 What are the odds? Last fiddled with by vasyannyasha on 2018-01-12 at 15:25 Reason: too many zeros
 2018-01-12, 17:10 #2 firejuggler     Apr 2010 Over the rainbow 50328 Posts what number did you test?
 2018-01-12, 17:18 #3 vasyannyasha     "Vasiliy" Apr 2017 Ukraine 32×7 Posts 111-digit number from aliquot sequence
 2018-01-12, 17:20 #4 vasyannyasha     "Vasiliy" Apr 2017 Ukraine 32×7 Posts GMP-ECM write only about odds for 35-digit factor and bigger.
 2018-01-12, 17:38 #5 VictordeHolland     "Victor de Hollander" Aug 2011 the Netherlands 23×3×72 Posts Assuming you're testing a number in a normal range (100-400 digits) With GMP-ECM t20 = 86 curves with B1=11,000 B2=1,873,422 1000 curves = more than 11 x t20 So after 1000 curves you should've found a 20 digit factor 11 times on average. Chance to miss a 20 digit factor after 1000 curves is very small e^(-11) = ~ 0.0000167 = 0.00167% Are you sure your software has compiled ok? Is the memory ok? Willing to share the number in question? So we can run a couple of hundred curves to check if GMP-ECM finds the factor on different hardware? Last fiddled with by VictordeHolland on 2018-01-12 at 17:50 Reason: number?
 2018-01-12, 17:52 #6 vasyannyasha     "Vasiliy" Apr 2017 Ukraine 32×7 Posts Before this I find dozens of different factors with ECM. Memory work fine. Maybe this was bug for 30 seconds. After this P-1 test used 2 GB and find this factor. Last fiddled with by vasyannyasha on 2018-01-12 at 18:02 Reason: My English
2018-01-13, 17:51   #7
WraithX

Mar 2006

11·43 Posts

Quote:
 Originally Posted by VictordeHolland Assuming you're testing a number in a normal range (100-400 digits) With GMP-ECM t20 = 86 curves with B1=11,000 B2=1,873,422 1000 curves = more than 11 x t20 So after 1000 curves you should've found a 20 digit factor 11 times on average. Chance to miss a 20 digit factor after 1000 curves is very small e^(-11) = ~ 0.0000167 = 0.00167% Are you sure your software has compiled ok? Is the memory ok? Willing to share the number in question? So we can run a couple of hundred curves to check if GMP-ECM finds the factor on different hardware?
According to my new web page, the chance to miss a 12 digit factor after 1000 curves at B1=11e3 is around 2.304 * 10^-203. My page shows 81e-6 (about 0.000081 or 0.0081%) to miss a 20-digit factor after 1000 curves at B1=11e3.

I think the differences in our numbers stems from GMP-ECM v7 recommending 106.7632 curves to find a 20-digit factor with B1=11e3 with param=1. I think you might be using param=0.

You can read more about my web page to calculate ecm success/failure chances here:

Last fiddled with by WraithX on 2018-01-13 at 17:53

2018-01-13, 18:28   #8
Batalov

"Serge"
Mar 2008
Phi(4,2^7658614+1)/2

223548 Posts

Quote:
 Originally Posted by vasyannyasha I didn't find 12-digit factor after 1000 curves with B1=11000 What are the odds?
If your darts do not hit the dart board, the odds are that there could be a problem with darts, or with the dart board, or with the air between the darts and the darts board, or with the thrower of darts. (In no particular order.)