Bug 99.99%

[paulunderwood]

Originally Posted by Ruicar

but 100 its mean is not primes my result is res64

Yes and that will be double checked by someone else.

[Ruicar]

sorry i understand my mistakes was on begin ok sorry for losing your time was the first time i use it and the last this take to much time for one worker and i made a big mistake you have reason can not be primes and not a mersenne

[Ruicar]

don't check it because is useless number

[Batalov]

What ... on earth ... is going on here?
75 messages? I already wrote (up there) to you that this is not twitter.
Don't

post

ten

messages

all one line long!

If you've got nothing to say - then don't!

[ MOD HAT ON ]Ruicar, you've been warned. [MOD HAT OFF/]

[Uncwilly]

I will post a long explanation message.

Unlike Trial Factoring, the Lucas-Lehmer (L-L) test is a long complicated test with many iterations (steps). The number of steps required depends on the size of the number being tested. If we are testing 2^1234567-1, about 1234567 steps are required (+-2, I forget). [Remember, that to be a Mersenne Prime, 1234567 also has to be prime. It is impossible otherwise.] Because we have to go through all of those steps (if it is prime we have to do them all to get the final answer, if it is not prime we have to do them all to get the final answer.) So, the program Prime95 displays how far along it is. (If it did not, you would have no feedback for the entire test, some tests can take months or years.) So, as the percent completed gets closer to 100%, there is no change in the likelihood that the number is prime, all of the Lucas-Lehmer tests must go to 100%.

Once the last step of the L-L test is complete, the result of the calculation is either zero (and the number is prime) or some very large number, the residual [think of it like the remainder in division, if it is not zero, it is not evenly divisible]. To make sure that the calculations did not make an error and miss a prime, the GIMPS project will have another user test the number again. If the residual matches, we know that both calculations matched. The RES64 that you see is the last little bit of the residual. That is all that we need to keep and compare to make sure the calculations match (to a reasonable certainty). Since the RES64 that your machine generated is not all zero, your number is not a Meresenne Prime. The error code your machine reported is zero, so it is likely that the calculation is good with no errors.


On a separate point of information: Your user log in information for the Mersenne Forum (mersenneforum.org) is not shared to the GIMPS project on PrimeNet (mersenne.org) nor vice versa. You must login into PrimeNet to manually submit your result. If you set up your copy of Prime95 normally, it should automatically submit the result. Make sure that you have set your PrimeNet user id in your copy of the Prime95 software.

I think that we have some forum members that speak French. So if that would be easier for you, post in French.
[Machine translation]
Je pense que nous avons des membres du forum qui parlent français. Donc, si cela vous serait plus facile, postez en français.[/Traduction automatique]

As noted by others: please post longer single posts, not several short ones. It makes it easier to follow your thoughts and gives you more room to develop your explanations.

If you still are having problems submitting your results after a few more tries (with more explanation of what exactly you have done [all of the steps]) you can if needed e-mail the result file to George

[Siegmund]

It means you have proven that that number is NOT prime.

A Lucas-Lehmer test runs ALL THE WAY TO THE END, and THEN gives you an answer, prime or not prime. Getting to 90% does not mean that you are 90%-sure you really have a prime number.

When I first joined the project (wow - 17 and a half years ago now) I had the same feeling.. I checked the program every day, it took 3 weeks to test a number, and towards the end I thought "wow! It's checked 99% of what it has to check and it still hasn't found any factors!" and was disappointed to see the final result. Then I went and read how the test worked, and knew what to expect next time.

[0PolarBearsHere]

Originally Posted by Siegmund

When I first joined the project (wow - 17 and a half years ago now) I had the same feeling.. I checked the program every day, it took 3 weeks to test a number, and towards the end I thought "wow! It's checked 99% of what it has to check and it still hasn't found any factors!" and was disappointed to see the final result. Then I went and read how the test worked, and knew what to expect next time.

Yes. Unfortunately when you don't need to factorise a number to determine whether it's a mersenne prime, these kinds of thoughts will occur. I think mine may have been the same.

[VBCurtis]

Originally Posted by Ruicar

if he reach 99.99 is not product by 2 primes?

You keep saying "99.99". As someone already told you in this thread, the percent is a progress report for the test, NOT the probability your test is an actual prime number.

Pretend the program is telling you "this test will take 7 days for a result. Your test is 99% done, please wait one more hour for a result."
An hour later "this test is 99.99% done, please wait one minute for the result of the test."
After 99.99%, the next update tells you the result of the test. Yours was not prime. That's all we know- the test is a yes or no result, prime or not prime. We don't know if it has two factors or 29, but we know (because you ran the test) that your number is not prime.

The test takes 7 (or more) days, and nothing is known about the test until it completes.
If your medical doctor took a blood sample to test for a disease, and told you you would have your result in 7 days, do you think you're 86% likely to have the disease because he didn't call you back after 6 days?

[Ruicar]

was that
/[Nov 5 22:44] Iteration: 66050000 / 66066061 [99.97%], ms/iter: 7.899, ETA: 00:02:06
[Nov 5 22:46] Iteration: 66060000 / 66066061 [99.99%], ms/iter: 7.909, ETA: 00:00:47
[Nov 5 22:46] M66066061 is not prime. Res64: 8F2D01829E4F843F. We10: 73F49AE0,9114609,00000000
[Nov 5 22:46] Setting affinity to run helper thread 1 on any logical CPU.
[Nov 5 22:46] Setting affinity to run helper thread 2 on any logical CPU.
[Nov 5 22:46] Setting affinity to run helper thread 3 on any logical CPU.
[Nov 5 22:46] Starting primality test of M66066061 using FMA3 FFT length 3456K, Pass1=768, Pass2=4608, 4 threads
[Nov 5 22:48] Iteration: 10000 / 66066061 [0.01%], ms/iter: 8.192, ETA: 6d 06:19
[Nov 5 22:49] Iteration: 20000 / 66066061 [0.03%], ms/iter: 8.187, ETA: 6d 06:11
[Nov 5 22:51] Iteration: 30000 / 66066061 [0.04%], ms/iter: 8.670, ETA: 6d 15:02
[Nov 5 22:52] Iteration: 40000 / 66066061 [0.06%], ms/iter: 8.717, ETA: 6d 15:52
he restart and no setting affinity to run sorry for all the post

[axn]

66066061 = 2399 * 27539 is a composite number and cannot be the index of a mersenne prime. This number is not tracked by GIMPS.

[pepi37]

We have PP before few months: I think it is PP clone
But maybe member is child and doesnot understund what we write . Or he dont know to read