[Mission Accomplished] Let's finish primality verification through Mp#48*, M(57 885 161)

[kriesel]

Now 19 left, average delta increased, max delta is ~172.5K. Judging from progress per exponent in the past 3 days, the lowest and highest remaining will complete within a day or two, the rest will probably expire and be reissued, by 5 days from now; add 7 days for run on fast Cat0-qualified systems and we get a projection of 2021-10-03 for completion of verification to Mp#48*. The possibility of stall and expiration without poach on any one or more of the ~17 reissued could delay that. It would take more than a started-and-then-stalled 30 day expiration and a no-progress-reported 10 day expiration on the same reissued exponent, and no poaches, to delay to 2021-11-16, which is 44 days after 2021-10-03.
The alternate method is in close agreement with the 56M historical analogy method which now gives 2021-10-02.

From https://www.mersenne.org/report_milestones/

Progress toward next GIMPS milestones (last updated 2021-09-21 12:45:12 UTC, updates every 15 minutes)

• All exponents below 57 555 247 have been tested and verified.
• All exponents below 104 405 113 have been tested at least once.
• Countdown to first time checking all exponents below 105 million: 3
• Countdown to first time checking all exponents below 106 million: 265
• Countdown to verifying all tests below M(57 885 161): 19
• Countdown to verifying all tests below 58 million: 26
• Countdown to verifying all tests below 59 million: 2 354 (1 665 available)
• Countdown to verifying all tests below M(74 207 281): 271 961
• Countdown to verifying all tests below M(77 232 917): 327 512
• Countdown to verifying all tests below M(82 589 933): 426 215

Originally Posted by kriesel

or maybe reaching Mp#48* comes in the middle, only ~2.5 times longer than the previous experience.

That ~2.5 longer should have been (4.83+1)/2 ~2.9 times as long. Also nearly 5-fold longer should have been nearly 5-fold as long.

[greenskull]

Oh, no!..



For me, the question remains -- how assignments with long overdue deadlines will be processed.
Will they be forcibly taken away from the previous owners. Whether they will be processed from scratch or from the point where everything stopped.
If the latest date is October 10, then does this mean that the cherished date will not come earlier than this moment.

[Uncwilly]

Originally Posted by greenskull

Will they be forcibly taken away from the previous owners.

When they expire, they get reassigned to someone else. If the slowpoke turns in the result before the other person does, oh well.

Whether they will be processed from scratch or from the point where everything stopped.

PrimeNet does not receive the interim files. The data transfer and storage are not currently worth it. So, the person with the new assignment has to start afresh.

If the latest date is October 10, then does this mean that the cherished date will not come earlier than this moment.

The slowest assignment that expires furthest in the future will then have to be reassigned and then run by who ever gets it. If that is a person that takes 10 days to complete such an assignment normally and their machine goes down for a few days over a long weekend, it may be 15 days after the current latest expiry.

Or other things can happen....

[greenskull]

Ok. Thanx!
This means there are rather big chances that my prediction will come true.

[kriesel]

Originally Posted by greenskull

Oh, no!..

Attachment 25734

For me, the question remains -- how assignments with long overdue deadlines will be processed.
Will they be forcibly taken away from the previous owners. Whether they will be processed from scratch or from the point where everything stopped.
If the latest date is October 10, then does this mean that the cherished date will not come earlier than this moment.

Oh, yesss.
A little late to be learning the rules of operation for the system you've made predictions about.
Don't know where you get October 10 from. Estimated completion, for an exponent that will expire and get reassigned far sooner, so the current estimated completion is irrelevant, perhaps.
Run a line through the uppermost red x in your plot, at the slope of your red prediction line, and get ~Oct 5. Way outside +-25 days from Nov 16.

Originally Posted by greenskull

Ok. Thanx!
This means there are rather big chances that my prediction will come true.

Um no. It means there are 17 small chances, for which multiple things would have to go your way to produce unexpectedly long delay. Nov 16 is far enough away that multiple successive assignments on the same exponent would need to expire before completion, and at least one of them would need to start reporting progress before stalling. That, after assignments go only to Cat0 qualified systems known reliable at not having expirations for the past 4 months. Read and understand the Cat0 assignment requirements. Plus it also requires that no one poaches the stalled exponent(s).
An initial delay of a day or two or three from assignment to start of progress is normal in prime95 or mprime. See prime95's days of work to queue up setting in Options, Preferences.


FYI there is hardware that can do such LLDC or the equivalent PRP/GEC/proof-generation, in several HOURS.

2021-03-16 09:37:08 asr2/radeonvii0 57034589 LL        0 loaded: 0000000000000004
2021-03-16 09:37:51 asr2/radeonvii0 57034589 LL   100000   0.18%;  431 us/it; ETA 0d 06:49; 51ebf4cd88bbcb14
...
2021-03-16 16:27:55 asr2/radeonvii0 57034589 LL 57034587 100.00%;  453 us/it; ETA 0d 00:00; c6d59ecf237f149e
2021-03-16 16:27:55 asr2/radeonvii0 waiting for the Jacobi check to finish..
2021-03-16 16:28:10 asr2/radeonvii0 57034589 OK 57000000 (jacobi == -1)
2021-03-16 16:28:10 asr2/radeonvii0 {"status":"C", "exponent":"57034589", "worktype":"LL", "res64":"c6d59ecf237f149e", "fft-length":"3145728", "shift-count":"0", "program":{"name":"gpuowl", "version":"v6.11-364-g36f4e2a"}, "user":"kriesel", "computer":"asr2/radeonvii0", "aid":"74720917554D8F6A27FABDFF1149FFC1", "timestamp":"2021-03-16 21:28:10 UTC"}

Actual duration 16:28:10 - 09:37:08 = 6:51:02 (h:mm:ss)
Add ~2.7% or ~11 minutes for 57.75M: ~7 hours 2 minutes.

[retina]

Originally Posted by greenskull

This means there are rather big chances that my prediction will come true.

You can "help it along" towards reaching your predicted goal by grabbing the last reassigned exponent and running it slowly making your best effort to complete it quickly.

[greenskull]

Originally Posted by retina

You can "help it along" towards reaching your predicted goal by grabbing the last reassigned exponent and running it slowly making your best effort to complete it quickly.

I've already thought about it :)
But I am more of an observer than a participant. And I am committed to the principles of fair play and non-interference.

[greenskull]

Originally Posted by kriesel

A little late to be learning the rules of operation for the system you've made predictions about.

You are very strict with me.
The game doesn't end there.

And I am very good at long distances than at short ones.
My earliest forecast, which I gave here on April 4, 2019, regarding the crossing of this level, was for September 12, 2021 ± 45 days:
https://mersenneforum.org/showpost.p...5&postcount=12

And I think we can already start predicting the next level.
I suggested the method of scoring for predictions above:
https://www.mersenneforum.org/showpo...postcount=3442

[kriesel]

Originally Posted by retina

You can "help it along" towards reaching your predicted goal by grabbing the last reassigned exponent and running it slowly making your best effort to complete it quickly.

quickly meaning here, at a rate consistent with his latest prediction that was based on more application of statistics than understanding. Assuming he both has at least one cat0 rated system and is lucky enough to draw on it one of the 8 exponents that get reassigned in the next 3 days, that is.

[kriesel]

Originally Posted by greenskull

And I am very good at long distances than at short ones.
My earliest forecast, which I gave here on April 4, 2019, regarding the crossing of this level, was for September 12, 2021 ± 45 days:
https://mersenneforum.org/showpost.p...5&postcount=12

which states in part:

Originally Posted by greenskull

We could estimate that all tests below M57885161 will be completed on 12 September 2021 ±45 days.
https://ibb.co/b11f3tt

Assuming the validity of my hypothesis, it can be assumed that at least 3 new Mersenne Prime numbers will be open by this date.

If that "open" meant found, your hypothesis would seem to be contradicted by the facts. The number of new Mersenne prime discoveries disclosed as occurring between 2018 Dec 8 and today remains zero. (It's unclear what that hypothesis is, or where it's clearly stated, if at all.)

predicted at https://www.mersenneforum.org/showpo...postcount=3072

48 mio -- on 19 July 2019 ±45 days
49 mio -- on 14 October 2019 ±45 days
50 mio -- on 07 January 2020 ±45 days.

(Not sure why you're using "mio" there; Google translate with language detect converts that to "my" in English. Nor what your native language may be. Not English, seems apparent. 49M would be pretty standard forum usage.)

actual (delta days computed by https://www.timeanddate.com/date/duration.html)
48M 2019-09-05 +48 days (miss)
49M 2019-11-19 +36 (hit at +80% of error bar value)
50M 2020-02-23 +47 days (miss)

+-45 days is a 91 day wide target, rather broad as the side of a barn, for a process that TAKES ~91 days on average per million of milestone advance recently.
(# of milestones by year: 2020: 5; 2019: 3; 2018:4; 2017: 4; etc
https://www.mersenne.org/report_milestones/)

Note that the average error in your earlier predictions above was +43.7 days, quite close to the difference between our current projections for reaching Mp#48* of +46 days, and to the 45 day error bar value. IIRC you changed your estimation method since, and chose not to compensate for observed past consistent prediction bias demonstrated by 48M, 49M, 50M. The 2021-09-12 old predicted date for Mp#48* has already passed. It would likely work out thanks to the broad error bar of +-45 days, except for having replaced (repudiated) it repeatedly since with predictions months later in date predicted.

[kruoli]

Originally Posted by kriesel

Not sure why you're using "mio" there; Google translate with language detect converts that to "my" in English. Nor what your native language may be. Not English, seems apparent.

At least in German, this is an abbreviation of "million". But Google Translate is also correct if the Italian word is meant (at least if you checked for the word in isolation). When using the automatic language selection and "10 mio", it correctly translates as "10 millions" for me (example I used). Somehow it still insisted that the original language was Italian, so maybe they use that abbreviation, too?