mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Aliquot Sequences (https://www.mersenneforum.org/forumdisplay.php?f=90)
-   -   Team sieve #2: c136 from 4788:2381 (https://www.mersenneforum.org/showthread.php?t=11690)

jrk 2009-04-23 19:20

[b]Reserving 11.7M to 12.4M[/b].

mdettweiler 2009-04-23 20:00

[quote=Batalov;170730]What is important is how many unique relns you had at this point.
(Check the log. Then you may be able to make a better estimate.)
For a number of this size, you'd need at least 20-22M unique relns.

[I]filtering wants 276621 more relations[/I]
is not [I]always[/I] accurate.[/quote]
Okay, here you go:
[I]found 3904684 duplicates and 17754868 unique relations[/I]
In that case, you guys may as well load up 2 or 3 million more (total) relations. Looks like we'll need it. :smile:

Regarding "filtering wants more relations" not being accurate: yes, I've noticed that too. Seems that it's impossible to *really* know how many more relations are needed except by trial and error, and by rough estimates based on # of unique relations needed by similar jobs.

jasonp 2009-04-24 03:24

[QUOTE=mdettweiler;170737]
Regarding "filtering wants more relations" not being accurate: yes, I've noticed that too. Seems that it's impossible to *really* know how many more relations are needed except by trial and error, and by rough estimates based on # of unique relations needed by similar jobs.[/QUOTE]
Willemien Ekkelkamp is working on [url="http://www.springerlink.com/content/747k17x78rt63341/"]a method[/url] that can extrapolate the number of relations needed based on a little test sieving, and for line sieving her results are a very accurate approximation to the real world (within 1-2%). I think some folks at EPFL are trying to improve the technique to work with lattice sieving.

But for now we only have the lazy way, that relies on history.

jrk 2009-04-24 03:25

I'll have almost 1.6M relations done in about 90 minutes. Will this be enough for another filtering run or should I reserve another range?

mdettweiler 2009-04-24 04:38

[quote=jrk;170766]I'll have almost 1.6M relations done in about 90 minutes. Will this be enough for another filtering run or should I reserve another range?[/quote]
Probably won't hurt to do another range. I won't be able to start the postprocessing until tomorrow morning (4/24 EDT) anyway, so even if we would have enough otherwise, the extra relations would help produce a smaller matrix.

Batalov 2009-04-24 04:59

[quote=jasonp;170765]Willemien Ekkelkamp is working on [URL="http://www.springerlink.com/content/747k17x78rt63341/"]a method[/URL] that can extrapolate the number of relations needed based on a little test sieving, and for line sieving her results are a very accurate approximation to the real world (within 1-2%)...[/quote]
For a moment I felt that you were pulling our (collective) leg(s).

It sounded like Eddie Izzard's sketch about made-up names...
"-- What shall we call our son so he does not get the sh!t kicked out of him at school? We shall call him [I]Englebert Humperdinck[/I]! Yes, that'll work."

Yes, I know, that's a real name. Sorry!

jrk 2009-04-24 05:12

11.7M to 12.4M done. 1581432 relations. Uploaded: [URL="http://jaysonking.com/aq/4788-2381-c136-11.7M_to_12.4M.out.bz2"]11.7M_to_12.4M[/URL] (64.3 MB)

[B]Reserving 12.4M to 12.8M[/B].

jrk 2009-04-24 10:45

12.4M to 12.8M done. 897146 relations. Uploaded: [URL="http://jaysonking.com/aq/4788-2381-c136-12.4M_to_12.8M.out.bz2"]12.4M_to_12.8M[/URL] (36.5 MB)

jasonp 2009-04-24 13:31

[QUOTE=Batalov;170774]For a moment I felt that you were pulling our (collective) leg(s).
[/QUOTE]
By publishing in Lecture Notes in Computer Science using an assumed name? That's a rather elaborate joke...

jrk 2009-04-26 00:43

Is the postprocessing working now?

mdettweiler 2009-04-26 04:14

[quote=jrk;170980]Is the postprocessing working now?[/quote]
Oops! Sorry about that--I'd apparently forgotten to download the last two relations files and thus hadn't started yet. My bad. :rolleyes:

Anyway, both files are downloaded and processed now. I'll get the postprocessing started (for real :grin:) sometime tomorrow morning/afternoon (4/26 EDT).

* 11.7M-12.4M jrk (done, 1581432 rels)
* 12.4M-12.8M jrk (done, 897146 rels)
Total relations received: 26277524 (131.4% of 20M rels)


All times are UTC. The time now is 09:33.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.