mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   FactorDB (https://www.mersenneforum.org/forumdisplay.php?f=94)
-   -   The "Hey YOU" Thread (https://www.mersenneforum.org/showthread.php?t=18489)

EdH 2016-10-26 01:28

[QUOTE=henryzz;445767]My work is submitted. Doesn't seem to have brought it below 550k numbers at 309 digits.[/QUOTE]
The spike has shown no noticeable decrease. I wonder if the graph is current and if the labels are true. Now, my curiosity is piqued...

chris2be8 2016-10-26 15:43

It's brought some numbers I hadn't checked into the accessible range, so I'll check them for algebraic factors. Thanks for the help.

Chris

EdH 2016-10-26 22:15

In my earlier limited view of the last visible 309 digit composites, it appeared that they were provided in index order. However, a more broad inspection has shown that the case, in fact, is different. This makes my earlier script even less valuable...

(I know, some of you are asking, "How could it possibly be less valuable than it was?":smile:

chris2be8 2017-07-17 16:20

[QUOTE=chris2be8;441740] I'll now start running against composites with no known factors. Which will take a few months. But should have more numbers end up fully factored, at least to start with.
[/QUOTE]

Finished at 60,000 digits (there are not many composites with no known factors in the range where status unknown isn't checked).

The final tally:
52363400 checked.
1809152 at least partly factored.
127059 fully factored.

And there will be some more where I lost the response from factordb. And some fully factored by factordb after I reduced them to under 70 digits.

I was originally planning to run it once a year. But since it took a year to run I'll wait a few weeks before restarting it against numbers with status unknown.

Chris

swellman 2017-08-18 00:08

U208052 is found to be C - now what?
 
I searched the forum for an answer to this, but could not find anything. Apologies if this question has been dealt with before.

While exploring the [url=http://home.earthlink.net/~elevensmooth/Progress.html]ElevenSmooth site[/url], we were inspired to plug the full M3326400 into factordb for grins, and found the largest factor is listed as [url=http://factordb.com/index.php?id=1100000000495254317]U208052[/url]. Ryan Propper ran 50 Miller-Rabin tests on it (using his own code), and all came back "not a prime". While this result certainly isn't shocking, it would settle the question of whether any further work is warranted on this particular cofactor. And no, no one is going to do much with a composite so large. But the issue is still open.

Is there a process to request to convert a U to a C without actually finding a factor? Presumably someone must verify compositeness on a trusted system with vetted software, correct?

Or are it's factors already known, hidden in plain sight? I'm not ashamed to admit I don't fully understand all the intertwined branches of so large (and smooth!) a Mersennes number. Any advice would be greatly appreciated.

Has anyone heard from William of ElevenSmooth lately?

axn 2017-08-18 03:29

[QUOTE=swellman;465812]Ryan Propper ran 50 Miller-Rabin tests on it (using his own code), and all came back "not a prime". [/quote]
WTF? Why? 1 negative result is sufficient.

[QUOTE=swellman;465812]While this result certainly isn't shocking, it would settle the question of whether any further work is warranted on this particular cofactor.[/quote]
As such, there is no efficient way to ECM _just_ this cofactor. You can use P95 to ECM its parent (2^1663200+1)

[QUOTE=swellman;465812]Is there a process to request to convert a U to a C without actually finding a factor?[/quote]
Do you have a login for factordb? AFAIK, there used to be a feature for trusted users to manually change the status of a number (but it may have been removed). You could write to Syd. FTR, factordb will automatically do PRP test for numbers < 20k digits, and for numbers between 20k digits and x digits, users can "Assign" them for PRP test. I don't know what 'x' is, but obviously it is not big enough to cover this composite.

[QUOTE=swellman;465812]Or are it's factors already known, hidden in plain sight?[/QUOTE]
If someone knew additional factors of this number, presumably they would've reported it to factordb.

swellman 2017-08-18 10:54

[QUOTE=axn;465845]WTF? Why? 1 negative result is sufficient.[/quote]

50 instances launched in parallel. Saved days in the (very) off chance it was prime. Wasted 49 instances if it was composite. Flip a coin.

[quote]
As such, there is no efficient way to ECM _just_ this cofactor. You can use P95 to ECM its parent (2^1663200+1)[/quote]
Thanks. Never actually used P95 but maybe it's time.

[quote]
Do you have a login for factordb? AFAIK, there used to be a feature for trusted users to manually change the status of a number (but it may have been removed). You could write to Syd. FTR, factordb will automatically do PRP test for numbers < 20k digits, and for numbers between 20k digits and x digits, users can "Assign" them for PRP test. I don't know what 'x' is, but obviously it is not big enough to cover this composite.[/quote]

No, but I will pursue one. Is Syd the only person who can issue an ID? I sense he's pretty busy with RL these days.

[quote]
If someone knew additional factors of this number, presumably they would've reported it to factordb.[/QUOTE]

Agreed. I'm just unsure how the ElevenSmooth project was managed. There are about fifty composite cofactors (plus one unknown) listed in Factordb that do not appear on the 11S site. I'm assuming they're all interrelated, i.e. find a factor of a smaller composite on 11S and one of the bigger terms will also benefit when the factor is reported to fdb.

axn 2017-08-18 13:32

[QUOTE=swellman;465856]50 instances launched in parallel. Saved days in the (very) off chance it was prime. Wasted 49 instances if it was composite. Flip a coin.[/quote]
:rolleyes: To each their own, I say :smile:

[QUOTE=swellman;465856]No, but I will pursue one. Is Syd the only person who can issue an ID? I sense he's pretty busy with RL these days.[/quote]
Yes, Syd is the only admin. To get an id, you can use [url]http://www.factordb.com/login.php?register=1[/url]. But to get special privileges, you'll have to write to Syd.

[QUOTE=swellman;465856]Agreed. I'm just unsure how the ElevenSmooth project was managed. There are about fifty composite cofactors (plus one unknown) listed in Factordb that do not appear on the 11S site. I'm assuming they're all interrelated, i.e. find a factor of a smaller composite on 11S and one of the bigger terms will also benefit when the factor is reported to fdb.[/QUOTE]

The composite cofactors come from the algebraic factorization. As such there is no point in tracking them. Factordb allows it, so someone (possibly even william) must've entered them. And you're right, factordb stores them in such a way that factor reported for one number (say C) is automatically available for all numbers that have C as a composite cofactor.

schickel 2017-08-18 13:42

As of now, there is no interface for changing a number's status available anymore.

There was an option available to flip a number from C to P for a while, but bad things happened while it was there. Unfortunately Syd did not also provide an option to do the revse - flip a number from P to C - so it could be rechecked if a number was marked P in error.

swellman 2017-08-18 14:30

[QUOTE=axn;465860]
Yes, Syd is the only admin. To get an id, you can use [url]http://www.factordb.com/login.php?register=1[/url]. But to get special privileges, you'll have to write to Syd.

[/quote]

Thanks. From schickel's post after yours, it sounds like the special privileges thing has gone away but I then don't need/want it.

[quote]
The composite cofactors come from the algebraic factorization. As such there is no point in tracking them. Factordb allows it, so someone (possibly even william) must've entered them. And you're right, factordb stores them in such a way that factor reported for one number (say C) is automatically available for all numbers that have C as a composite cofactor.[/QUOTE]

Thanks. Will focus on the composites list on the 11S site, with reports of any factors found to be made to factordb of course.

swellman 2017-08-18 14:32

[QUOTE=schickel;465862]As of now, there is no interface for changing a number's status available anymore.

There was an option available to flip a number from C to P for a while, but bad things happened while it was there. Unfortunately Syd did not also provide an option to do the revse - flip a number from P to C - so it could be rechecked if a number was marked P in error.[/QUOTE]

No worries, we will have to do it the old fashioned way - split it with a hammer! A really big hammer...


All times are UTC. The time now is 05:48.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.