Trippple Checks

[GP2]

Originally Posted by ATH

I managed to run my own double check by mistake. I forgot to check them:
http://mersenne.org/M44356541

If anyone got time at some point to do another run.

I will run it.

[GP2]

Some triple checks where Madpoo mismatched the first test. His result is almost certainly correct.

Anyone want these?

DoubleCheck=54515387,73,1
DoubleCheck=54556591,73,1
DoubleCheck=55096187,73,1
DoubleCheck=55239377,73,1
DoubleCheck=55633247,73,1
DoubleCheck=55866413,73,1
DoubleCheck=56035241,73,1
DoubleCheck=57031193,73,1
DoubleCheck=57076309,73,1
DoubleCheck=58156037,73,1
DoubleCheck=58996009,73,1
DoubleCheck=59086387,73,1
DoubleCheck=59095571,73,1
DoubleCheck=59108381,73,1

[Mark Rose]

Originally Posted by GP2

Some triple checks where Madpoo mismatched the first test. His result is almost certainly correct.

Anyone want these?

DoubleCheck=54515387,73,1
DoubleCheck=54556591,73,1
DoubleCheck=55096187,73,1
DoubleCheck=55239377,73,1
DoubleCheck=55633247,73,1
DoubleCheck=55866413,73,1
DoubleCheck=56035241,73,1
DoubleCheck=57031193,73,1
DoubleCheck=57076309,73,1
DoubleCheck=58156037,73,1
DoubleCheck=58996009,73,1
DoubleCheck=59086387,73,1
DoubleCheck=59095571,73,1
DoubleCheck=59108381,73,1

Queued.

[Madpoo]

Originally Posted by GP2

Some triple checks where Madpoo mismatched the first test. His result is almost certainly correct.
...

Thanks for spotting those. I block out my own results when looking for mismatches to run...I forget about the ones I've done strategic double-checking on and got mismatches. Of course, I assume mine are right, so...

[Madpoo]

These are always fun...

M45075623

[moebius]

This one is funny too...
45305951

[Madpoo]

These 4 tests need a triple (and 44611877 needs a quad) check... I did one of the tests already.

DoubleCheck=42051857,72,1
DoubleCheck=44611877,72,1
DoubleCheck=56713931,73,1
DoubleCheck=68842393,75,1

No rush, since I'm pretty sure mine is correct, but just in case anyone felt like finishing those off.

[rudi_m]

Originally Posted by Madpoo

These 4 tests need a triple (and 44611877 needs a quad) check...

I've queued them.

[GP2]

Can someone poach M47092121 and complete the triple check quickly?

AdvancedTest=47092121

This is the first LL test I ran on the new c5 instance type on AWS, and it didn't match the first-time test.

The trouble is, this wasn't a strategic double check. The machine that did the first-time check has no prior bad results (although it has only two verified results, in the 26M and 27M ranges).

On the c4 instance type there have been no bad results out of a few thousand, so I hope this is a coincidence and not some bug or problem with c5 or mprime.

There was one unusual thing I did during the run of this exponent: I suspended the Linux process with kill -s SIGSTOP in order to run some tests on Mlucas and then resumed it with kill -s SIGCONT. I think I did that on c4 instances a few times in the past without any harmful effects.

[Madpoo]

Originally Posted by GP2

Can someone poach M47092121 and complete the triple check quickly?
...
This is the first LL test I ran on the new c5 instance type on AWS, and it didn't match the first-time test.

The trouble is, this wasn't a strategic double check. The machine that did the first-time check has no prior bad results (although it has only two verified results, in the 26M and 27M ranges).

I wouldn't worry about it, your result is probably fine. The strategic stuff is only a best guess anyway. While it's true in general that past performance predicts future results, there are exceptions... good past results may wind up with bad results for any # of reasons (memory flaked out, the user decided to try an overclock, heat wave, etc). And vice versa... bad results galore, and then the user fixed something and now they're rock solid.

I've run into quite a few of those myself where I didn't match either previous result (leading to the need for quad checks) and nothing really stood out about the past performance of either system.

[GP2]

Originally Posted by Madpoo

I wouldn't worry about it, your result is probably fine.

Way too much of a coincidence for my liking. We'll see. I'm now double-checking a bunch of other exponents that were tested by that user/computer.