M40, what went wrong?

[trif]

Originally Posted by jeff8765

What are the odds that a test would give a false positive?

Whatever the historical error rate is times 2^64.

[jeff8765]

So is it almost impossible for their to be an accidental false positive? Also when will the next double check be done?

[Darak]

My verification run of M#40 just completed....not prime.

I can't imagine being George and seeing the result come up as not prime.... He's sitting there at the computer and is "HERE IT COMES HERE IT COMES; OH BABY!!!!" "UGH!!!"

I can not think of a better example for the policy of not informing the media until verification runs have been made. As it stands now no damage is done regardless if it pans out to be a fluke or not.

[Prime95]

Originally Posted by jeff8765

So is it almost impossible for their to be an accidental false positive? Also when will the next double check be done?

Yes, the chance of a false positive from a working program is zero.

I've been thinking about how prime95 could report a false positive. Past posts have mentioned that if some memory corruption event occurs that zeroes all the FFT data then the next iteration will generate -2. However, it is possible that the memory corruption could also screw up the code that subtracts 2 from each iteration. Ever since the random shift was introduced in version 17, the subtract 2 code uses a variable called units_bit to decide which FFT word to subtract 2 from. If that variable was also clobbered, then I suppose its possible no subtract 2 takes place and you get a lot of iterations with zero as the result.

Next double-check will be done on Friday.

[jeff8765]

Is it possible for someone to delibrately report a false positive with the program or would it be impossible to send a false positve to Primenet?

[Prime95]

More bad news to report. I've just gotten a hold of the results.txt file from the original run. There were five SUM(INPUTS) != SUM(OUTPUTS) error during the run. Historically, five errors during a run means there is less than a 50% chance the final LL result will be correct.

Now I have to try and figure out what safety checks can be added to prevent false alarms.

So far, I have 2 changes planned:

1) Prime95 does not tell the server how many errors occurred for an LL test with a prime result. I'll fix this. Knowing their were 5 errors would have raised red flags earlier on.

2) I'll change prime95 to NOT delete the save files when a new prime is found. When a prime is reported, I'll ask for the save file and rerun the last 30 minutes of the test. I think this would have helped in this case. This measure also deters a hacker. It might be easy for a hacker to spoof a prime report - but he'll have difficulty generating a save file that reports the number as prime after the final 30 minutes of LL testing.

There is still a chance this really is prime, we'll know on Friday.

[Prime95]

Originally Posted by jeff8765

Is it possible for someone to delibrately report a false positive with the program or would it be impossible to send a false positve to Primenet?

It would not be hard for a hacker to figure out how to do this. I am confident that is not the case here.

And if you think I feel bad, imagine how this poor fellow feels....

[trif]

Originally Posted by Prime95

It would not be hard for a hacker to figure out how to do this. I am confident that is not the case here.

In my experience, those who have worked to accumulate the knowledge and skill necessary to do something like this very rarely go around doing destructive, malicious, or stupid things. Those who do are either 1) immature and doing it to impress someone else or 2) being paid to do it. Since it's hard to imagine 2, and there are a lot better targets out there for 1, it's unlikely that someone would maliciously report a prime that wasn't.

When I was working with distributed.net, we did have people who tried easy things to run up their stats (some succeeded for awhile). But we had at least a half a dozen people (that I know of) who reverse engineered the communications code (which was held secret like Prime95's), and as such, could have faked as many false results as they liked. There was never any indication that any of these people did so. People who do real work understand the value of work done by others, and are much less likely to go around destroying things for the hell of it.

[Boulder]

If you think of any bright sides here, at least George will surely come up with new ideas to make prime95 more accurate once again (as he already has given things a lot of thought).

I really feel for the poor soul who found the candidate :(

[jocelynl]

Xyzzy told us it was a triple check. So I think it's from the batch that George released in the 8M.

[AP]

No, it would have been the largest known prime so the exponent has to be > 13.466.917......

Regards
Achim