mersenneforum.org  

Go Back   mersenneforum.org > Factoring Projects > Cunningham Tables

Reply
 
Thread Tools
Old 2021-11-26, 00:56   #1
charybdis
 
charybdis's Avatar
 
Apr 2020

547 Posts
Default Polynomial selection for 2,2246M c221

This is one of only two remaining unambiguous GNFS numbers from the 1987 edition of the Cunningham book (base 2 <1200 bits), the other being 2,1109+ c225.

Bob Silverman suggested that the forum run polynomial selection for this number, and Greg has agreed to sieve it with NFS@Home once a suitable poly is found.

The composite is
Code:
24023387191766184217094927222587025087468868508838209688355658860897738047963529318736265313199625192712082581481439831563094403250831892547179972207284600405234081309183045488404642574713121441527058807512178264038863337
This will be a degree-6 search. The obvious target for a good score is Gimarel's score of 3.36e-16 for 3,697+, which to the best of my knowledge is the only GNFS-221 that has ever been run before. The c220 record is 4.178e-16, so it's reasonable to believe that 3.5e-16 or better is possible.

For those who prefer to use CADO, the parameters used to find the aforementioned c220 record polynomial might be a useful starting point: these are P=12M, incr=420, nq=46656, sopteffort=10, ropteffort=100. I am not claiming that these are optimal. The (pre-sizeopt) leading coefficient was in the 35M-40M range; unlike with degree-5, there did not appear to be a significant decrease in average polynomial quality with larger leading coefficients, so there was no need to move to higher values of incr.

NFS@Home have a number of SNFS jobs queued up, so don't worry if you've been sieving 3,748+ for the last month and would like a break: there's no rush. It'll probably be about a week before I start any searching myself.
charybdis is offline   Reply With Quote
Old 2021-11-26, 22:46   #2
swellman
 
swellman's Avatar
 
Jun 2012

24·7·29 Posts
Default

Sounds fun. Love your recipe for CADO, though I may increase ropteffort and sopteffort a bit.

One question - is tasks.wutimeout = 24000 (default) sufficient?

Reserving 1e3 < c6 < 2e6.

ETA: adrange = 1680

Last fiddled with by swellman on 2021-11-27 at 00:55 Reason: Add adrange
swellman is online now   Reply With Quote
Old 2021-11-27, 02:01   #3
charybdis
 
charybdis's Avatar
 
Apr 2020

22316 Posts
Default

Quote:
Originally Posted by swellman View Post
One question - is tasks.wutimeout = 24000 (default) sufficient?
Depends if you set ropteffort very high. With ropteffort=100 you'll be fine. 200 might potentially need more time on old/slow CPUs.
Your adrange is small enough that the length of the stage-1+sizeopt tasks won't be an issue.
charybdis is offline   Reply With Quote
Old 2021-11-27, 02:26   #4
swellman
 
swellman's Avatar
 
Jun 2012

24·7·29 Posts
Default

FWIW, msieve is looking for an e-score of at least 2.71e-16 to > 3.12e-16. I’m running some msieve-GPU for c6 < 1M. Not expecting much from msieve on a C221 but blind squirrels etc.

CADO is running on one of my old quad core laptops. It is SLOW but fingers crossed a flare pops up.

Thank you for the advice @charybdis!
swellman is online now   Reply With Quote
Old 2021-11-27, 06:38   #5
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

11·461 Posts
Default

Quote:
Originally Posted by charybdis View Post
For those who prefer to use CADO, the parameters used to find the aforementioned c220 record polynomial might be a useful starting point: these are P=12M, incr=420, nq=46656, sopteffort=10, ropteffort=100. I am not claiming that these are optimal. The (pre-sizeopt) leading coefficient was in the 35M-40M range; unlike with degree-5, there did not appear to be a significant decrease in average polynomial quality with larger leading coefficients, so there was no need to move to higher values of incr.
If there is little dropoff in score for larger leading coeff's, doesn't that argue for incr-4620 rather than 420? That is, we should take advantage of the slightly better typical poly that an extra small-factor of c6 provides, since the downside of searching large c6 range doesn't seem to be a problem?
VBCurtis is offline   Reply With Quote
Old 2021-11-27, 14:38   #6
charybdis
 
charybdis's Avatar
 
Apr 2020

547 Posts
Default

Quote:
Originally Posted by VBCurtis View Post
If there is little dropoff in score for larger leading coeff's, doesn't that argue for incr-4620 rather than 420? That is, we should take advantage of the slightly better typical poly that an extra small-factor of c6 provides, since the downside of searching large c6 range doesn't seem to be a problem?
Perhaps. While I didn't see much dropoff up to 40M, that doesn't mean there won't be by 400M. I suppose the prudent thing to do would be to run smallish ranges (100k?) at 40M and 400M and compare the "exp_E" scores of the polynomials at the bottom of the priority queue from each run.
charybdis is offline   Reply With Quote
Old 2021-11-29, 08:14   #7
Gimarel
 
Apr 2010

22·47 Posts
Default

3.5e-16 is possible:
Code:
# norm 7.406551e-16 alpha -9.691472 e 3.501e-16 rroots 6
skew: 7000972.35
c0: 53635249950771487474963143079350712070525911980
c1: 34662949307835344131518242206719949955943
c2: -72498102600242091149685770153699749
c3: 3018341741417463336008367359
c4: 7813123654476017869721
c5: 45283110042666
c6: -25200
Y0: -1113552721024727138053628986496268142
Y1: 1434019800661087662503
Gimarel is offline   Reply With Quote
Old 2021-11-29, 21:06   #8
swellman
 
swellman's Avatar
 
Jun 2012

62608 Posts
Default



Very nice!

A new record for C221, but can an even higher score be found? 3.7-3.8e-16 perhaps?

I’ll keep searching but you’ve set the high water mark!
swellman is online now   Reply With Quote
Old 2021-11-29, 23:13   #9
charybdis
 
charybdis's Avatar
 
Apr 2020

10001000112 Posts
Default

Quote:
Originally Posted by Gimarel View Post
3.5e-16 is possible:
Very nice!

This is good enough that I'll stay away for longer than I was planning; there are lots of new base-2 Cunninghams to do...
charybdis is offline   Reply With Quote
Old 2021-11-29, 23:43   #10
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

10011110011112 Posts
Default

Would Greg like some test-sieving done?
I can try to quantify the tradeoff between 34LP and 34/35 for this job, for instance.
This poly is "good enough", but this is a very big job and saving even 1-2% of sieve time is quite a few workunits from the elves; a 2% savings is roughly a C190 sieve job in time saved! So, we should continue to look for a 3.6.
There's some chance of "spin" still, too.

Edit: I'm busy with ECM pretesting a couple jobs for 15e queue to meet the upcoming "challenge" demand, I'll work on CADO select on this job in a couple weeks once I get two jobs posted to 15e.

Last fiddled with by VBCurtis on 2021-11-29 at 23:45
VBCurtis is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Polynomial selection Max0526 NFS@Home 9 2017-05-20 08:57
Best way to scale polynomial selection pastcow Msieve 6 2013-05-08 09:01
Updated polynomial selection jasonp Msieve 65 2011-05-01 19:06
2^877-1 polynomial selection fivemack Factoring 47 2009-06-16 00:24
Polynomial selection CRGreathouse Factoring 2 2009-05-25 07:55

All times are UTC. The time now is 17:34.


Sat Dec 4 17:34:16 UTC 2021 up 134 days, 12:03, 1 user, load averages: 0.79, 1.11, 1.22

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

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.