Why "DoubleCheck"?

[yankeesmarc923]

hey everyone

saw the article on cnn.com...searched for the marsenne site and downloaded the software. im currently on test 18 of the Lucas-Lehmer iterations. i have one question. my worktodo file says "DoubleCheck=10831231,64,1." so basically i will be double checking numbers? not testing new ones?

lol, its fine with me...anyway i can help. are the numbers only checked twice or do they just continuously rotate all the already checked numbers through us petty folk with the weak computers. anyway...anything i can do to help would be great.

[nfortino]

Assuming the two results match, exponents are only checked twice. If you wish, you can tell the program to take any type of work you want by opening the program, and selecting primenet under the test menu. Before you do so, check to see when your current exponent will finish by going to staus under the same menu. A first time check will take about four times as long as the double check you have now. If you are feeling pacient, there is nothing wrong with tacking an exponent which will take a long time (unless it is more than six months, in which case you probably stick to double checks.)

[Complex33]

Welcome aboard!

Prime95 has determined that the speed of your computer is most suitable for double checking work. There is a historic error rate of around 2% for LL tests, therefore to prevent any missed primes every exponent is checked twice and the residuals compared. The double check is usually run a couple of years after the first time test. You can by all means test first time exponents if you so wish, it will take a lot of patience though on your hardware. Again welcome and good luck!

[yankeesmarc923]

thanks guys...i've started on doublechecking M10831231 i was wondering what an average "per iteration time" is. lol...with my computer mine is probably 10 times the average! any info would be appreciated. thanks!

[GP2]

Originally posted by yankeesmarc923
are the numbers only checked twice or do they just continuously rotate all the already checked numbers through us petty folk with the weak computers. anyway...anything i can do to help would be great.

When a number is Lucas-Lehmer tested, the end result is just a very big number. If that number is zero, you've just discovered a prime.

If the number is non-zero, then the last 64 bits are saved and recorded (this is the "residue"). When a second test is done (double check) on the same exponent -- often years later -- a new residue is found, and normally it should match the previous one. The odds against an accidental match are 1 in 2⁶⁴ (about 16 billion billion).

In a few cases, the residues don't match and a triple check needs to be done. This is fairly rare, however since the project has tested millions of exponents, there have even been a small number of cases where quadruple or higher checks were needed to find two matching residues for an exponent.


There is never any need to re-send already double-checked exponents as "busywork"... the server will just send you the next highest available number. We'll never run out of numbers.

[Prime Monster]

Originally posted by yankeesmarc923
thanks guys...i've started on doublechecking M10831231 i was wondering what an average "per iteration time" is. lol...with my computer mine is probably 10 times the average! any info would be appreciated. thanks!

You can check your computers processor against similar processors on the Benchmarks page: http://www.mersenne.org/bench.htm

PM