Strategic Double Clicking

[sdbardwick]

Originally Posted by PBMcL

Another from my glacial (but reliable) old Mac. 36674201 needs a TC.

We matched.

[frmky]

38746493 needs a quad check.

[Mark Rose]

Originally Posted by frmky

38746493 needs a quad check.

Mine.

[UBR47K]

Triple Checks would be appreciated on the following CuLu results:

DoubleCheck=36848387,71,1
DoubleCheck=38671361,71,1
DoubleCheck=66921341,74,1
DoubleCheck=64847711,75,1

[fivemack]

Based on a discussion elsewhere in the forum: madpoo, do you have the database access sufficient to produce a list of numbers where the reported residue is 00000000000000__ ?

[Madpoo]

Originally Posted by fivemack

Based on a discussion elsewhere in the forum: madpoo, do you have the database access sufficient to produce a list of numbers where the reported residue is 00000000000000__ ?

I do, but bear in mind that except for known primes, the ones where it's masked like that usually have the unmasked part "02", or they were false positives.

(I know, spoiler alert, there are NOT any hidden primes that haven't been announced yet, LOL)

[Dubslow]

Originally Posted by Madpoo

I do, but bear in mind that except for known primes, the ones where it's masked like that usually have the unmasked part "02", or they were false positives.

I believe that's his point. This is the Strategic Double Checking Thread after all.

[Madpoo]

Originally Posted by Dubslow

I believe that's his point. This is the Strategic Double Checking Thread after all.

When I first poked around in the data, I found all the tests with 00000000000000__ in them and ran my own double (or triple) checks. Of course most of those ended with 02 (also an 03 and 04) and were bad, and then there were one or two goofy false positives where other people had already done double/triple checks.

My goal was to a) satisfy my own curiosity and b) prevent the speculation that would occur if someone saw a masked residue like that.

EDIT: Examples:
M66921341
M62891251
M79060367 (trust me when I say the masked part is NOT "00")

[fivemack]

Unless I am misreading the status page something rotten, 79060367 hasn't been double-checked and is not assigned to anybody.

Nearly-all-zeroes residue nearly always means 'the FFT data got zeroed at some point' (possibly 'a checkpoint file was written while the disc was full and then prime95 restarted from the all-zeroes checkpoint')

[LaurV]

Originally Posted by Madpoo

M79060367 (trust me when I say the masked part is NOT "00")

First, this seems to be unassigned and there is no residue showing non-zero stuff (like your double check you talk about). Then, his point was that if such cases still exist (obviously yes, as you just proved), then it should be nice to have a list with them, so someone can pick worktodo from it and eliminate all the doubt. Can you generate such a list and make it public?

Third, even if all are zero, it may not mean that the number is prime. This is only the tail of the full residue. Statistically, we can get a result with all the last 16 hex digits being zero every 18,446,744,073,709,551,616 exponents...

[Madpoo]

Originally Posted by fivemack

Unless I am misreading the status page something rotten, 79060367 hasn't been double-checked and is not assigned to anybody.

Nearly-all-zeroes residue nearly always means 'the FFT data got zeroed at some point' (possibly 'a checkpoint file was written while the disc was full and then prime95 restarted from the all-zeroes checkpoint')

Yeah, that one was newer and I hadn't seen it yet. I'd do a DC on it but ever since I figured out why it's (somewhat) common for an error to result in a residue of 0x2 I'm not too worried about those.

Eventually it'll get double and triple checked, but it's clear the first test was bad.