20180314, 21:29  #1 
Mar 2018
Cambridge, MA
2 Posts 
Why have ECM testing for known nonprime Mersenne numbers?
Out of curiosity, I requested that my computer get ECM assignments. I got assignments for numbers like M48437, M65413, etc. These are relatively small Mersenne numbers that we already know are not prime. Why is GIMPS asking my computer to test them further?

20180314, 22:34  #2 
6809 > 6502
"""""""""""""""""""
Aug 2003
101×103 Posts
2^{2}·2,393 Posts 
Part of the GIMPS entire project is finding factors for exponents. There are folks looking at them. Lower numbers like the ones you mentioned are now best served by ECM, not TF or other methods.

20180314, 23:11  #3 
P90 years forever!
Aug 2002
Yeehaw, FL
2^{2}·1,873 Posts 
If your computer is rather slow or not running many hours a day, then the server may have determined it would take too long to run a doublecheck test, instead giving you the quicker ECM assignment.
You can force the work type you prefer in the Test / Worker Windows dialog box. 
20180315, 03:24  #4 
Mar 2018
Cambridge, MA
2_{10} Posts 

20180315, 13:44  #5 
Sep 2003
2·5·7·37 Posts 
You yourself requested ECM testing for your computer. In practice, ECM is only done for Mersenne exponents that already have known factors.
ECM is not costeffective for exponents that haven't been firsttime tested (in the 70 million range) or exponents that haven't been doublechecked (in the 40 million range). Using ECM to look for factors of such exponents would take longer than just doing LucasLehmer (LL) tests. It is costeffective to do trialfactoring (TF) or P−1 testing to search for factors of such exponents, in order to save time and skip some LucasLehmer tests. However, as mentioned above, it isn't costeffective to do ECM if your goal is solely to look for Mersenne primes. However, ECM is capable of finding factors that TF and P−1 aren't able to find, and some consider it interesting to search for additional factors just for fun. This is usually done only for fairly small exponents. Another reason that the software provides ECM is that some older computers simply can't do LL testing fast enough for exponents in the ranges currently being tested. They might take a year or more to do a single exponent, and in that time their assignment would probably expire, and the exponent would be reassigned to another user. However, those older computers can still do interesting work by searching for additional factors using ECM. 
20180316, 07:43  #6 
Sep 2002
Oeiras, Portugal
3·487 Posts 
Well, not quite. A considerable amount of the global ECM effort, I would say the largest part of it, is geared towards finding factors for numbers that, although already LL tested and DCed, thus being necessarily not prime, have not yet any known factor.
Last fiddled with by lycorn on 20180316 at 07:43 
20180316, 14:00  #7 
Sep 2003
101000011110_{2} Posts 
Yes, I should have written "already known to be composite" rather than "already have known factors". Just got muddled there for some reason. I've even run a few ECM tests myself on exponents with no known factors.

20181205, 19:56  #8 
Dec 2015
Local Group
2^{3} Posts 
I was recently assigned ECM work via the "Whatever makes sense" option (for the first time that I can recall), had this same question, and found this thread. Finding factors through ECM just for fun is fine... I'm wondering though if there's anything it's useful for otherwise... i.e. they'll end world hunger or get us to Mars faster.
Last fiddled with by section31 on 20181205 at 19:58 
20181205, 20:14  #9 
If I May
"Chris Halsall"
Sep 2002
Barbados
2×4,787 Posts 

20181206, 05:26  #10  
Sep 2018
3·23 Posts 
Quote:
Quote:
Also, sometimes GIMPS, via "Whatever makes sense", will assign ECM because for whatever reason, on a new install, the computer (client) reports as a 100mhz. After crunching ECM for a few days the server realizes it is fine and then assigns a DC instead. 

20181206, 15:15  #11 
Feb 2017
Nowhere
2^{6}×71 Posts 
I suppose it is also possible that, TF and P1 being done, and the large cofactor still being composite, ECM might split off another factor, after which the remaining cofactor might be a very large PRP. That would be a nice feather in someone's cap...

Thread Tools  
Similar Threads  
Thread  Thread Starter  Forum  Replies  Last Post 
Sum of prime divisors for Mersenne Numbers?  kurtulmehtap  Math  3  20110119 18:48 
Are Mersenne numbers more likely tobe prime?  davieddy  Lounge  23  20080614 17:50 
Mersenne Prime Factors of v.large numbers  devarajkandadai  Miscellaneous Math  6  20060104 22:44 
A property of prime Mersenne numbers under LLT  T.Rex  Math  12  20050912 07:56 
ECM on Mersenne numbers with prime exponents  biwema  Math  5  20040421 04:44 