mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Lounge (https://www.mersenneforum.org/forumdisplay.php?f=7)
-   -   Bad LL Tests (https://www.mersenneforum.org/showthread.php?t=161)

outlnder 2002-10-17 01:08

Bad LL Tests
 
I just looked at the "Bad tests" document found on the Status Page.

What will be happening to those exponents?

Will they be released to be redone, or have they already been triple checked and have 2 matching results??

Prime95 2002-10-17 04:00

They have already been triple-checked. The only way we know which tests are bad is when we finally get a matching double-check.

outlnder 2002-10-17 08:45

Are there any plans on telling the people who have turned in multiple bad tests??

This would let them know which machines had errors.

I noticed that I had 2 bad tests. In fact the last one on the list was mine. Fortunately I have corrected all the problems on those machines(crosses fingers and prays).

lycorn 2002-10-17 17:54

I´ve also been browsing that file, and noticed several exponents with an error count of 00000000. How were they found to be bad?
Second point: George says all exponents in that file have been triple-checked, so why are they still in the list? Because 3 different results were returned :shock: , or the result kept in the list is the one found to be wrong? In that case the exponent should be also in the [i]double-checked[/i] file, which I do not think is the case.
Third: I´ve done several double-checks over the last months, and I can find most of the exponents in the [i]double-checked[/i] file. There are a few I can´t find, neither in that file nor in the [i]bad[/i] file. And no, they are not the last ones I have tested... What could be the reason?

garo 2002-10-17 18:00

A result for an exponent is found to be bad when there are three results submitted and two of them match. Then the third one is said to be bad since it didn't match whereas the others did. Since the chances of two incorrect reults matching are od the order of 1/2^64 this is a reasonable assumption.
This list, bad.zip is a list of all results that are known to be bad.

The list hrf3.zip (tested once but not doublechecked) actually also contains those exponents for which two results have been returned and they did not match. In this case, obviously one and maybe even both are bad results but we won't know which one till two matching results have been turned in. Check this hrf3.zip for your "missing" exponents. If all these exponents are coming from the same machine then maybe something is wrong with that machine. Nothing certain, just a guess.

Kevin 2002-10-18 01:22

About informing people of errors, maybe something could be built-in so when an error occurs, a box will pop open the next time the person attempts to open the Prime95 window, telling them something may be wrong with their hardware, or something of the sort.

Would exponents that need triple checking qualify as "priority" exponents?

binarydigits 2002-10-19 12:11

[quote="Kevin"]Would exponents that need triple checking qualify as "priority" exponents?[/quote]

Generally yes, but not necessarily. Double checks are now running on exponents from about 5.6M to about 8.6M. Any mismatches being currently reported would certainly need priority triple checks if the exponent is from the lower end of that range, but not if from the higher end (at least, not yet).

lycorn 2002-10-20 21:22

Garo
Thanks for your hint. I checked hrf3.zip and in fact my "missing" exponents were there. Now I fully understand how the whole thing works. My "missing"exponents came from different machines and they all produced no errors, so I hope they are the correct results ;)

garo 2002-10-21 00:12

Yeah it's probably the other test that's bad then.
garo.


All times are UTC. The time now is 00:18.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.