mersenne.ca

[LaurV]

Originally Posted by James Heinrich

bad sectors on the HDD

Grrr... man, don't use that HDD anymore! Just scrap it and you will save yourself a lot of headache in the future. Trust me, I destroyed more than 14 HDDs in my "career", only in computers that I owned and used at home (including thunderstrucked HDD, I think I told this story).
I could re-download all the important data for "just in case", if you resend me the link (last sync was sometimes mid of the last year, like 3 or 4 GB all the factors to 10G, iirc). I will also contribute a little bit toward you buying a new HDD (check with Mike).

[James Heinrich]

Originally Posted by LaurV

Grrr... man, don't use that HDD anymore!

I guess I should have been clearer, to me it was obvious that that HDD's useful life is over. I had actually already been considering replacing that drive with a newer/larger one for space reasons, so as soon as I found a problem I clicked "order". The replacement should be here in a few days.

To be clear, this is a HDD in my home system (not the server), and all the data is replicated on a second machine so there is little danger of data loss. The interesting bits of data are packaged up for export every Sunday morning on the export page (I haven't packaged up the 2020 data yet but will soon).
edit: 2020 archives are now finalized

[petrw1]

https://www.mersenne.ca/exponent/41771999

[SethTro]

I was looking at https://www.mersenne.ca/tjaoi.php

If it's not a hard query it would be interesting to know how many number (maybe even in each bit range) where TJAOI has found the only factor.
If that's easy then additionally GHz saved over all exponents where TJAOI has found the singular factor.

[James Heinrich]

Originally Posted by SethTro

If it's not a hard query it would be interesting to know how many number (maybe even in each bit range) where TJAOI has found the only factor.

It's possible, but it's not a trivial query.

[LaurV]

Last similar "argument" raised by me some time ago had the result, someone posted a list of 5 (?) such factors, from which, after some investigation, only 1 or 2 were "new TF", the rest were found by ECM or were not "first factor", or were not "new". He found a lot of "second" and "third" and "so on" factors, I mean A LOT, but didn't factor many exponents which had no known factor before. I mean, by TF. As he worked the lower bitlevels "by q", or "by k" (it is still unsettled how he did it, but it is clear he didn't do it "by p"), any "new" factor he would have found below 66 bits (he surpassed the 65 bits boundary, I think, and still raising slowly) would mean a factor "missed" by GIMPS. There are not many of those.

[James Heinrich]

Originally Posted by LaurV

someone posted a list of 5 (?) such factors, from which, after some investigation, only 1 or 2 were "new TF", the rest were found by ECM or were not "first factor", or were not "new".

It's possible my query is faulty, but I was not able to find any exponents for which TJAOI has the only known factor.

SELECT *, COUNT(*) AS `howmany` FROM `prime_numbers_0` WHERE (`exponent` IN ( SELECT `exponent` FROM `primenet_results_archive_2013` JOIN `primenet_results_archive_users` USING (`primenet_results_user_id`) WHERE (`primenet_results_archive_users`.`user_name` = "TJAOI"))) GROUP BY `exponent` ASC HAVING (`howmany` > 1);

Finding exponents where a specific user found the first known factor is a little more complex, I may come back and look at that again later.

[Viliam Furik]

Originally Posted by James Heinrich

It's possible my query is faulty, but I was not able to find any exponents for which TJAOI has the only known factor.

Not possible, it's sure. A recent example is 5055943. Occasionally he turns in a few exponents around 2M-5M with new single factors found by TF.

By my guesstimates, there should be roughly 500 such exponents.

[James Heinrich]

Originally Posted by Viliam Furik

Not possible, it's sure. A recent example is 5055943

Thanks. My query was broken in two different ways
Now it returns more sane values

SELECT `exponent`, `timestamp_found` FROM `known_factors_0` WHERE (`exponent` IN (SELECT `exponent` FROM `primenet_results_archive_2021` JOIN `primenet_results_archive_users` USING (`primenet_results_user_id`) WHERE (`result_type` = "F") AND (`primenet_results_archive_users`.`user_name` = "TJAOI"))) GROUP BY `exponent` ASC HAVING (COUNT(*) = 1);

Count of exponents by year where TJAOI has found the only known factor:

2021: 13
2020: 519
2019: 143
2018: 121
2017: 68
2016: 113
2015: 217
2014: 211
2013: 5

The 1410 exponents in question are attached, sorted by date.

Note that this is not quite the same question as looking for where TJAOI found the first known factor and someone else may have subsequently found another factor (or, indeed, TJAOI found a second factor for the same exponent). I have not looked for these but I suspect there won't be many.

[petrw1]

I found the source code quite easily about 3 years ago but either, I've gone brain-dead or it's not as accessible as it used to be.

I believe the formulas have changed because the percentages are not the same as they used to be.

I'd like to try to translate the basic calculation into an Excel macro to help me determine the best bounds to use for large blocks of P-1 assignments.

Thanks

[James Heinrich]

The P-1 probability code was re-designed by Mihai Preda in 2020-Aug.
His original code was, at the time, posted on https://github.com/preda/gpuowl/tree/master/pm1/ but I no longer see the file there that I used. I think I remember something about Mihai including a PM1 probability calculator utility with gpuowl so you might want to look into that (I could be mistaken).

I have attached my PHP version of Mihai's code, I'm not sure how easily it translates to Excel.