lowest 10M exponents available - pick your choice

[tallguy]

Originally Posted by tha

Test=33801619,68,1

ETA May 11th...

[Batalov]

Originally Posted by cochet

When there is an error code, like that :
33600877,Serge Batalov,homeQ6600,97A2A2D3914C6C__,00000300

do the signification is the necessity of LLtest again for the first time ?

Of course, it is definitely a new candidat for M45 ... just take it and check. Be my guest. How many LLs did you run? or do you do only numerology?

_________diatribe follows_________

Seriously, errors happen. But their impact is less when it is a "restart from the last saved file" after the cause of the momentary problem was removed. All 3 errors happened at once and the mere fact that the restarts occured is always recorded in the save file and I don't f... ehh ...fiddle with the save files. Yeah, there were errors and I am not proud of them. Could have avoided them.

In fact, I was upset with the error messages and I followed up with a number of stability tests for my comp and for reproducibility of the results -
1) I've checked that all the save files from which I recovered had iteration counters earlier than those in which errors were reported (both 'p' and 'q' files - but I've renamed 'q' files into 'p' and restarted, to be more sure. Now I would have used my InterimFiles things, but then I didn't have that option set.)
2) I reran two of the siblings (they ran on neigboring threads on the quad) of this candidate on another computer - one had 00000400 code (also 4 bangs in a row exactly at the same time), another had no errors. The residues matched.
3) the reason of that glitch was also clear, it was a one-off thing. Crap happens. We live, we learn.
4) users who never set "round off error check = ON" will have a lot of no-error results, but I did set it ...et non, je ne regrette rien.

So I did some double-checks, but entirely for my own satisfaction. The only real credit is finding the prime (well, and not missing one, true). I am now confident that my residues have as good as anybody's chance to be fine - they are restarted from before the errors and they matched the alternate runs. And learned a few safety tricks like InterimFiles=1000000 which I wouldn't have otherwise. heck! I ran out of coffee! --end of a diatribe.

[markr]

I thought the question a reasonable one. Error codes 0000cc00 with cc>0 are ominous, with many bad results in the past. (As an aside, any thoughts on how well historical data will agree with current results, with changing program versions etc etc?) On its own - without the extra information showing the great care taken with this particular result - that error code would be grounds for considering a re-run as a "first-time" test. For example, 33604247. The non-zero error codes really stand out in these reports!

[Batalov]

That's why I wholeheartedly asked him to run it! Seriously. But his game is different, isn't it?

From you I can take criticism, absolutely seriously, and it was for the serious readers of this thread what I wrote above. But from the new candidat for M45 numerologist I will only take this first time-check for an answer. (I will gladly take this first time-check from anyone else willing to do it, as well.) What if the residues match? It's no use for me to run it. Or is there? look at this - http://v5www.mersenne.org/report_ll/?exp_lo=33904823

"Those who do look for the means, those who don't provide excuses". Those who never did anything will not get in the "bad.zip" file, too, guaranteed.

[markr]

Fair enough! I have no disagreement.

You can double-check one of your own assignments, but it's not the best way. Better for the double-check to be done on a different machine, and that is more readily seen to be done if it's a different user. The results for 33904823 count as verified since they have different random shifts:

Originally Posted by the latest lucas_v.txt

33904823,freephoenix,dell19,WZ1,7802CE096233D894,12908069,00000000
33904823,freephoenix,dell19,WZ1,7802CE096233D894,2758752,00000000

Originally Posted by Batalov

Those who never did anything will not get in the "bad.zip" file, too, guaranteed.

I have three entries in bad.txt - does that count? Actually, one thing that does make me is my bad/verified ratio < 0.5%.

[davieddy]

Originally Posted by Batalov

What if the residues match? It's no use for me to run it. Or is there? look at this - http://v5www.mersenne.org/report_ll/?exp_lo=33904823

If different shifts are used, I think we can accept a doublecheck
on the same computer.

[mdettweiler]

Originally Posted by davieddy

If different shifts are used, I think we can accept a doublecheck
on the same computer.

But what if that computer is unstable? Would different shifts make help with that?

[tha]

Originally Posted by Anonymous

But what if that computer is unstable? Would different shifts make help with that?

Yes.

[davieddy]

Originally Posted by Anonymous

But what if that computer is unstable? Would different shifts make help with that?

If the problem is "instability", it would be a mammoth fluke
to duplicate identical erroneous residues even with the same
shift.
The shift deals with the case of a repeatable error (eg roundoff)
which might well get duplicated on a different computer.

Using a different computer and user addresses the problem
(hopefully not widespread) of fraud and conspiracy.

[mdettweiler]

Originally Posted by davieddy

If the problem is "instability", it would be a mammoth fluke
to duplicate identical erroneous residues even with the same
shift.
The shift deals with the case of a repeatable error (eg roundoff)
which might well get duplicated on a different computer.

Using a different computer and user addresses the problem
(hopefully not widespread) of fraud and conspiracy.

Oh, I see. Okay, I guess as long as unstable machines will never produce the same bad residual twice (or at least very rarely do so), then that would be fine--I didn't know that they worked like that.

[JHagerson]

I took 33801673. ETC 15 June.