20200917, 00:18  #1 
Aug 2020
2·3·19 Posts 
Is it realistic to calculate and store the following info
For every (composite) mersenne number, if it had a factor of x bits/digits, how likely would have the factor been already discovered.
If we store by digits from 21 to 70 then there will be 50 floating point numbers for each exponent. Storing by bits for the same range will yield about 166. Either way it's not a huge demand for storage, so the only problem is to calculate it correctly. Last fiddled with by Ensigm on 20200917 at 00:21 
20200917, 02:08  #2 
Apr 2010
Over the rainbow
100111010110_{2} Posts 
Yes it is possible but it is useless info. there is 3 method of factoring Mersenne.
PM1, Trial factoring, and ECM. PM1 is probalistic. 53853801 has a factor. When I used it , the algo calculated that I had a 7.04% chance of finding a factor. It doesn't say anything about the size. TF has a 1/bit level chance of finding a factor. again unrelated to size. while ECM (eliptic curve method) work by size of factor. (called T level). If you complete T35, there is no chance of any factor below 35 digit on that particular M exist. Edit : Sorry for the multiple edit. I wanted to make myself clear. And sorry for anyone quoting me in between edit Last fiddled with by firejuggler on 20200917 at 02:26 
20200917, 02:14  #3  
"Curtis"
Feb 2005
Riverside, CA
2×2,311 Posts 
Quote:
It's more accurate to say that after a T35 is done, chances of a 30 or smaller digit factor is nil (1/e^6 or worse). A 33 digit factor has about a 10% chance (1/e^2, roughly) to be missed by a T35. 

20200917, 07:38  #4  
Aug 2020
114_{10} Posts 
Quote:
If a bit level has been TF'ed, then if a factor is in that level, the probability of discovering that factor would be 100%. 

20200917, 16:01  #5 
"Curtis"
Feb 2005
Riverside, CA
2·2,311 Posts 
This is correct he meant the odds of finding a factor have nothing to do with the size of the input number. That's why the data you want to collect would not be useful none of the 3 methods he mentioned have odds changed by the size of the input.

20200917, 19:54  #6  
Aug 2020
2·3·19 Posts 
Quote:
I never mentioned the size of the input number. What I was talking about can be paraphrased as, in your words, "the chance of missing an xdigit factor (if one exists)". For each exponent, this will produce a line chart with bit or digit as the xaxis and probability as the Y axis. This can be useful when people manually do curves with bounds that are not the typical values of T25, T30, etc. Under current rules I assume they are rounded down to the closest typical value? (Such as if I do 100 curves with B1=1e5, B2=1e7, they will only be counted as 100 T25 curves.) 

20200917, 20:05  #7  
Bamboozled!
"πΊππ·π·π"
May 2003
Down not across
10100100010000_{2} Posts 
Quote:
What I do with the ECMNET server running at 83.217.167.17:8194 is first to multiply the number of curves by the B1 reported by a client and use that as a measure of work done. Each candidate in the server has its own record of the amount of work done by that measure. That total is used to set an appropriate B1 for a client when the candidate is allocated to it. 

20200917, 21:09  #8 
"Curtis"
Feb 2005
Riverside, CA
2×2,311 Posts 
mersennne.org does something similar I run ECM with B1=80k, B2=100*B1, and every 5 curves I run earns credit of 8 standard B1=50k curves. It's not a perfect conversion, but it works well enough for B1 near the expected Tlevel.
Where it breaks down is for B1 choices far smaller than useful e.g. M1277; 1000 curves at B1 =1e6 are useless there, but get credited at just over 1 curve @850M. We hope people don't do that... 
Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Is it possible to convert Prime95 to UWP and launch it in Microsoft Store?  Bulldozer  Software  7  20190830 14:37 
New York: Corpse Wheeled to CheckCashing Store  ewmayer  Lounge  2  20080115 16:18 
Best way to store huge data  HiddenWarrior  Data  18  20051011 03:53 
store prices  Fusion_power  Puzzles  7  20030831 01:37 
I hope the new Apple store in Chicago opening today...  Paulie  Software  4  20030716 14:19 