mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Aliquot Sequences (https://www.mersenneforum.org/forumdisplay.php?f=90)
-   -   Subproject #11: 2^2 * 3 - Lose the '3' (https://www.mersenneforum.org/showthread.php?t=16075)

LaurV 2011-09-25 11:02

[QUOTE=schickel;272591]As promised, [URL="ftp://ftp.frontiernet.net/pub/users/aliquot/AllSeq.zip"]here[/URL] is a current status file for everything.....I left it sorted by sequence this time.[/QUOTE]

Ok, I dlded the list picked up 996666 (love the looking :P), fed it to yafu. The C110 at the end splits in:

[CODE]PRP18 = 508589975260964627
PRP92 = 80099422866747069882885138925003246019962693978819307076036046509103508938007134955902232051
[/CODE]

ecm curve 58/90, less then a minute. Maybe it helps a dime.. (I never was an Aliquot fan till now :D)

Mini-Geek 2011-09-25 12:00

[QUOTE=schickel;272673]Hmmm....I left the file sorted by size, thinking it might make it easier for people to pick one out. Should we resort the list by sequence?[/QUOTE]

I prefer by size, because that's what really makes a difference in trying to test a sequence, not where it started or how long it's been going on. I think the clustering of reservations near the smaller sequences can be mostly attributed to the fact that they're also at the top of the list. (in my case, I can tell you that's the case)
[QUOTE=LaurV;272687]Ok, I dlded the list picked up 996666 (love the looking :P), fed it to yafu. The C110 at the end splits in:

[CODE]PRP18 = 508589975260964627
PRP92 = 80099422866747069882885138925003246019962693978819307076036046509103508938007134955902232051
[/CODE]

ecm curve 58/90, less then a minute. Maybe it helps a dime.. (I never was an Aliquot fan till now :D)[/QUOTE]
Thanks, and welcome! Here are some links to get you started:
[url]http://www.mersenneforum.org/showthread.php?t=12140[/url]
[url]http://gilchrist.ca/jeff/factoring/index.html[/url]
[url]http://gilchrist.ca/jeff/factoring/nfs_beginners_guide.html[/url]
[url]http://www.mersennewiki.org/index.php/Aliquot_Sequences[/url]
And yes, that does help a bit. Hopefully anyone trying to pick up that number would run ECM on it first, but it's possible to assume ECM had already been run, which would have meant a lot of extra GNFS work that is completely unnecessary.
In case you're not familiar with it, we log all our work into the FactorDB. [URL="http://factordb.com/sequences.php?se=1&aq=996666"]Here is its entry for 996666[/URL]. It includes your factors for that C110.

bchaffin 2011-09-25 16:01

[QUOTE=Mini-Geek;272453]Are we deleting once reserved, or adding a reservation tag? For now, I'll add a reservation tag, as it's easier to delete than to recover the info.[/QUOTE]

I like this method, since it allows me to parse the top post so my workers can avoid the reserved sequences -- I just made this change.

As always, if there are any sequences you want me to stay away from which are not in the top post of either the main reservations thread or this thread, let me know. (I'm also still avoiding everything from project 400 until that subproject is declared officially done, though that's probably irrelevant since those sequences are typically way above the workers' threshold.)

LaurV 2011-09-26 02:05

[QUOTE=Mini-Geek;272689]
In case you're not familiar with it, we log all our work into the FactorDB. [URL="http://factordb.com/sequences.php?se=1&aq=996666"]Here is its entry for 996666[/URL]. It includes your factors for that C110.[/QUOTE]

Of course, I reported them :D And from FactorDB I took the C110 of the last term (the zip does not contains the terms, as you know, but only the heads of the sequences). Decided to take it one more step. Already few k relations from the 150 k needed (yafu siqs). I will check your links, thanks.

LaurV 2011-09-26 07:48

[QUOTE=schickel;272591]As promised, [URL="ftp://ftp.frontiernet.net/pub/users/aliquot/AllSeq.zip"]here[/URL] is a current status file for everything.....I left it sorted by sequence this time.[/QUOTE]

Now, that I stuck my ears into this domain and can't get them off, I think it would be better to have a "*Cxxx" at the end of each line of this file. To see at a glance how big is the cofactor. Of course we can guess it from the provided factors, but it is not "at a glance". Eventually, xxx padded to 3 characters (filled with 0 in front). This would make it easier for newcomers (like me) who want just to "try" few more steps, to sort it by the cofactor's size and pick a sequence with a convenient cofactor... My range is, say, below 110 digits, so I would like to pick a cofactor in this range and play with it few hours/days. When is factorized, I report the result, without going too deep into aliquotes. So, one can contribute from time to time when he has some spare core and the right mood.

Edit:
1. I am pushing C110 of 866th term of 996666 at least one step (I like this number, that's why, easy to remember)
2. I am pushing C96 of 2474th term of 256950 at least few steps (it is the smallest composite in the list, that's why)
in the next days. I hope I am not stepping on anybody's tail

schickel 2011-09-26 08:53

[QUOTE=LaurV;272749]Now, that I stuck my ears into this domain and can't get them off, I think it would be better to have a "*Cxxx" at the end of each line of this file. To see at a glance how big is the cofactor. Of course we can guess it from the provided factors, but it is not "at a glance". Eventually, xxx padded to 3 characters (filled with 0 in front). This would make it easier for newcomers (like me) who want just to "try" few more steps, to sort it by the cofactor's size and pick a sequence with a convenient cofactor... My range is, say, below 110 digits, so I would like to pick a cofactor in this range and play with it few hours/days. When is factorized, I report the result, without going too deep into aliquotes. So, one can contribute from time to time when he has some spare core and the right mood.[/quote]It would take some additional effort on my part. I use the "[URL="http://factordb.com/elf.php?seq=996666&type=1"]Download .elf file[/URL]" link to pull down the complete sequence. To get the size of the remaining composite, I would have to either divide out the small factors or do a search on the initial number on the line to parse out the size. If I get some time, I'll see what is going to be easier....[quote]Edit:
1. I am pushing C110 of 866th term of 996666 at least one step (I like this number, that's why, easy to remember)
2. I am pushing C96 of 2474th term of 256950 at least few steps (it is the smallest composite in the list, that's why)
in the next days. I hope I am not stepping on anybody's tail[/QUOTE]Just check the reservation [URL="http://www.mersenneforum.org/showthread.php?t=11588"]thread[/URL] to see what is officially reserved. If you want exclusivity on a sequence, post in the reservation thread, otherwise the DB workers or bchaffin's sequence workers will factor any number that is small enough (currently ~106 digits for bchaffin's workers, considerably smaller for the other DB workers).

Checking, right now 256950 is reserved by fivemack, and the sequence just finished a downdriver run. Since it has the 2^2 * 7 driver, it is going to increase until an escape occurs. It is very likely that he has local work that has not been uploaded to the DB.

996666 is currently unreserved, but is currently driven by 2^4 * 31, so it is also on an upward trend. Just run it until it's beyond your reach.....

If you want something a little more manageable, look for a smaller sequence with just [TEX]2^n[/TEX] as the leading factor, that way it's likely to decrease the next line....

LaurV 2011-09-26 10:19

[QUOTE=schickel;272754]
Checking, right now 256950 is reserved by fivemack, and the sequence just finished a downdriver run. Since it has the 2^2 * 7 driver, it is going to increase until an escape occurs. It is very likely that he has local work that has not been uploaded to the DB.
[/QUOTE]

C96 was almost finished when I read your post (about 85k relations from 97k needed) so I let it finish. It splits into:

PRP52 = 3723350663243243784127076674821693188932287078906061
PRP44 = 90710817240745853809351361240337236523389537

I am giving up on it, I don't want to step on fivemack's toes. As you said, most probably he has these factors too, on his local logs. One C96 is not big deal. (later edit: could not resist, that c72 looked juicy, so I did another 2 steps)

[QUOTE]
996666 is currently unreserved, but is currently driven by 2^4 * 31, so it is also on an upward trend. Just run it until it's beyond your reach.....
[/QUOTE]Ok, then I will try to get it on, some more steps.

[QUOTE]
If you want something a little more manageable, look for a smaller sequence with just [TEX]2^n[/TEX] as the leading factor, that way it's likely to decrease the next line....[/QUOTE]Thanks for the advice. I read some theory on the subject (drivers, guides, downdrivers, etc), but I still did not fully understood it, and I don't have the "feeling" of what is happening and why.

Mini-Geek 2011-09-26 22:18

[QUOTE=LaurV;272760]Thanks for the advice. I read some theory on the subject (drivers, guides, downdrivers, etc), but I still did not fully understood it, and I don't have the "feeling" of what is happening and why.[/QUOTE]

To help you get the "feel" of it: the sequences are calculated by summing the factors of a number. Lots of little prime factors make for lots of total factors, which make the sequence grow quickly. Few enough little factors means the sequence will drop. Certain groups of factors (especially perfect numbers, which (when even) always correspond to Mersenne numbers) tend to stick around; most of these make the sequence grow (i.e. drivers/guides), while others tend to make it smaller (downdriver, and 2^x, which I call "downguides", since they let the sequence drop if there aren't any other fairly small factors).
It may be informative to find a part of a sequence that has some ups and downs, and where the size is under 80 digits, and rerun it, watching the aliqueit output...growing with certain drivers, hovering around the same size, or dropping, and how they change.

Mini-Geek 2011-09-27 00:45

Releasing [URL="http://factordb.com/sequences.php?se=1&aq=247840"]247840[/URL]. It acquired the 2^2 * 7 driver and hit a c111 (full ECM) in one iteration.

bchaffin 2011-09-28 02:21

Releasing 243402, which is done for this project. It did manage to lose the 3 and even decreased for a little while, but then picked up 2^4 * 31.

I'll take 252330 next.

LaurV 2011-09-30 04:18

@schickel

Well, I took 996666 from T865 (T=term) to T880 (after a proper reservation, as you advised me before), it is a "good" sequence, never made problems and gave me to factor only composites below C102. Still going on it.

I use to report the factors as I find them. All is manual job, I don't like the fact that I can't (or I don't know how to) customize programs like aliqueit112.exe to launch the factoring program I want, with the parameters I want, so I do the job manually when I have time and a free core for a couple of hours). This is in fact good, as someone else (the server, yoyo's boincs) could find factors for the small composites (when and if they appear) faster that I can do, but is bad because they step on my fingers and we do the same job twice.

That is why I posted, I just want to let you know FYI that the fact you mentioned, about the factorDB not lending out composites from proper reserved sequences, is not quite true. As I expected, last evening someone (maybe one of yoyo's boincs) factored the C82 of T879 and reported the result much faster then I could :smile:. No big deal, if the sequence is making steps, no matter who pushed it forward. I will jostle it when is floundered.:smile:


All times are UTC. The time now is 16:08.

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